{"repo_id":"LLaMA-Adapter","entity_id":"py:setup","uri":"program://LLaMA-Adapter/module/setup#L1-L6","kind":"module","name":"setup","path":"setup.py","language":"python","start_line":1,"end_line":6,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom setuptools import find_packages, setup\n\nsetup(name=\"llama\", version=\"0.0.0\", packages=find_packages())","source_hash":"2ef5fd81323c276fca0f7106c2f0d98d12499ec49e5b7ee8de222a315e9153d7","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:example","uri":"program://LLaMA-Adapter/module/example#L1-L117","kind":"module","name":"example","path":"example.py","language":"python","start_line":1,"end_line":117,"context_start_line":1,"context_end_line":117,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport json\nimport os\nimport sys\nimport time\nfrom pathlib import Path\nfrom typing import Tuple\n\nimport fire\nimport torch\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import LLaMA, ModelArgs, Tokenizer, Transformer\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n    quantizer: bool=False,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")\n    adapter_checkpoint = torch.load(adapter_path, map_location=\"cpu\")\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(max_seq_len=max_seq_len, max_batch_size=max_batch_size, quantizer=quantizer, **params)\n    model_args.adapter_layer = int(adapter_checkpoint[\"adapter_query.weight\"].shape[0] / model_args.adapter_len)\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    print(model)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    model.load_state_dict(checkpoint, strict=False)\n    model.load_state_dict(adapter_checkpoint, strict=False)\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    temperature: float = 0.1,\n    top_p: float = 0.75,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n    quantizer: bool = False,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    generator = load(ckpt_dir, tokenizer_path, adapter_path, local_rank, world_size, max_seq_len, max_batch_size, quantizer)\n    instructs = [\n        \"Tell me about alpacas.\",\n        \"Tell me about the president of Mexico in 2019.\",\n        \"Tell me about the king of France in 2019.\",\n        \"List all Canadian provinces in alphabetical order.\",\n        \"Write a Python program that prints the first 10 Fibonacci numbers.\",\n        \"Write a program that prints the numbers from 1 to 100. But for multiples of three print 'Fizz' instead of the number and for the multiples of five print 'Buzz'. For numbers which are multiples of both three and five print 'FizzBuzz'.\",  # noqa: E501\n        \"Tell me five words that rhyme with 'shock'.\",\n        \"Translate the sentence 'I have no mouth but I must scream' into Spanish.\",\n        \"Count up from 1 to 500.\",\n    ]\n    prompts = [PROMPT_DICT[\"prompt_no_input\"].format_map({\"instruction\": x, \"input\": \"\"}) for x in instructs]\n\n    results = generator.generate(prompts, max_gen_len=512, temperature=temperature, top_p=top_p)\n\n    for result in results:\n        print(result)\n        print(\"\\n==================================\\n\")\n\n\nif __name__ == \"__main__\":\n    fire.Fire(main)","source_hash":"6c23b7ee72462699c540418da3efb30bfb4b4dcc5d82e513cc14c99635883587","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:example.setup_model_parallel","uri":"program://LLaMA-Adapter/function/example.setup_model_parallel#L31-L41","kind":"function","name":"setup_model_parallel","path":"example.py","language":"python","start_line":31,"end_line":41,"context_start_line":11,"context_end_line":61,"code":"import fire\nimport torch\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import LLaMA, ModelArgs, Tokenizer, Transformer\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n    quantizer: bool=False,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")","source_hash":"6c23b7ee72462699c540418da3efb30bfb4b4dcc5d82e513cc14c99635883587","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:example.load","uri":"program://LLaMA-Adapter/function/example.load#L44-L78","kind":"function","name":"load","path":"example.py","language":"python","start_line":44,"end_line":78,"context_start_line":24,"context_end_line":98,"code":"        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n    quantizer: bool=False,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")\n    adapter_checkpoint = torch.load(adapter_path, map_location=\"cpu\")\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(max_seq_len=max_seq_len, max_batch_size=max_batch_size, quantizer=quantizer, **params)\n    model_args.adapter_layer = int(adapter_checkpoint[\"adapter_query.weight\"].shape[0] / model_args.adapter_len)\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    print(model)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    model.load_state_dict(checkpoint, strict=False)\n    model.load_state_dict(adapter_checkpoint, strict=False)\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    temperature: float = 0.1,\n    top_p: float = 0.75,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n    quantizer: bool = False,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    generator = load(ckpt_dir, tokenizer_path, adapter_path, local_rank, world_size, max_seq_len, max_batch_size, quantizer)\n    instructs = [\n        \"Tell me about alpacas.\",\n        \"Tell me about the president of Mexico in 2019.\",","source_hash":"6c23b7ee72462699c540418da3efb30bfb4b4dcc5d82e513cc14c99635883587","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:example.main","uri":"program://LLaMA-Adapter/function/example.main#L81-L113","kind":"function","name":"main","path":"example.py","language":"python","start_line":81,"end_line":113,"context_start_line":61,"context_end_line":117,"code":"    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")\n    adapter_checkpoint = torch.load(adapter_path, map_location=\"cpu\")\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(max_seq_len=max_seq_len, max_batch_size=max_batch_size, quantizer=quantizer, **params)\n    model_args.adapter_layer = int(adapter_checkpoint[\"adapter_query.weight\"].shape[0] / model_args.adapter_len)\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    print(model)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    model.load_state_dict(checkpoint, strict=False)\n    model.load_state_dict(adapter_checkpoint, strict=False)\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    temperature: float = 0.1,\n    top_p: float = 0.75,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n    quantizer: bool = False,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    generator = load(ckpt_dir, tokenizer_path, adapter_path, local_rank, world_size, max_seq_len, max_batch_size, quantizer)\n    instructs = [\n        \"Tell me about alpacas.\",\n        \"Tell me about the president of Mexico in 2019.\",\n        \"Tell me about the king of France in 2019.\",\n        \"List all Canadian provinces in alphabetical order.\",\n        \"Write a Python program that prints the first 10 Fibonacci numbers.\",\n        \"Write a program that prints the numbers from 1 to 100. But for multiples of three print 'Fizz' instead of the number and for the multiples of five print 'Buzz'. For numbers which are multiples of both three and five print 'FizzBuzz'.\",  # noqa: E501\n        \"Tell me five words that rhyme with 'shock'.\",\n        \"Translate the sentence 'I have no mouth but I must scream' into Spanish.\",\n        \"Count up from 1 to 500.\",\n    ]\n    prompts = [PROMPT_DICT[\"prompt_no_input\"].format_map({\"instruction\": x, \"input\": \"\"}) for x in instructs]\n\n    results = generator.generate(prompts, max_gen_len=512, temperature=temperature, top_p=top_p)\n\n    for result in results:\n        print(result)\n        print(\"\\n==================================\\n\")\n\n\nif __name__ == \"__main__\":\n    fire.Fire(main)","source_hash":"6c23b7ee72462699c540418da3efb30bfb4b4dcc5d82e513cc14c99635883587","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:utils.quantization","uri":"program://LLaMA-Adapter/module/utils.quantization#L1-L49","kind":"module","name":"utils.quantization","path":"utils/quantization.py","language":"python","start_line":1,"end_line":49,"context_start_line":1,"context_end_line":49,"code":"import torch\nimport bitsandbytes as bnb\n\n\n'''\nlit-llama\n'''\nclass Linear8bitLt(bnb.nn.Linear8bitLt):\n    \"\"\"Wraps `bnb.nn.Linear8bitLt` and enables instantiation directly on the device and\n    re-quantizaton when loading the state dict.\n\n\n    \"\"\"\n\n    def __init__(self, *args, **kwargs):\n        super().__init__(*args, **kwargs, has_fp16_weights=False, threshold=6.0)\n\n        # We quantize the initial weight here so we don't end up filling the device\n        # memory with float32 weights which could lead to OOM.\n        self._quantize_weight(self.weight.data)\n\n    def _load_from_state_dict(self, local_state_dict, *args, **kwargs):\n\n        # There is only one key that ends with `*.weight`, the other one is the bias\n        weight_key = next(\n            (name for name in local_state_dict.keys() if name.endswith(\"weight\")),\n            None,\n        )\n        if weight_key is None:\n            return\n\n        # Load the weight from the state dict and re-quantize it\n        weight = local_state_dict.pop(weight_key)\n        self._quantize_weight(weight)\n\n        # If there is a bias, let nn.Module load it\n        if local_state_dict:\n            super()._load_from_state_dict(local_state_dict, *args, **kwargs)\n\n    def _quantize_weight(self, weight: torch.Tensor) -> None:\n\n        # This code is taken and adapted from `bnb.nn.Int8Params.cuda()`\n        B = weight.contiguous().half().cuda()\n        CB, CBt, SCB, SCBt, coo_tensorB = bnb.functional.double_quant(B)\n        del CBt\n        del SCBt\n        self.weight.data = CB\n        setattr(self.weight, \"CB\", CB)\n        setattr(self.weight, \"SCB\", SCB)","source_hash":"8936905ab12763c28c0324a49457f6680d9ed1c568a746d284ca64bf506a2f60","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:utils.quantization.Linear8bitLt","uri":"program://LLaMA-Adapter/class/utils.quantization.Linear8bitLt#L8-L49","kind":"class","name":"Linear8bitLt","path":"utils/quantization.py","language":"python","start_line":8,"end_line":49,"context_start_line":1,"context_end_line":49,"code":"import torch\nimport bitsandbytes as bnb\n\n\n'''\nlit-llama\n'''\nclass Linear8bitLt(bnb.nn.Linear8bitLt):\n    \"\"\"Wraps `bnb.nn.Linear8bitLt` and enables instantiation directly on the device and\n    re-quantizaton when loading the state dict.\n\n\n    \"\"\"\n\n    def __init__(self, *args, **kwargs):\n        super().__init__(*args, **kwargs, has_fp16_weights=False, threshold=6.0)\n\n        # We quantize the initial weight here so we don't end up filling the device\n        # memory with float32 weights which could lead to OOM.\n        self._quantize_weight(self.weight.data)\n\n    def _load_from_state_dict(self, local_state_dict, *args, **kwargs):\n\n        # There is only one key that ends with `*.weight`, the other one is the bias\n        weight_key = next(\n            (name for name in local_state_dict.keys() if name.endswith(\"weight\")),\n            None,\n        )\n        if weight_key is None:\n            return\n\n        # Load the weight from the state dict and re-quantize it\n        weight = local_state_dict.pop(weight_key)\n        self._quantize_weight(weight)\n\n        # If there is a bias, let nn.Module load it\n        if local_state_dict:\n            super()._load_from_state_dict(local_state_dict, *args, **kwargs)\n\n    def _quantize_weight(self, weight: torch.Tensor) -> None:\n\n        # This code is taken and adapted from `bnb.nn.Int8Params.cuda()`\n        B = weight.contiguous().half().cuda()\n        CB, CBt, SCB, SCBt, coo_tensorB = bnb.functional.double_quant(B)\n        del CBt\n        del SCBt\n        self.weight.data = CB\n        setattr(self.weight, \"CB\", CB)\n        setattr(self.weight, \"SCB\", SCB)","source_hash":"8936905ab12763c28c0324a49457f6680d9ed1c568a746d284ca64bf506a2f60","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:utils.quantization.__init__","uri":"program://LLaMA-Adapter/function/utils.quantization.__init__#L15-L20","kind":"function","name":"__init__","path":"utils/quantization.py","language":"python","start_line":15,"end_line":20,"context_start_line":1,"context_end_line":40,"code":"import torch\nimport bitsandbytes as bnb\n\n\n'''\nlit-llama\n'''\nclass Linear8bitLt(bnb.nn.Linear8bitLt):\n    \"\"\"Wraps `bnb.nn.Linear8bitLt` and enables instantiation directly on the device and\n    re-quantizaton when loading the state dict.\n\n\n    \"\"\"\n\n    def __init__(self, *args, **kwargs):\n        super().__init__(*args, **kwargs, has_fp16_weights=False, threshold=6.0)\n\n        # We quantize the initial weight here so we don't end up filling the device\n        # memory with float32 weights which could lead to OOM.\n        self._quantize_weight(self.weight.data)\n\n    def _load_from_state_dict(self, local_state_dict, *args, **kwargs):\n\n        # There is only one key that ends with `*.weight`, the other one is the bias\n        weight_key = next(\n            (name for name in local_state_dict.keys() if name.endswith(\"weight\")),\n            None,\n        )\n        if weight_key is None:\n            return\n\n        # Load the weight from the state dict and re-quantize it\n        weight = local_state_dict.pop(weight_key)\n        self._quantize_weight(weight)\n\n        # If there is a bias, let nn.Module load it\n        if local_state_dict:\n            super()._load_from_state_dict(local_state_dict, *args, **kwargs)\n\n    def _quantize_weight(self, weight: torch.Tensor) -> None:","source_hash":"8936905ab12763c28c0324a49457f6680d9ed1c568a746d284ca64bf506a2f60","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:utils.quantization._load_from_state_dict","uri":"program://LLaMA-Adapter/function/utils.quantization._load_from_state_dict#L22-L38","kind":"function","name":"_load_from_state_dict","path":"utils/quantization.py","language":"python","start_line":22,"end_line":38,"context_start_line":2,"context_end_line":49,"code":"import bitsandbytes as bnb\n\n\n'''\nlit-llama\n'''\nclass Linear8bitLt(bnb.nn.Linear8bitLt):\n    \"\"\"Wraps `bnb.nn.Linear8bitLt` and enables instantiation directly on the device and\n    re-quantizaton when loading the state dict.\n\n\n    \"\"\"\n\n    def __init__(self, *args, **kwargs):\n        super().__init__(*args, **kwargs, has_fp16_weights=False, threshold=6.0)\n\n        # We quantize the initial weight here so we don't end up filling the device\n        # memory with float32 weights which could lead to OOM.\n        self._quantize_weight(self.weight.data)\n\n    def _load_from_state_dict(self, local_state_dict, *args, **kwargs):\n\n        # There is only one key that ends with `*.weight`, the other one is the bias\n        weight_key = next(\n            (name for name in local_state_dict.keys() if name.endswith(\"weight\")),\n            None,\n        )\n        if weight_key is None:\n            return\n\n        # Load the weight from the state dict and re-quantize it\n        weight = local_state_dict.pop(weight_key)\n        self._quantize_weight(weight)\n\n        # If there is a bias, let nn.Module load it\n        if local_state_dict:\n            super()._load_from_state_dict(local_state_dict, *args, **kwargs)\n\n    def _quantize_weight(self, weight: torch.Tensor) -> None:\n\n        # This code is taken and adapted from `bnb.nn.Int8Params.cuda()`\n        B = weight.contiguous().half().cuda()\n        CB, CBt, SCB, SCBt, coo_tensorB = bnb.functional.double_quant(B)\n        del CBt\n        del SCBt\n        self.weight.data = CB\n        setattr(self.weight, \"CB\", CB)\n        setattr(self.weight, \"SCB\", SCB)","source_hash":"8936905ab12763c28c0324a49457f6680d9ed1c568a746d284ca64bf506a2f60","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:utils.quantization._quantize_weight","uri":"program://LLaMA-Adapter/function/utils.quantization._quantize_weight#L40-L49","kind":"function","name":"_quantize_weight","path":"utils/quantization.py","language":"python","start_line":40,"end_line":49,"context_start_line":20,"context_end_line":49,"code":"        self._quantize_weight(self.weight.data)\n\n    def _load_from_state_dict(self, local_state_dict, *args, **kwargs):\n\n        # There is only one key that ends with `*.weight`, the other one is the bias\n        weight_key = next(\n            (name for name in local_state_dict.keys() if name.endswith(\"weight\")),\n            None,\n        )\n        if weight_key is None:\n            return\n\n        # Load the weight from the state dict and re-quantize it\n        weight = local_state_dict.pop(weight_key)\n        self._quantize_weight(weight)\n\n        # If there is a bias, let nn.Module load it\n        if local_state_dict:\n            super()._load_from_state_dict(local_state_dict, *args, **kwargs)\n\n    def _quantize_weight(self, weight: torch.Tensor) -> None:\n\n        # This code is taken and adapted from `bnb.nn.Int8Params.cuda()`\n        B = weight.contiguous().half().cuda()\n        CB, CBt, SCB, SCBt, coo_tensorB = bnb.functional.double_quant(B)\n        del CBt\n        del SCBt\n        self.weight.data = CB\n        setattr(self.weight, \"CB\", CB)\n        setattr(self.weight, \"SCB\", SCB)","source_hash":"8936905ab12763c28c0324a49457f6680d9ed1c568a746d284ca64bf506a2f60","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.demo","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.demo#L1-L20","kind":"module","name":"llama_adapter_v2_multimodal7b.demo","path":"llama_adapter_v2_multimodal7b/demo.py","language":"python","start_line":1,"end_line":20,"context_start_line":1,"context_end_line":20,"code":"import cv2\nimport llama\nimport torch\nfrom PIL import Image\n\ndevice = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\nllama_dir = \"/path/to/LLaMA/\"\n\n# choose from BIAS-7B, LORA-BIAS-7B, CAPTION-7B.pth\nmodel, preprocess = llama.load(\"BIAS-7B\", llama_dir, device)\nmodel.eval()\n\nprompt = llama.format_prompt('Please introduce this painting.')\nimg = Image.fromarray(cv2.imread(\"../docs/logo_v1.png\"))\nimg = preprocess(img).unsqueeze(0).to(device)\n\nresult = model.generate(img, [prompt])[0]\n\nprint(result)","source_hash":"4aa8d6e038dbd5203bfaf76e378cc880d5ffe3cd6d7e567fd92e18bebada90d7","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.main_finetune","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.main_finetune#L1-L205","kind":"module","name":"llama_adapter_v2_multimodal7b.main_finetune","path":"llama_adapter_v2_multimodal7b/main_finetune.py","language":"python","start_line":1,"end_line":205,"context_start_line":1,"context_end_line":205,"code":"import torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import FinetuneDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_finetune import train_one_epoch\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('llama_adapterV2 pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=64, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--epochs', default=400, type=int)\n    parser.add_argument('--accum_iter', default=1, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='7B', type=str,\n                        help='Type of LLaMA model') #\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='path to checkpoint from pretrain stage')\n    parser.add_argument('--max_words', default=512, type=int,\n                        help='max number of input words')\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.05,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=None, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--blr', type=float, default=1e-3, metavar='LR',\n                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')\n    parser.add_argument('--min_lr', type=float, default=0., metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_epochs', type=int, default=40, metavar='N',\n                        help='epochs to warmup LR')\n\n    # Dataset parameters\n    parser.add_argument('--data_config', default='configs/data/finetune/EN.yaml', type=str,\n                        help='dataset config path')\n    parser.add_argument('--num_workers', default=10, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n\n    parser.add_argument('--output_dir', default='./output',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')\n    model = LLaMA_adapter(llama_ckpt_dir, llama_tokenzier_path)\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    print(\"Trainable Params:\")\n    print([(key, val.shape) for key, val in model.named_parameters() if val.requires_grad])\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # training detail\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n    misc.load_model(model_without_ddp, args.pretrained_path)\n\n\n    dataset_train = FinetuneDataset(args.data_config, transform=transform_train,\n                                max_words=args.max_words, tokenizer_path=llama_tokenzier_path)\n    print(dataset_train)\n    num_tasks = misc.get_world_size()\n    global_rank = misc.get_rank()\n    sampler_train = torch.utils.data.DistributedSampler(\n        dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True\n    )\n    print(\"Sampler_train = %s\" % str(sampler_train))\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train, sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # SummaryWrite\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train,\n            optimizer, device, epoch, loss_scaler,\n            log_writer=log_writer,\n            args=args\n        )\n\n        if args.output_dir and (epoch % 5 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,\n                loss_scaler=loss_scaler, epoch=epoch)\n\n        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},\n                     'epoch': epoch,\n                     **{f'val_{k}': v for k, v in train_stats.items()}}\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"671fe114125e0c1e86a43b56d22a85b59683c3fe053378e5743eeaa5c1506dfa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.main_finetune.get_args_parser","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.main_finetune.get_args_parser#L23-L85","kind":"function","name":"get_args_parser","path":"llama_adapter_v2_multimodal7b/main_finetune.py","language":"python","start_line":23,"end_line":85,"context_start_line":3,"context_end_line":105,"code":"from torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import FinetuneDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_finetune import train_one_epoch\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('llama_adapterV2 pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=64, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--epochs', default=400, type=int)\n    parser.add_argument('--accum_iter', default=1, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='7B', type=str,\n                        help='Type of LLaMA model') #\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='path to checkpoint from pretrain stage')\n    parser.add_argument('--max_words', default=512, type=int,\n                        help='max number of input words')\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.05,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=None, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--blr', type=float, default=1e-3, metavar='LR',\n                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')\n    parser.add_argument('--min_lr', type=float, default=0., metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_epochs', type=int, default=40, metavar='N',\n                        help='epochs to warmup LR')\n\n    # Dataset parameters\n    parser.add_argument('--data_config', default='configs/data/finetune/EN.yaml', type=str,\n                        help='dataset config path')\n    parser.add_argument('--num_workers', default=10, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n\n    parser.add_argument('--output_dir', default='./output',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')","source_hash":"671fe114125e0c1e86a43b56d22a85b59683c3fe053378e5743eeaa5c1506dfa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.main_finetune.main","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.main_finetune.main#L88-L197","kind":"function","name":"main","path":"llama_adapter_v2_multimodal7b/main_finetune.py","language":"python","start_line":88,"end_line":197,"context_start_line":68,"context_end_line":205,"code":"                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')\n    model = LLaMA_adapter(llama_ckpt_dir, llama_tokenzier_path)\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    print(\"Trainable Params:\")\n    print([(key, val.shape) for key, val in model.named_parameters() if val.requires_grad])\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # training detail\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n    misc.load_model(model_without_ddp, args.pretrained_path)\n\n\n    dataset_train = FinetuneDataset(args.data_config, transform=transform_train,\n                                max_words=args.max_words, tokenizer_path=llama_tokenzier_path)\n    print(dataset_train)\n    num_tasks = misc.get_world_size()\n    global_rank = misc.get_rank()\n    sampler_train = torch.utils.data.DistributedSampler(\n        dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True\n    )\n    print(\"Sampler_train = %s\" % str(sampler_train))\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train, sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # SummaryWrite\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train,\n            optimizer, device, epoch, loss_scaler,\n            log_writer=log_writer,\n            args=args\n        )\n\n        if args.output_dir and (epoch % 5 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,\n                loss_scaler=loss_scaler, epoch=epoch)\n\n        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},\n                     'epoch': epoch,\n                     **{f'val_{k}': v for k, v in train_stats.items()}}\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"671fe114125e0c1e86a43b56d22a85b59683c3fe053378e5743eeaa5c1506dfa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.main_pretrain","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.main_pretrain#L1-L202","kind":"module","name":"llama_adapter_v2_multimodal7b.main_pretrain","path":"llama_adapter_v2_multimodal7b/main_pretrain.py","language":"python","start_line":1,"end_line":202,"context_start_line":1,"context_end_line":202,"code":"import torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import PretrainDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_pretrain import train_one_epoch\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('llama_adapterV2 pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=64, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--epochs', default=400, type=int)\n    parser.add_argument('--accum_iter', default=1, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='7B', type=str,\n                        help='Type of LLaMA model') #\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--max_words', default=96, type=int,\n                        help='max number of input words')\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.05,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=None, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--blr', type=float, default=1e-3, metavar='LR',\n                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')\n    parser.add_argument('--min_lr', type=float, default=0., metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_epochs', type=int, default=40, metavar='N',\n                        help='epochs to warmup LR')\n\n    # Dataset parameters\n    parser.add_argument('--data_config', default='configs/data/pretrain/EN.yaml', type=str,\n                        help='dataset config path')\n    parser.add_argument('--num_workers', default=10, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n\n    parser.add_argument('--output_dir', default='./output',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--split_epoch', type=int, default=50)\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')\n    model = LLaMA_adapter(llama_ckpt_dir, llama_tokenzier_path, phase=\"pretrain\")\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    print(\"Trainable Params:\")\n    print([(key, val.shape) for key, val in model.named_parameters() if val.requires_grad])\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # training detail\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n\n\n    dataset_train = PretrainDataset(args.data_config, transform=transform_train,\n                                max_words=args.max_words, tokenizer_path=llama_tokenzier_path)\n    print(dataset_train)\n    num_tasks = misc.get_world_size()\n    global_rank = misc.get_rank()\n    sampler_train = misc.DistributedSubEpochSampler(\n        dataset_train, num_replicas=num_tasks, rank=global_rank, split_epoch=args.split_epoch, shuffle=True\n    )\n    print(\"Sampler_train = %s\" % str(sampler_train))\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train, sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # SummaryWrite\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train,\n            optimizer, device, epoch, loss_scaler,\n            log_writer=log_writer,\n            args=args\n        )\n\n        if args.output_dir and (epoch % 2 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,\n                loss_scaler=loss_scaler, epoch=epoch)\n\n        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},\n                     'epoch': epoch}\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"25f96771b527c0cf0ac998c008639df7231da9c09660c1f448c7451fa94e1d64","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.main_pretrain.get_args_parser","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.main_pretrain.get_args_parser#L23-L84","kind":"function","name":"get_args_parser","path":"llama_adapter_v2_multimodal7b/main_pretrain.py","language":"python","start_line":23,"end_line":84,"context_start_line":3,"context_end_line":104,"code":"from torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import PretrainDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_pretrain import train_one_epoch\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('llama_adapterV2 pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=64, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--epochs', default=400, type=int)\n    parser.add_argument('--accum_iter', default=1, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='7B', type=str,\n                        help='Type of LLaMA model') #\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--max_words', default=96, type=int,\n                        help='max number of input words')\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.05,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=None, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--blr', type=float, default=1e-3, metavar='LR',\n                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')\n    parser.add_argument('--min_lr', type=float, default=0., metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_epochs', type=int, default=40, metavar='N',\n                        help='epochs to warmup LR')\n\n    # Dataset parameters\n    parser.add_argument('--data_config', default='configs/data/pretrain/EN.yaml', type=str,\n                        help='dataset config path')\n    parser.add_argument('--num_workers', default=10, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n\n    parser.add_argument('--output_dir', default='./output',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--split_epoch', type=int, default=50)\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')","source_hash":"25f96771b527c0cf0ac998c008639df7231da9c09660c1f448c7451fa94e1d64","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.main_pretrain.main","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.main_pretrain.main#L87-L194","kind":"function","name":"main","path":"llama_adapter_v2_multimodal7b/main_pretrain.py","language":"python","start_line":87,"end_line":194,"context_start_line":67,"context_end_line":202,"code":"    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--split_epoch', type=int, default=50)\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')\n    model = LLaMA_adapter(llama_ckpt_dir, llama_tokenzier_path, phase=\"pretrain\")\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    print(\"Trainable Params:\")\n    print([(key, val.shape) for key, val in model.named_parameters() if val.requires_grad])\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # training detail\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n\n\n    dataset_train = PretrainDataset(args.data_config, transform=transform_train,\n                                max_words=args.max_words, tokenizer_path=llama_tokenzier_path)\n    print(dataset_train)\n    num_tasks = misc.get_world_size()\n    global_rank = misc.get_rank()\n    sampler_train = misc.DistributedSubEpochSampler(\n        dataset_train, num_replicas=num_tasks, rank=global_rank, split_epoch=args.split_epoch, shuffle=True\n    )\n    print(\"Sampler_train = %s\" % str(sampler_train))\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train, sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # SummaryWrite\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train,\n            optimizer, device, epoch, loss_scaler,\n            log_writer=log_writer,\n            args=args\n        )\n\n        if args.output_dir and (epoch % 2 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,\n                loss_scaler=loss_scaler, epoch=epoch)\n\n        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},\n                     'epoch': epoch}\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"25f96771b527c0cf0ac998c008639df7231da9c09660c1f448c7451fa94e1d64","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.engine_finetune","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.engine_finetune#L1-L77","kind":"module","name":"llama_adapter_v2_multimodal7b.engine_finetune","path":"llama_adapter_v2_multimodal7b/engine_finetune.py","language":"python","start_line":1,"end_line":77,"context_start_line":1,"context_end_line":77,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, imgs) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        imgs = imgs.to(device, non_blocking=True)\n        with torch.cuda.amp.autocast():\n             c_loss, m_loss = model(examples, labels, imgs)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n        loss_scaler(loss, optimizer, parameters=model.parameters(),\n                    update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.engine_finetune.train_one_epoch","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.engine_finetune.train_one_epoch#L12-L77","kind":"function","name":"train_one_epoch","path":"llama_adapter_v2_multimodal7b/engine_finetune.py","language":"python","start_line":12,"end_line":77,"context_start_line":1,"context_end_line":77,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, imgs) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        imgs = imgs.to(device, non_blocking=True)\n        with torch.cuda.amp.autocast():\n             c_loss, m_loss = model(examples, labels, imgs)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n        loss_scaler(loss, optimizer, parameters=model.parameters(),\n                    update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.gradio_app","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.gradio_app#L1-L91","kind":"module","name":"llama_adapter_v2_multimodal7b.gradio_app","path":"llama_adapter_v2_multimodal7b/gradio_app.py","language":"python","start_line":1,"end_line":91,"context_start_line":1,"context_end_line":91,"code":"import cv2\nimport gradio as gr\nimport torch\nfrom PIL import Image\n\nimport llama\n\n\ndevice = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\nllama_dir = \"/path/to/LLaMA/\"\n\nmodel, preprocess = llama.load(\"BIAS-7B\", llama_dir, device)\nmodel.half()\nmodel.eval()\n\ndef multi_modal_generate(\n    img_path: str,\n    prompt: str,\n    max_gen_len=256,\n    temperature: float = 0.1,\n    top_p: float = 0.75,\n):\n    try:\n        img = Image.fromarray(cv2.imread(img_path))\n    except:\n        return \"\"\n\n    img = preprocess(img).unsqueeze(0).half().to(device)\n    prompt = llama.format_prompt(prompt)\n\n    result = model.generate(img, [prompt], \n                            max_gen_len=max_gen_len, \n                            temperature=temperature, \n                            top_p=top_p)\n    print(result[0])\n    return result[0]\n\n\ndef create_multi_modal_demo():\n    with gr.Blocks() as instruct_demo:\n        with gr.Row():\n            with gr.Column():\n                img = gr.Image(label='Input', type='filepath')\n                question = gr.Textbox(lines=2, label=\"Prompt\")\n                max_len = gr.Slider(minimum=1, maximum=512,\n                                    value=256, label=\"Max length\")\n                with gr.Accordion(label='Advanced options', open=False):\n                    temp = gr.Slider(minimum=0, maximum=1,\n                                     value=0.1, label=\"Temperature\")\n                    top_p = gr.Slider(minimum=0, maximum=1,\n                                      value=0.75, label=\"Top p\")\n\n                run_botton = gr.Button(\"Run\")\n\n            with gr.Column():\n                outputs = gr.Textbox(lines=10, label=\"Output\")\n\n        inputs = [img, question, max_len, temp, top_p]\n\n        examples = [\n            [\"../docs/logo_v1.png\", \"Please introduce this painting.\", 256, 0.1, 0.75],\n        ]\n\n        gr.Examples(\n            examples=examples,\n            inputs=inputs,\n            outputs=outputs,\n            fn=multi_modal_generate,\n            cache_examples=False\n        )\n        run_botton.click(fn=multi_modal_generate,\n                         inputs=inputs, outputs=outputs)\n    return instruct_demo\n\n\ndescription = \"\"\"\n# LLaMA-Adapter V2🚀\nThe official demo for **LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model**.\n\nPlease refer to our [arXiv paper](https://arxiv.org/abs/2304.15010) and [github](https://github.com/ZrrSkywalker/LLaMA-Adapter) for more details.\n\nThe demo for **LLaMA-Adapter V1** is available at: [Huggingface Spaces](https://huggingface.co/spaces/csuhan/LLaMA-Adapter).\n\"\"\"\n\nwith gr.Blocks(css=\"h1,p {text-align: center;}\") as demo:\n    gr.Markdown(description)\n    with gr.TabItem(\"Multi-Modal Interaction\"):\n        create_multi_modal_demo()\n\ndemo.queue(api_open=True, concurrency_count=1).launch(share=True)","source_hash":"800829419bded0f9a137c7489c91d1883db81fc4ca25045b5176b75c169f6eb2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.gradio_app.multi_modal_generate","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.gradio_app.multi_modal_generate#L17-L37","kind":"function","name":"multi_modal_generate","path":"llama_adapter_v2_multimodal7b/gradio_app.py","language":"python","start_line":17,"end_line":37,"context_start_line":1,"context_end_line":57,"code":"import cv2\nimport gradio as gr\nimport torch\nfrom PIL import Image\n\nimport llama\n\n\ndevice = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\nllama_dir = \"/path/to/LLaMA/\"\n\nmodel, preprocess = llama.load(\"BIAS-7B\", llama_dir, device)\nmodel.half()\nmodel.eval()\n\ndef multi_modal_generate(\n    img_path: str,\n    prompt: str,\n    max_gen_len=256,\n    temperature: float = 0.1,\n    top_p: float = 0.75,\n):\n    try:\n        img = Image.fromarray(cv2.imread(img_path))\n    except:\n        return \"\"\n\n    img = preprocess(img).unsqueeze(0).half().to(device)\n    prompt = llama.format_prompt(prompt)\n\n    result = model.generate(img, [prompt], \n                            max_gen_len=max_gen_len, \n                            temperature=temperature, \n                            top_p=top_p)\n    print(result[0])\n    return result[0]\n\n\ndef create_multi_modal_demo():\n    with gr.Blocks() as instruct_demo:\n        with gr.Row():\n            with gr.Column():\n                img = gr.Image(label='Input', type='filepath')\n                question = gr.Textbox(lines=2, label=\"Prompt\")\n                max_len = gr.Slider(minimum=1, maximum=512,\n                                    value=256, label=\"Max length\")\n                with gr.Accordion(label='Advanced options', open=False):\n                    temp = gr.Slider(minimum=0, maximum=1,\n                                     value=0.1, label=\"Temperature\")\n                    top_p = gr.Slider(minimum=0, maximum=1,\n                                      value=0.75, label=\"Top p\")\n\n                run_botton = gr.Button(\"Run\")\n\n            with gr.Column():\n                outputs = gr.Textbox(lines=10, label=\"Output\")","source_hash":"800829419bded0f9a137c7489c91d1883db81fc4ca25045b5176b75c169f6eb2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.gradio_app.create_multi_modal_demo","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.gradio_app.create_multi_modal_demo#L40-L74","kind":"function","name":"create_multi_modal_demo","path":"llama_adapter_v2_multimodal7b/gradio_app.py","language":"python","start_line":40,"end_line":74,"context_start_line":20,"context_end_line":91,"code":"    max_gen_len=256,\n    temperature: float = 0.1,\n    top_p: float = 0.75,\n):\n    try:\n        img = Image.fromarray(cv2.imread(img_path))\n    except:\n        return \"\"\n\n    img = preprocess(img).unsqueeze(0).half().to(device)\n    prompt = llama.format_prompt(prompt)\n\n    result = model.generate(img, [prompt], \n                            max_gen_len=max_gen_len, \n                            temperature=temperature, \n                            top_p=top_p)\n    print(result[0])\n    return result[0]\n\n\ndef create_multi_modal_demo():\n    with gr.Blocks() as instruct_demo:\n        with gr.Row():\n            with gr.Column():\n                img = gr.Image(label='Input', type='filepath')\n                question = gr.Textbox(lines=2, label=\"Prompt\")\n                max_len = gr.Slider(minimum=1, maximum=512,\n                                    value=256, label=\"Max length\")\n                with gr.Accordion(label='Advanced options', open=False):\n                    temp = gr.Slider(minimum=0, maximum=1,\n                                     value=0.1, label=\"Temperature\")\n                    top_p = gr.Slider(minimum=0, maximum=1,\n                                      value=0.75, label=\"Top p\")\n\n                run_botton = gr.Button(\"Run\")\n\n            with gr.Column():\n                outputs = gr.Textbox(lines=10, label=\"Output\")\n\n        inputs = [img, question, max_len, temp, top_p]\n\n        examples = [\n            [\"../docs/logo_v1.png\", \"Please introduce this painting.\", 256, 0.1, 0.75],\n        ]\n\n        gr.Examples(\n            examples=examples,\n            inputs=inputs,\n            outputs=outputs,\n            fn=multi_modal_generate,\n            cache_examples=False\n        )\n        run_botton.click(fn=multi_modal_generate,\n                         inputs=inputs, outputs=outputs)\n    return instruct_demo\n\n\ndescription = \"\"\"\n# LLaMA-Adapter V2🚀\nThe official demo for **LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model**.\n\nPlease refer to our [arXiv paper](https://arxiv.org/abs/2304.15010) and [github](https://github.com/ZrrSkywalker/LLaMA-Adapter) for more details.\n\nThe demo for **LLaMA-Adapter V1** is available at: [Huggingface Spaces](https://huggingface.co/spaces/csuhan/LLaMA-Adapter).\n\"\"\"\n\nwith gr.Blocks(css=\"h1,p {text-align: center;}\") as demo:\n    gr.Markdown(description)\n    with gr.TabItem(\"Multi-Modal Interaction\"):\n        create_multi_modal_demo()\n\ndemo.queue(api_open=True, concurrency_count=1).launch(share=True)","source_hash":"800829419bded0f9a137c7489c91d1883db81fc4ca25045b5176b75c169f6eb2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.engine_pretrain","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.engine_pretrain#L1-L77","kind":"module","name":"llama_adapter_v2_multimodal7b.engine_pretrain","path":"llama_adapter_v2_multimodal7b/engine_pretrain.py","language":"python","start_line":1,"end_line":77,"context_start_line":1,"context_end_line":77,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, imgs) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        imgs = imgs.to(device, non_blocking=True)\n        with torch.cuda.amp.autocast():\n             c_loss, m_loss = model(examples, labels, imgs)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n        loss_scaler(loss, optimizer, parameters=model.parameters(),\n                    update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.engine_pretrain.train_one_epoch","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.engine_pretrain.train_one_epoch#L12-L77","kind":"function","name":"train_one_epoch","path":"llama_adapter_v2_multimodal7b/engine_pretrain.py","language":"python","start_line":12,"end_line":77,"context_start_line":1,"context_end_line":77,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, imgs) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        imgs = imgs.to(device, non_blocking=True)\n        with torch.cuda.amp.autocast():\n             c_loss, m_loss = model(examples, labels, imgs)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n        loss_scaler(loss, optimizer, parameters=model.parameters(),\n                    update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.extract_adapter_from_checkpoint","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.util.extract_adapter_from_checkpoint#L1-L52","kind":"module","name":"llama_adapter_v2_multimodal7b.util.extract_adapter_from_checkpoint","path":"llama_adapter_v2_multimodal7b/util/extract_adapter_from_checkpoint.py","language":"python","start_line":1,"end_line":52,"context_start_line":1,"context_end_line":52,"code":"import torch\n\ndef save(full_model, path, model_type = 'BIAS'):\n    if model_type == 'BIAS':\n        keys = [\n            f'visual_blocks.{i}.{key}.{suffix}'\n            for i in range(8)\n            for key in ['norm1', 'attn.qkv', 'attn.proj', 'norm2', 'mlp.fc1', 'mlp.fc2']\n            for suffix in ['weight', 'bias']\n        ] + [\n            f'llama.layers.{i}.{key}'\n            for i in range(32)\n            for key in ['attention.gate', 'attention.wq.bias', 'attention.wo.bias', 'feed_forward.w1.bias', 'feed_forward.w2.bias', 'feed_forward.w3.bias', 'attention_norm.weight', 'ffn_norm.weight']\n        ] + [\n            f'{base_key}.{suffix}'\n            for base_key in ['clip_proj_norm', 'visual_proj_norm', 'visual_proj', 'clip_proj']\n            for suffix in ['weight', 'bias']\n        ] + ['llama.norm.weight', 'visual_query.weight', 'adapter_query.weight']\n\n    \n    elif model_type == 'LORA':\n        keys = [\n            f'visual_blocks.{i}.{key}.{suffix}'\n            for i in range(8)\n            for key in [f'norm{j}' for j in range(1, 3)] + ['attn.qkv', 'attn.proj', 'mlp.fc1', 'mlp.fc2']\n            for suffix in ['weight', 'bias']\n        ] + [\n            f'llama.layers.{i}.{key}'\n            for i in range(32)\n            for key in ['attention.gate', 'attention.wq.bias', 'attention.wo.bias', 'feed_forward.w1.bias', 'feed_forward.w2.bias', 'feed_forward.w3.bias', 'attention_norm.weight', 'ffn_norm.weight']\n                + [f'attention.lora_wk_l{j}.weight' for j in range(1, 3)]\n                + [f'attention.lora_wo_l{j}.weight' for j in range(1, 3)]\n                + [f'feed_forward.lora_w{k}_l{j}.weight' for k in range(1, 4) for j in range(1, 3)]\n                + [f'attention.lora_wq_l{j}.weight' for j in range(1, 3)]\n                + [f'attention.lora_wv_l{j}.weight' for j in range(1, 3)]\n                + ['attention.new_gate']\n        ] + [\n            f'{base_key}.{suffix}'\n            for base_key in ['clip_proj_norm', 'visual_proj_norm', 'visual_proj', 'clip_proj']\n            for suffix in ['weight', 'bias']\n        ] + ['llama.norm.weight', 'visual_query.weight', 'adapter_query.weight']\n\n    ## TODO: Add other model types\n\n    full_model_state_dict = full_model.state_dict()\n    small_weights = {key: full_model_state_dict[key] for key in keys}\n    if model_type == 'BIAS':\n        wrapped_small_weights = {'model': small_weights,'config': {'w_bias': True, 'w_lora': False, 'lora_rank': 16}}\n    elif model_type == 'LORA':\n        wrapped_small_weights = {'model': small_weights,'config': {'w_bias': True, 'w_lora': True,  'lora_rank': 16}}\n    # Save the wrapped small weights\n    torch.save(wrapped_small_weights, path)","source_hash":"cdf86474579900b917d88f2488ef9b151ba11a87246765b5ec3640b5c83d4838","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.extract_adapter_from_checkpoint.save","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.extract_adapter_from_checkpoint.save#L3-L52","kind":"function","name":"save","path":"llama_adapter_v2_multimodal7b/util/extract_adapter_from_checkpoint.py","language":"python","start_line":3,"end_line":52,"context_start_line":1,"context_end_line":52,"code":"import torch\n\ndef save(full_model, path, model_type = 'BIAS'):\n    if model_type == 'BIAS':\n        keys = [\n            f'visual_blocks.{i}.{key}.{suffix}'\n            for i in range(8)\n            for key in ['norm1', 'attn.qkv', 'attn.proj', 'norm2', 'mlp.fc1', 'mlp.fc2']\n            for suffix in ['weight', 'bias']\n        ] + [\n            f'llama.layers.{i}.{key}'\n            for i in range(32)\n            for key in ['attention.gate', 'attention.wq.bias', 'attention.wo.bias', 'feed_forward.w1.bias', 'feed_forward.w2.bias', 'feed_forward.w3.bias', 'attention_norm.weight', 'ffn_norm.weight']\n        ] + [\n            f'{base_key}.{suffix}'\n            for base_key in ['clip_proj_norm', 'visual_proj_norm', 'visual_proj', 'clip_proj']\n            for suffix in ['weight', 'bias']\n        ] + ['llama.norm.weight', 'visual_query.weight', 'adapter_query.weight']\n\n    \n    elif model_type == 'LORA':\n        keys = [\n            f'visual_blocks.{i}.{key}.{suffix}'\n            for i in range(8)\n            for key in [f'norm{j}' for j in range(1, 3)] + ['attn.qkv', 'attn.proj', 'mlp.fc1', 'mlp.fc2']\n            for suffix in ['weight', 'bias']\n        ] + [\n            f'llama.layers.{i}.{key}'\n            for i in range(32)\n            for key in ['attention.gate', 'attention.wq.bias', 'attention.wo.bias', 'feed_forward.w1.bias', 'feed_forward.w2.bias', 'feed_forward.w3.bias', 'attention_norm.weight', 'ffn_norm.weight']\n                + [f'attention.lora_wk_l{j}.weight' for j in range(1, 3)]\n                + [f'attention.lora_wo_l{j}.weight' for j in range(1, 3)]\n                + [f'feed_forward.lora_w{k}_l{j}.weight' for k in range(1, 4) for j in range(1, 3)]\n                + [f'attention.lora_wq_l{j}.weight' for j in range(1, 3)]\n                + [f'attention.lora_wv_l{j}.weight' for j in range(1, 3)]\n                + ['attention.new_gate']\n        ] + [\n            f'{base_key}.{suffix}'\n            for base_key in ['clip_proj_norm', 'visual_proj_norm', 'visual_proj', 'clip_proj']\n            for suffix in ['weight', 'bias']\n        ] + ['llama.norm.weight', 'visual_query.weight', 'adapter_query.weight']\n\n    ## TODO: Add other model types\n\n    full_model_state_dict = full_model.state_dict()\n    small_weights = {key: full_model_state_dict[key] for key in keys}\n    if model_type == 'BIAS':\n        wrapped_small_weights = {'model': small_weights,'config': {'w_bias': True, 'w_lora': False, 'lora_rank': 16}}\n    elif model_type == 'LORA':\n        wrapped_small_weights = {'model': small_weights,'config': {'w_bias': True, 'w_lora': True,  'lora_rank': 16}}\n    # Save the wrapped small weights\n    torch.save(wrapped_small_weights, path)","source_hash":"cdf86474579900b917d88f2488ef9b151ba11a87246765b5ec3640b5c83d4838","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.util.misc#L1-L413","kind":"module","name":"llama_adapter_v2_multimodal7b.util.misc","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":1,"end_line":413,"context_start_line":1,"context_end_line":413,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\nimport urllib\nfrom tqdm import tqdm\n\nimport torch\nimport torch.utils.data\nimport torch.distributed as dist\nfrom torch._six import inf\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    print(\"GPU::\", args.gpu)\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(model_without_ddp, path):\n    if path.startswith('https'):\n        checkpoint = torch.hub.load_state_dict_from_url(\n            path, map_location='cpu', check_hash=True)\n    else:\n        checkpoint = torch.load(path, map_location='cpu')\n    new_checkpoint = {}\n    for key, value in checkpoint['model'].items():\n        key = key.replace(\"llma\", \"llama\")\n        new_checkpoint[key] = value\n    print(model_without_ddp.load_state_dict(new_checkpoint, strict=False))\n    print(\"Load checkpoint %s\" % path)\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        return download_target\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:\n                    break\n                output.write(buffer)\n                loop.update(len(buffer))\n\n\n    return download_target","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.SmoothedValue","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.util.misc.SmoothedValue#L27-L86","kind":"class","name":"SmoothedValue","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":27,"end_line":86,"context_start_line":7,"context_end_line":106,"code":"# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\nimport urllib\nfrom tqdm import tqdm\n\nimport torch\nimport torch.utils.data\nimport torch.distributed as dist\nfrom torch._six import inf\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.MetricLogger","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.util.misc.MetricLogger#L89-L170","kind":"class","name":"MetricLogger","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":89,"end_line":170,"context_start_line":69,"context_end_line":190,"code":"    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.setup_for_distributed","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.setup_for_distributed#L173-L187","kind":"function","name":"setup_for_distributed","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":173,"end_line":187,"context_start_line":153,"context_end_line":207,"code":"                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.is_dist_avail_and_initialized","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.is_dist_avail_and_initialized#L190-L195","kind":"function","name":"is_dist_avail_and_initialized","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":190,"end_line":195,"context_start_line":170,"context_end_line":215,"code":"            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.get_world_size","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.get_world_size#L198-L201","kind":"function","name":"get_world_size","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":198,"end_line":201,"context_start_line":178,"context_end_line":221,"code":"\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.get_rank","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.get_rank#L204-L207","kind":"function","name":"get_rank","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":204,"end_line":207,"context_start_line":184,"context_end_line":227,"code":"            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.is_main_process","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.is_main_process#L210-L211","kind":"function","name":"is_main_process","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":210,"end_line":211,"context_start_line":190,"context_end_line":231,"code":"def is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.save_on_master","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.save_on_master#L214-L216","kind":"function","name":"save_on_master","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":214,"end_line":216,"context_start_line":194,"context_end_line":236,"code":"        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.init_distributed_mode","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.init_distributed_mode#L219-L252","kind":"function","name":"init_distributed_mode","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":219,"end_line":252,"context_start_line":199,"context_end_line":272,"code":"    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    print(\"GPU::\", args.gpu)\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.NativeScalerWithGradNormCount","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.util.misc.NativeScalerWithGradNormCount#L255-L281","kind":"class","name":"NativeScalerWithGradNormCount","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":255,"end_line":281,"context_start_line":235,"context_end_line":301,"code":"        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    print(\"GPU::\", args.gpu)\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.get_grad_norm_","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.get_grad_norm_#L284-L296","kind":"function","name":"get_grad_norm_","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":284,"end_line":296,"context_start_line":264,"context_end_line":316,"code":"            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.save_model","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.save_model#L299-L316","kind":"function","name":"save_model","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":299,"end_line":316,"context_start_line":279,"context_end_line":336,"code":"\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(model_without_ddp, path):\n    if path.startswith('https'):\n        checkpoint = torch.hub.load_state_dict_from_url(\n            path, map_location='cpu', check_hash=True)\n    else:\n        checkpoint = torch.load(path, map_location='cpu')\n    new_checkpoint = {}\n    for key, value in checkpoint['model'].items():\n        key = key.replace(\"llma\", \"llama\")\n        new_checkpoint[key] = value\n    print(model_without_ddp.load_state_dict(new_checkpoint, strict=False))\n    print(\"Load checkpoint %s\" % path)\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.load_model","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.load_model#L319-L330","kind":"function","name":"load_model","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":319,"end_line":330,"context_start_line":299,"context_end_line":350,"code":"def save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(model_without_ddp, path):\n    if path.startswith('https'):\n        checkpoint = torch.hub.load_state_dict_from_url(\n            path, map_location='cpu', check_hash=True)\n    else:\n        checkpoint = torch.load(path, map_location='cpu')\n    new_checkpoint = {}\n    for key, value in checkpoint['model'].items():\n        key = key.replace(\"llma\", \"llama\")\n        new_checkpoint[key] = value\n    print(model_without_ddp.load_state_dict(new_checkpoint, strict=False))\n    print(\"Load checkpoint %s\" % path)\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.all_reduce_mean","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.all_reduce_mean#L333-L341","kind":"function","name":"all_reduce_mean","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":333,"end_line":341,"context_start_line":313,"context_end_line":361,"code":"            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(model_without_ddp, path):\n    if path.startswith('https'):\n        checkpoint = torch.hub.load_state_dict_from_url(\n            path, map_location='cpu', check_hash=True)\n    else:\n        checkpoint = torch.load(path, map_location='cpu')\n    new_checkpoint = {}\n    for key, value in checkpoint['model'].items():\n        key = key.replace(\"llma\", \"llama\")\n        new_checkpoint[key] = value\n    print(model_without_ddp.load_state_dict(new_checkpoint, strict=False))\n    print(\"Load checkpoint %s\" % path)\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.add_weight_decay","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.add_weight_decay#L344-L356","kind":"function","name":"add_weight_decay","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":344,"end_line":356,"context_start_line":324,"context_end_line":376,"code":"        checkpoint = torch.load(path, map_location='cpu')\n    new_checkpoint = {}\n    for key, value in checkpoint['model'].items():\n        key = key.replace(\"llma\", \"llama\")\n        new_checkpoint[key] = value\n    print(model_without_ddp.load_state_dict(new_checkpoint, strict=False))\n    print(\"Load checkpoint %s\" % path)\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.DistributedSubEpochSampler","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.util.misc.DistributedSubEpochSampler#L359-L390","kind":"class","name":"DistributedSubEpochSampler","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":359,"end_line":390,"context_start_line":339,"context_end_line":410,"code":"        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        return download_target\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:\n                    break\n                output.write(buffer)\n                loop.update(len(buffer))","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.download","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.download#L392-L413","kind":"function","name":"download","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":392,"end_line":413,"context_start_line":372,"context_end_line":413,"code":"        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        return download_target\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:\n                    break\n                output.write(buffer)\n                loop.update(len(buffer))\n\n\n    return download_target","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.__init__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.__init__#L361-L369","kind":"function","name":"__init__","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":361,"end_line":369,"context_start_line":341,"context_end_line":389,"code":"        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.update","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.update#L94-L101","kind":"function","name":"update","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":94,"end_line":101,"context_start_line":74,"context_end_line":121,"code":"        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.synchronize_between_processes","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.synchronize_between_processes#L119-L121","kind":"function","name":"synchronize_between_processes","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":119,"end_line":121,"context_start_line":99,"context_end_line":141,"code":"                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.median","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.median#L59-L61","kind":"function","name":"median","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":59,"end_line":61,"context_start_line":39,"context_end_line":81,"code":"\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.avg","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.avg#L64-L66","kind":"function","name":"avg","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":64,"end_line":66,"context_start_line":44,"context_end_line":86,"code":"\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.global_avg","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.global_avg#L69-L70","kind":"function","name":"global_avg","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":69,"end_line":70,"context_start_line":49,"context_end_line":90,"code":"        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.max","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.max#L73-L74","kind":"function","name":"max","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":73,"end_line":74,"context_start_line":53,"context_end_line":94,"code":"        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.value","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.value#L77-L78","kind":"function","name":"value","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":77,"end_line":78,"context_start_line":57,"context_end_line":98,"code":"\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.__str__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.__str__#L111-L117","kind":"function","name":"__str__","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":111,"end_line":117,"context_start_line":91,"context_end_line":137,"code":"        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.__getattr__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.__getattr__#L103-L109","kind":"function","name":"__getattr__","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":103,"end_line":109,"context_start_line":83,"context_end_line":129,"code":"            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.add_meter","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.add_meter#L123-L124","kind":"function","name":"add_meter","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":123,"end_line":124,"context_start_line":103,"context_end_line":144,"code":"    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.log_every","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.log_every#L126-L170","kind":"function","name":"log_every","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":126,"end_line":170,"context_start_line":106,"context_end_line":190,"code":"        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.print","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.print#L179-L185","kind":"function","name":"print","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":179,"end_line":185,"context_start_line":159,"context_end_line":205,"code":"                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.__call__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.__call__#L261-L275","kind":"function","name":"__call__","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":261,"end_line":275,"context_start_line":241,"context_end_line":295,"code":"\n    args.distributed = True\n\n    print(\"GPU::\", args.gpu)\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.state_dict","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.state_dict#L277-L278","kind":"function","name":"state_dict","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":277,"end_line":278,"context_start_line":257,"context_end_line":298,"code":"\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.load_state_dict","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.load_state_dict#L280-L281","kind":"function","name":"load_state_dict","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":280,"end_line":281,"context_start_line":260,"context_end_line":301,"code":"\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.__len__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.__len__#L371-L372","kind":"function","name":"__len__","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":371,"end_line":372,"context_start_line":351,"context_end_line":392,"code":"            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.__iter__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.__iter__#L374-L387","kind":"function","name":"__iter__","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":374,"end_line":387,"context_start_line":354,"context_end_line":407,"code":"    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        return download_target\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.misc.set_epoch","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.misc.set_epoch#L389-L390","kind":"function","name":"set_epoch","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":389,"end_line":390,"context_start_line":369,"context_end_line":410,"code":"        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        return download_target\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:\n                    break\n                output.write(buffer)\n                loop.update(len(buffer))","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.lr_sched","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.util.lr_sched#L1-L21","kind":"module","name":"llama_adapter_v2_multimodal7b.util.lr_sched","path":"llama_adapter_v2_multimodal7b/util/lr_sched.py","language":"python","start_line":1,"end_line":21,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs \n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * \\\n            (1. + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs)))\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"4ab5d5633bda0be9173ec91570bb3050326d942582ded2267702b53c3ac87c2c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.lr_sched.adjust_learning_rate","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.lr_sched.adjust_learning_rate#L9-L21","kind":"function","name":"adjust_learning_rate","path":"llama_adapter_v2_multimodal7b/util/lr_sched.py","language":"python","start_line":9,"end_line":21,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs \n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * \\\n            (1. + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs)))\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"4ab5d5633bda0be9173ec91570bb3050326d942582ded2267702b53c3ac87c2c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.evaluate_mme","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.util.evaluate_mme#L1-L184","kind":"module","name":"llama_adapter_v2_multimodal7b.util.evaluate_mme","path":"llama_adapter_v2_multimodal7b/util/evaluate_mme.py","language":"python","start_line":1,"end_line":184,"context_start_line":1,"context_end_line":184,"code":"import os\nimport glob\nimport argparse\nfrom tqdm import tqdm\nimport PIL\nfrom PIL import Image\nimport torch\nimport torch.distributed as dist\nfrom torch.utils.data import Dataset\nimport cv2\nfrom llama.llama_adapter import LLaMA_adapter\n\nDATA_DIR = \"./MME_Benchmark_release_version\"\n\ndef get_image(image):\n    if type(image) is str:\n        try:\n            return Image.open(image).convert(\"RGB\")\n        except Exception as e:\n            print(f\"Fail to read image: {image}\")\n            exit(-1)\n    elif type(image) is Image.Image:\n        return image\n    elif type(image) is PIL.JpegImagePlugin.JpegImageFile:\n        return image\n    elif type(image) is PIL.PngImagePlugin.PngImageFile:\n        return image\n    elif type(image) is PIL.MpoImagePlugin.MpoImageFile:\n        return image\n    else:\n        raise NotImplementedError(f\"Invalid type of Image: {type(image)}\")\n\n\nclass MMEDataset(Dataset):\n    def __init__(\n        self,\n        dataset_name\n    ):\n        self.dataset_name = dataset_name\n        self.dataset = []\n        jpg_sets = [\"artwork\", \"celebrity\", \"color\", \"count\", \"existence\", \"landmark\", \"OCR\", \"position\", \"posters\", \"scene\"]\n        png_sets = [\"code_reasoning\", \"commonsense_reasoning\", \"numerical_calculation\", \"text_translation\"]\n        image_suffix = '.jpg' if dataset_name in jpg_sets else \".png\"\n\n        assert (dataset_name in jpg_sets) or (dataset_name in png_sets), f\"Invalid dataset name for MME benchmark: {dataset_name}\"\n\n        if os.path.exists(f\"{DATA_DIR}/{dataset_name}/images\") and os.path.exists(f\"{DATA_DIR}/{dataset_name}/questions_answers_YN\"):\n            question_files = os.listdir(f\"{DATA_DIR}/{dataset_name}/questions_answers_YN\")\n            for question_file in question_files:\n                image_file_name = os.path.join(DATA_DIR, dataset_name, \"images\", question_file.replace('.txt', image_suffix))\n                with open(os.path.join(DATA_DIR, dataset_name, \"questions_answers_YN\", question_file), 'r', encoding='utf-8') as f:\n                    for line in f.readlines():\n                        try:\n                            question, gt_answer = line.replace('\\n', '').split('\\t')\n                            self.dataset.append({\n                                \"image_path\": image_file_name,\n                                \"gt_answers\": gt_answer,\n                                \"question\": question\n                            })\n                        except:\n                            pass\n\n        else:\n            question_files = glob.glob(f\"{DATA_DIR}/{dataset_name}/*.txt\")\n            for question_file in question_files:\n                image_file_name = question_file.replace(\".txt\", image_suffix)\n                with open(question_file, 'r', encoding='utf-8') as f:\n                    for line in f.readlines():\n                        try:\n                            question, gt_answer = line.replace('\\n', '').split('\\t')\n                            self.dataset.append({\n                                \"image_path\": image_file_name,\n                                \"gt_answers\": gt_answer,\n                                \"question\": question\n                            })\n                        except:\n                            pass\n\n    def __len__(self):\n        return len(self.dataset)\n\n    def __getitem__(self, idx):\n        return self.dataset[idx]\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('Single-turn (conversation) demo', add_help=False)\n    # Model parameters\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='directory containing pre-trained checkpoints')\n    parser.add_argument('--lora', default=16, type=int)\n    parser.add_argument('--output_path', default='/path/to/output_results', type=str)\n    return parser\n\n\nif __name__ == \"__main__\":\n    args = get_args_parser().parse_args()\n\n    device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\n    llama_dir = args.llama_path\n    llama_type = '7B'\n    llama_ckpt_dir = os.path.join(llama_dir, llama_type)\n    llama_tokenzier_path = os.path.join(llama_dir, 'tokenizer.model')\n    \n    model_path = args.pretrained_path\n    # load llama_adapter weights and model_cfg\n    print(f'Loading LLaMA-Adapter from {model_path}')\n    ckpt = torch.load(model_path, map_location='cpu')\n\n    w_bias = True\n    w_lora = args.lora > 0\n    print('Lora:', w_lora)\n    lora_rank = args.lora\n    model = LLaMA_adapter(\n        llama_ckpt_dir, llama_tokenzier_path,\n        max_seq_len=512, max_batch_size=1,\n        clip_model='ViT-L/14',\n        v_embed_dim=768, v_depth=8,\n        v_num_heads=16, v_mlp_ratio=4.0,\n        query_len=10, query_layer=31,\n        w_bias=w_bias,\n        w_lora=w_lora,\n        lora_rank=lora_rank,\n        w_new_gate=w_lora,  # for compatibility\n        phase='finetune')\n\n    load_result = model.load_state_dict(ckpt['model'], strict=False)\n    print(load_result)\n\n    model = model.to(device)\n    model.half()\n    model.eval()\n    preprocess = model.clip_transform\n\n    prompt_format = (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request using a single word or phrase.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    )\n\n    def multi_modal_generate(\n        img_path: str,\n        prompt: str,\n        max_gen_len=30,\n        temperature: float = 0,\n        top_p: float = 0.75,\n    ):\n        img = Image.fromarray(cv2.imread(img_path))\n        img = preprocess(img).unsqueeze(0).half().to(device)\n        prompt = prompt_format.format_map({'instruction': prompt})\n\n        result = model.generate(img, [prompt], \n                                max_gen_len=max_gen_len, \n                                temperature=temperature, \n                                top_p=top_p)\n        return result[0]\n\n\n    result = {}\n    dataset_names = [\"artwork\", \"celebrity\", \"color\", \"count\", \"existence\", \"OCR\", \"position\", \"posters\", \"scene\", \"code_reasoning\", \"commonsense_reasoning\", \"numerical_calculation\", \"text_translation\", \"landmark\"] # landmark (03d5e3bfc958be38.jpg)\n    answer_path = args.output_path\n    batch_size = 1\n\n    print(\"Starting...\")\n    for dataset_name in dataset_names:\n        dataset = MMEDataset(dataset_name)\n\n        predictions = []\n        with torch.no_grad():\n            for data in tqdm(dataset, desc=f\"Inferencing {dataset_name}\"):\n                pred = multi_modal_generate(data['image_path'], data['question'])            \n                predictions.append({'image_path': data['image_path'], 'question': data['question'], 'answer': pred, 'gt_answers': data['gt_answers']})\n\n        os.makedirs(answer_path, exist_ok=True)\n        prediction_file = os.path.join(answer_path, f\"{dataset_name}.txt\")\n        out_datas = [\n            f\"{data['image_path']}\\t{data['question']}\\t{data['gt_answers']}\\t{data['answer']}\"\n            for data in predictions\n        ]\n        with open(prediction_file, 'w') as f:\n            f.write('\\n'.join(out_datas))","source_hash":"d5272f273a26eb7d95fa9871bd0d6cd466b816f5c81cded28a669bda03955961","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.evaluate_mme.get_image","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.evaluate_mme.get_image#L15-L31","kind":"function","name":"get_image","path":"llama_adapter_v2_multimodal7b/util/evaluate_mme.py","language":"python","start_line":15,"end_line":31,"context_start_line":1,"context_end_line":51,"code":"import os\nimport glob\nimport argparse\nfrom tqdm import tqdm\nimport PIL\nfrom PIL import Image\nimport torch\nimport torch.distributed as dist\nfrom torch.utils.data import Dataset\nimport cv2\nfrom llama.llama_adapter import LLaMA_adapter\n\nDATA_DIR = \"./MME_Benchmark_release_version\"\n\ndef get_image(image):\n    if type(image) is str:\n        try:\n            return Image.open(image).convert(\"RGB\")\n        except Exception as e:\n            print(f\"Fail to read image: {image}\")\n            exit(-1)\n    elif type(image) is Image.Image:\n        return image\n    elif type(image) is PIL.JpegImagePlugin.JpegImageFile:\n        return image\n    elif type(image) is PIL.PngImagePlugin.PngImageFile:\n        return image\n    elif type(image) is PIL.MpoImagePlugin.MpoImageFile:\n        return image\n    else:\n        raise NotImplementedError(f\"Invalid type of Image: {type(image)}\")\n\n\nclass MMEDataset(Dataset):\n    def __init__(\n        self,\n        dataset_name\n    ):\n        self.dataset_name = dataset_name\n        self.dataset = []\n        jpg_sets = [\"artwork\", \"celebrity\", \"color\", \"count\", \"existence\", \"landmark\", \"OCR\", \"position\", \"posters\", \"scene\"]\n        png_sets = [\"code_reasoning\", \"commonsense_reasoning\", \"numerical_calculation\", \"text_translation\"]\n        image_suffix = '.jpg' if dataset_name in jpg_sets else \".png\"\n\n        assert (dataset_name in jpg_sets) or (dataset_name in png_sets), f\"Invalid dataset name for MME benchmark: {dataset_name}\"\n\n        if os.path.exists(f\"{DATA_DIR}/{dataset_name}/images\") and os.path.exists(f\"{DATA_DIR}/{dataset_name}/questions_answers_YN\"):\n            question_files = os.listdir(f\"{DATA_DIR}/{dataset_name}/questions_answers_YN\")\n            for question_file in question_files:\n                image_file_name = os.path.join(DATA_DIR, dataset_name, \"images\", question_file.replace('.txt', image_suffix))\n                with open(os.path.join(DATA_DIR, dataset_name, \"questions_answers_YN\", question_file), 'r', encoding='utf-8') as f:","source_hash":"d5272f273a26eb7d95fa9871bd0d6cd466b816f5c81cded28a669bda03955961","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.evaluate_mme.MMEDataset","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.util.evaluate_mme.MMEDataset#L34-L83","kind":"class","name":"MMEDataset","path":"llama_adapter_v2_multimodal7b/util/evaluate_mme.py","language":"python","start_line":34,"end_line":83,"context_start_line":14,"context_end_line":103,"code":"\ndef get_image(image):\n    if type(image) is str:\n        try:\n            return Image.open(image).convert(\"RGB\")\n        except Exception as e:\n            print(f\"Fail to read image: {image}\")\n            exit(-1)\n    elif type(image) is Image.Image:\n        return image\n    elif type(image) is PIL.JpegImagePlugin.JpegImageFile:\n        return image\n    elif type(image) is PIL.PngImagePlugin.PngImageFile:\n        return image\n    elif type(image) is PIL.MpoImagePlugin.MpoImageFile:\n        return image\n    else:\n        raise NotImplementedError(f\"Invalid type of Image: {type(image)}\")\n\n\nclass MMEDataset(Dataset):\n    def __init__(\n        self,\n        dataset_name\n    ):\n        self.dataset_name = dataset_name\n        self.dataset = []\n        jpg_sets = [\"artwork\", \"celebrity\", \"color\", \"count\", \"existence\", \"landmark\", \"OCR\", \"position\", \"posters\", \"scene\"]\n        png_sets = [\"code_reasoning\", \"commonsense_reasoning\", \"numerical_calculation\", \"text_translation\"]\n        image_suffix = '.jpg' if dataset_name in jpg_sets else \".png\"\n\n        assert (dataset_name in jpg_sets) or (dataset_name in png_sets), f\"Invalid dataset name for MME benchmark: {dataset_name}\"\n\n        if os.path.exists(f\"{DATA_DIR}/{dataset_name}/images\") and os.path.exists(f\"{DATA_DIR}/{dataset_name}/questions_answers_YN\"):\n            question_files = os.listdir(f\"{DATA_DIR}/{dataset_name}/questions_answers_YN\")\n            for question_file in question_files:\n                image_file_name = os.path.join(DATA_DIR, dataset_name, \"images\", question_file.replace('.txt', image_suffix))\n                with open(os.path.join(DATA_DIR, dataset_name, \"questions_answers_YN\", question_file), 'r', encoding='utf-8') as f:\n                    for line in f.readlines():\n                        try:\n                            question, gt_answer = line.replace('\\n', '').split('\\t')\n                            self.dataset.append({\n                                \"image_path\": image_file_name,\n                                \"gt_answers\": gt_answer,\n                                \"question\": question\n                            })\n                        except:\n                            pass\n\n        else:\n            question_files = glob.glob(f\"{DATA_DIR}/{dataset_name}/*.txt\")\n            for question_file in question_files:\n                image_file_name = question_file.replace(\".txt\", image_suffix)\n                with open(question_file, 'r', encoding='utf-8') as f:\n                    for line in f.readlines():\n                        try:\n                            question, gt_answer = line.replace('\\n', '').split('\\t')\n                            self.dataset.append({\n                                \"image_path\": image_file_name,\n                                \"gt_answers\": gt_answer,\n                                \"question\": question\n                            })\n                        except:\n                            pass\n\n    def __len__(self):\n        return len(self.dataset)\n\n    def __getitem__(self, idx):\n        return self.dataset[idx]\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('Single-turn (conversation) demo', add_help=False)\n    # Model parameters\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='directory containing pre-trained checkpoints')\n    parser.add_argument('--lora', default=16, type=int)\n    parser.add_argument('--output_path', default='/path/to/output_results', type=str)\n    return parser\n\n\nif __name__ == \"__main__\":\n    args = get_args_parser().parse_args()\n\n    device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\n    llama_dir = args.llama_path","source_hash":"d5272f273a26eb7d95fa9871bd0d6cd466b816f5c81cded28a669bda03955961","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.evaluate_mme.get_args_parser","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.evaluate_mme.get_args_parser#L86-L95","kind":"function","name":"get_args_parser","path":"llama_adapter_v2_multimodal7b/util/evaluate_mme.py","language":"python","start_line":86,"end_line":95,"context_start_line":66,"context_end_line":115,"code":"                image_file_name = question_file.replace(\".txt\", image_suffix)\n                with open(question_file, 'r', encoding='utf-8') as f:\n                    for line in f.readlines():\n                        try:\n                            question, gt_answer = line.replace('\\n', '').split('\\t')\n                            self.dataset.append({\n                                \"image_path\": image_file_name,\n                                \"gt_answers\": gt_answer,\n                                \"question\": question\n                            })\n                        except:\n                            pass\n\n    def __len__(self):\n        return len(self.dataset)\n\n    def __getitem__(self, idx):\n        return self.dataset[idx]\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('Single-turn (conversation) demo', add_help=False)\n    # Model parameters\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='directory containing pre-trained checkpoints')\n    parser.add_argument('--lora', default=16, type=int)\n    parser.add_argument('--output_path', default='/path/to/output_results', type=str)\n    return parser\n\n\nif __name__ == \"__main__\":\n    args = get_args_parser().parse_args()\n\n    device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\n    llama_dir = args.llama_path\n    llama_type = '7B'\n    llama_ckpt_dir = os.path.join(llama_dir, llama_type)\n    llama_tokenzier_path = os.path.join(llama_dir, 'tokenizer.model')\n    \n    model_path = args.pretrained_path\n    # load llama_adapter weights and model_cfg\n    print(f'Loading LLaMA-Adapter from {model_path}')\n    ckpt = torch.load(model_path, map_location='cpu')\n\n    w_bias = True\n    w_lora = args.lora > 0\n    print('Lora:', w_lora)","source_hash":"d5272f273a26eb7d95fa9871bd0d6cd466b816f5c81cded28a669bda03955961","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.evaluate_mme.__init__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.evaluate_mme.__init__#L35-L77","kind":"function","name":"__init__","path":"llama_adapter_v2_multimodal7b/util/evaluate_mme.py","language":"python","start_line":35,"end_line":77,"context_start_line":15,"context_end_line":97,"code":"def get_image(image):\n    if type(image) is str:\n        try:\n            return Image.open(image).convert(\"RGB\")\n        except Exception as e:\n            print(f\"Fail to read image: {image}\")\n            exit(-1)\n    elif type(image) is Image.Image:\n        return image\n    elif type(image) is PIL.JpegImagePlugin.JpegImageFile:\n        return image\n    elif type(image) is PIL.PngImagePlugin.PngImageFile:\n        return image\n    elif type(image) is PIL.MpoImagePlugin.MpoImageFile:\n        return image\n    else:\n        raise NotImplementedError(f\"Invalid type of Image: {type(image)}\")\n\n\nclass MMEDataset(Dataset):\n    def __init__(\n        self,\n        dataset_name\n    ):\n        self.dataset_name = dataset_name\n        self.dataset = []\n        jpg_sets = [\"artwork\", \"celebrity\", \"color\", \"count\", \"existence\", \"landmark\", \"OCR\", \"position\", \"posters\", \"scene\"]\n        png_sets = [\"code_reasoning\", \"commonsense_reasoning\", \"numerical_calculation\", \"text_translation\"]\n        image_suffix = '.jpg' if dataset_name in jpg_sets else \".png\"\n\n        assert (dataset_name in jpg_sets) or (dataset_name in png_sets), f\"Invalid dataset name for MME benchmark: {dataset_name}\"\n\n        if os.path.exists(f\"{DATA_DIR}/{dataset_name}/images\") and os.path.exists(f\"{DATA_DIR}/{dataset_name}/questions_answers_YN\"):\n            question_files = os.listdir(f\"{DATA_DIR}/{dataset_name}/questions_answers_YN\")\n            for question_file in question_files:\n                image_file_name = os.path.join(DATA_DIR, dataset_name, \"images\", question_file.replace('.txt', image_suffix))\n                with open(os.path.join(DATA_DIR, dataset_name, \"questions_answers_YN\", question_file), 'r', encoding='utf-8') as f:\n                    for line in f.readlines():\n                        try:\n                            question, gt_answer = line.replace('\\n', '').split('\\t')\n                            self.dataset.append({\n                                \"image_path\": image_file_name,\n                                \"gt_answers\": gt_answer,\n                                \"question\": question\n                            })\n                        except:\n                            pass\n\n        else:\n            question_files = glob.glob(f\"{DATA_DIR}/{dataset_name}/*.txt\")\n            for question_file in question_files:\n                image_file_name = question_file.replace(\".txt\", image_suffix)\n                with open(question_file, 'r', encoding='utf-8') as f:\n                    for line in f.readlines():\n                        try:\n                            question, gt_answer = line.replace('\\n', '').split('\\t')\n                            self.dataset.append({\n                                \"image_path\": image_file_name,\n                                \"gt_answers\": gt_answer,\n                                \"question\": question\n                            })\n                        except:\n                            pass\n\n    def __len__(self):\n        return len(self.dataset)\n\n    def __getitem__(self, idx):\n        return self.dataset[idx]\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('Single-turn (conversation) demo', add_help=False)\n    # Model parameters\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='directory containing pre-trained checkpoints')\n    parser.add_argument('--lora', default=16, type=int)\n    parser.add_argument('--output_path', default='/path/to/output_results', type=str)\n    return parser\n\n","source_hash":"d5272f273a26eb7d95fa9871bd0d6cd466b816f5c81cded28a669bda03955961","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.evaluate_mme.__len__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.evaluate_mme.__len__#L79-L80","kind":"function","name":"__len__","path":"llama_adapter_v2_multimodal7b/util/evaluate_mme.py","language":"python","start_line":79,"end_line":80,"context_start_line":59,"context_end_line":100,"code":"                            })\n                        except:\n                            pass\n\n        else:\n            question_files = glob.glob(f\"{DATA_DIR}/{dataset_name}/*.txt\")\n            for question_file in question_files:\n                image_file_name = question_file.replace(\".txt\", image_suffix)\n                with open(question_file, 'r', encoding='utf-8') as f:\n                    for line in f.readlines():\n                        try:\n                            question, gt_answer = line.replace('\\n', '').split('\\t')\n                            self.dataset.append({\n                                \"image_path\": image_file_name,\n                                \"gt_answers\": gt_answer,\n                                \"question\": question\n                            })\n                        except:\n                            pass\n\n    def __len__(self):\n        return len(self.dataset)\n\n    def __getitem__(self, idx):\n        return self.dataset[idx]\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('Single-turn (conversation) demo', add_help=False)\n    # Model parameters\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='directory containing pre-trained checkpoints')\n    parser.add_argument('--lora', default=16, type=int)\n    parser.add_argument('--output_path', default='/path/to/output_results', type=str)\n    return parser\n\n\nif __name__ == \"__main__\":\n    args = get_args_parser().parse_args()\n","source_hash":"d5272f273a26eb7d95fa9871bd0d6cd466b816f5c81cded28a669bda03955961","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.evaluate_mme.__getitem__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.evaluate_mme.__getitem__#L82-L83","kind":"function","name":"__getitem__","path":"llama_adapter_v2_multimodal7b/util/evaluate_mme.py","language":"python","start_line":82,"end_line":83,"context_start_line":62,"context_end_line":103,"code":"\n        else:\n            question_files = glob.glob(f\"{DATA_DIR}/{dataset_name}/*.txt\")\n            for question_file in question_files:\n                image_file_name = question_file.replace(\".txt\", image_suffix)\n                with open(question_file, 'r', encoding='utf-8') as f:\n                    for line in f.readlines():\n                        try:\n                            question, gt_answer = line.replace('\\n', '').split('\\t')\n                            self.dataset.append({\n                                \"image_path\": image_file_name,\n                                \"gt_answers\": gt_answer,\n                                \"question\": question\n                            })\n                        except:\n                            pass\n\n    def __len__(self):\n        return len(self.dataset)\n\n    def __getitem__(self, idx):\n        return self.dataset[idx]\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('Single-turn (conversation) demo', add_help=False)\n    # Model parameters\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='directory containing pre-trained checkpoints')\n    parser.add_argument('--lora', default=16, type=int)\n    parser.add_argument('--output_path', default='/path/to/output_results', type=str)\n    return parser\n\n\nif __name__ == \"__main__\":\n    args = get_args_parser().parse_args()\n\n    device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\n    llama_dir = args.llama_path","source_hash":"d5272f273a26eb7d95fa9871bd0d6cd466b816f5c81cded28a669bda03955961","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.util.evaluate_mme.multi_modal_generate","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.util.evaluate_mme.multi_modal_generate#L144-L159","kind":"function","name":"multi_modal_generate","path":"llama_adapter_v2_multimodal7b/util/evaluate_mme.py","language":"python","start_line":144,"end_line":159,"context_start_line":124,"context_end_line":179,"code":"        w_bias=w_bias,\n        w_lora=w_lora,\n        lora_rank=lora_rank,\n        w_new_gate=w_lora,  # for compatibility\n        phase='finetune')\n\n    load_result = model.load_state_dict(ckpt['model'], strict=False)\n    print(load_result)\n\n    model = model.to(device)\n    model.half()\n    model.eval()\n    preprocess = model.clip_transform\n\n    prompt_format = (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request using a single word or phrase.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    )\n\n    def multi_modal_generate(\n        img_path: str,\n        prompt: str,\n        max_gen_len=30,\n        temperature: float = 0,\n        top_p: float = 0.75,\n    ):\n        img = Image.fromarray(cv2.imread(img_path))\n        img = preprocess(img).unsqueeze(0).half().to(device)\n        prompt = prompt_format.format_map({'instruction': prompt})\n\n        result = model.generate(img, [prompt], \n                                max_gen_len=max_gen_len, \n                                temperature=temperature, \n                                top_p=top_p)\n        return result[0]\n\n\n    result = {}\n    dataset_names = [\"artwork\", \"celebrity\", \"color\", \"count\", \"existence\", \"OCR\", \"position\", \"posters\", \"scene\", \"code_reasoning\", \"commonsense_reasoning\", \"numerical_calculation\", \"text_translation\", \"landmark\"] # landmark (03d5e3bfc958be38.jpg)\n    answer_path = args.output_path\n    batch_size = 1\n\n    print(\"Starting...\")\n    for dataset_name in dataset_names:\n        dataset = MMEDataset(dataset_name)\n\n        predictions = []\n        with torch.no_grad():\n            for data in tqdm(dataset, desc=f\"Inferencing {dataset_name}\"):\n                pred = multi_modal_generate(data['image_path'], data['question'])            \n                predictions.append({'image_path': data['image_path'], 'question': data['question'], 'answer': pred, 'gt_answers': data['gt_answers']})\n\n        os.makedirs(answer_path, exist_ok=True)\n        prediction_file = os.path.join(answer_path, f\"{dataset_name}.txt\")\n        out_datas = [","source_hash":"d5272f273a26eb7d95fa9871bd0d6cd466b816f5c81cded28a669bda03955961","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.data.dataset","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.data.dataset#L1-L157","kind":"module","name":"llama_adapter_v2_multimodal7b.data.dataset","path":"llama_adapter_v2_multimodal7b/data/dataset.py","language":"python","start_line":1,"end_line":157,"context_start_line":1,"context_end_line":157,"code":"import torch\nimport yaml\nfrom torch.utils.data import Dataset\nfrom PIL import Image\nimport json\nimport llama.utils\nfrom llama import Tokenizer\nimport copy\nimport torchvision.transforms as transforms\nimport pandas as pd\nimport random\nimport cv2\n\ntry:\n    from torchvision.transforms import InterpolationMode\n    BICUBIC = InterpolationMode.BICUBIC\nexcept ImportError:\n    BICUBIC = Image.BICUBIC\n\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n# create data\ntransform_train = transforms.Compose([\n    transforms.RandomResizedCrop(size=(224, 224), scale=(0.9, 1.0), ratio=(0.75, 1.3333), interpolation=BICUBIC,\n                                 antialias=None),  # 3 is bicubic\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\nclass FinetuneDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        ann = []\n        for meta_path in self.config['META']:\n            meta_l = json.load(open(meta_path))\n            print(f\"{meta_path}: len {len(meta_l)}\")\n            ann += meta_l\n        self.ann = ann\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n        data_item = self.ann[index]\n        if 'image' in data_item.keys():\n            filename = data_item['image']\n            question = data_item['conversations'][0]['value']\n            answer = data_item['conversations'][1]['value']\n     \n            image = cv2.imread(filename)\n            image = Image.fromarray(image)\n            image = self.transform(image)\n            format_instruction = question\n            format_input = None\n        else:\n            image = torch.zeros(3, 224, 224)\n            format_instruction = data_item['instruction'],\n            format_input = data_item['input']\n            answer = data_item['output']\n        input1 = llama.utils.format_prompt(format_instruction, format_input)\n        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image\n\n\nclass PretrainDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        images, captions = [], []\n        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []\n        for x, y in zip(images, captions):\n            self.data_list.append({'url': x, 'caption': y})\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.data_list)\n\n    def __getitem__(self, index):\n        sample = self.data_list[index]\n        image_path, caption = sample['url'], sample['caption']\n        if isinstance(caption, list):\n            caption = random.choice(caption)\n        caption = str(caption)\n\n        image = cv2.imread(image_path)\n        image = Image.fromarray(image)\n        image = self.transform(image)\n\n        format_instruction = \"Generate caption of this image\"\n        input1 = llama.utils.format_prompt(format_instruction, None)\n        input2 = input1 + caption\n\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image","source_hash":"55d2ae8775c52c638ce4f2a13f0f9bfd9921264010bda2eb46998367cbe47b7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.data.dataset.FinetuneDataset","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.data.dataset.FinetuneDataset#L41-L96","kind":"class","name":"FinetuneDataset","path":"llama_adapter_v2_multimodal7b/data/dataset.py","language":"python","start_line":41,"end_line":96,"context_start_line":21,"context_end_line":116,"code":"PROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n# create data\ntransform_train = transforms.Compose([\n    transforms.RandomResizedCrop(size=(224, 224), scale=(0.9, 1.0), ratio=(0.75, 1.3333), interpolation=BICUBIC,\n                                 antialias=None),  # 3 is bicubic\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\nclass FinetuneDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        ann = []\n        for meta_path in self.config['META']:\n            meta_l = json.load(open(meta_path))\n            print(f\"{meta_path}: len {len(meta_l)}\")\n            ann += meta_l\n        self.ann = ann\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n        data_item = self.ann[index]\n        if 'image' in data_item.keys():\n            filename = data_item['image']\n            question = data_item['conversations'][0]['value']\n            answer = data_item['conversations'][1]['value']\n     \n            image = cv2.imread(filename)\n            image = Image.fromarray(image)\n            image = self.transform(image)\n            format_instruction = question\n            format_input = None\n        else:\n            image = torch.zeros(3, 224, 224)\n            format_instruction = data_item['instruction'],\n            format_input = data_item['input']\n            answer = data_item['output']\n        input1 = llama.utils.format_prompt(format_instruction, format_input)\n        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image\n\n\nclass PretrainDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        images, captions = [], []\n        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []","source_hash":"55d2ae8775c52c638ce4f2a13f0f9bfd9921264010bda2eb46998367cbe47b7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.data.dataset.PretrainDataset","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.data.dataset.PretrainDataset#L99-L157","kind":"class","name":"PretrainDataset","path":"llama_adapter_v2_multimodal7b/data/dataset.py","language":"python","start_line":99,"end_line":157,"context_start_line":79,"context_end_line":157,"code":"        input1 = llama.utils.format_prompt(format_instruction, format_input)\n        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image\n\n\nclass PretrainDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        images, captions = [], []\n        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []\n        for x, y in zip(images, captions):\n            self.data_list.append({'url': x, 'caption': y})\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.data_list)\n\n    def __getitem__(self, index):\n        sample = self.data_list[index]\n        image_path, caption = sample['url'], sample['caption']\n        if isinstance(caption, list):\n            caption = random.choice(caption)\n        caption = str(caption)\n\n        image = cv2.imread(image_path)\n        image = Image.fromarray(image)\n        image = self.transform(image)\n\n        format_instruction = \"Generate caption of this image\"\n        input1 = llama.utils.format_prompt(format_instruction, None)\n        input2 = input1 + caption\n\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image","source_hash":"55d2ae8775c52c638ce4f2a13f0f9bfd9921264010bda2eb46998367cbe47b7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.data.dataset.__init__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.data.dataset.__init__#L100-L122","kind":"function","name":"__init__","path":"llama_adapter_v2_multimodal7b/data/dataset.py","language":"python","start_line":100,"end_line":122,"context_start_line":80,"context_end_line":142,"code":"        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image\n\n\nclass PretrainDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        images, captions = [], []\n        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []\n        for x, y in zip(images, captions):\n            self.data_list.append({'url': x, 'caption': y})\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.data_list)\n\n    def __getitem__(self, index):\n        sample = self.data_list[index]\n        image_path, caption = sample['url'], sample['caption']\n        if isinstance(caption, list):\n            caption = random.choice(caption)\n        caption = str(caption)\n\n        image = cv2.imread(image_path)\n        image = Image.fromarray(image)\n        image = self.transform(image)\n\n        format_instruction = \"Generate caption of this image\"\n        input1 = llama.utils.format_prompt(format_instruction, None)\n        input2 = input1 + caption\n\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)","source_hash":"55d2ae8775c52c638ce4f2a13f0f9bfd9921264010bda2eb46998367cbe47b7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.data.dataset.__len__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.data.dataset.__len__#L124-L125","kind":"function","name":"__len__","path":"llama_adapter_v2_multimodal7b/data/dataset.py","language":"python","start_line":124,"end_line":125,"context_start_line":104,"context_end_line":145,"code":"        print(\"DATASET CONFIG:\")\n        print(self.config)\n        images, captions = [], []\n        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []\n        for x, y in zip(images, captions):\n            self.data_list.append({'url': x, 'caption': y})\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.data_list)\n\n    def __getitem__(self, index):\n        sample = self.data_list[index]\n        image_path, caption = sample['url'], sample['caption']\n        if isinstance(caption, list):\n            caption = random.choice(caption)\n        caption = str(caption)\n\n        image = cv2.imread(image_path)\n        image = Image.fromarray(image)\n        image = self.transform(image)\n\n        format_instruction = \"Generate caption of this image\"\n        input1 = llama.utils.format_prompt(format_instruction, None)\n        input2 = input1 + caption\n\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:","source_hash":"55d2ae8775c52c638ce4f2a13f0f9bfd9921264010bda2eb46998367cbe47b7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.data.dataset.__getitem__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.data.dataset.__getitem__#L127-L157","kind":"function","name":"__getitem__","path":"llama_adapter_v2_multimodal7b/data/dataset.py","language":"python","start_line":127,"end_line":157,"context_start_line":107,"context_end_line":157,"code":"        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []\n        for x, y in zip(images, captions):\n            self.data_list.append({'url': x, 'caption': y})\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.data_list)\n\n    def __getitem__(self, index):\n        sample = self.data_list[index]\n        image_path, caption = sample['url'], sample['caption']\n        if isinstance(caption, list):\n            caption = random.choice(caption)\n        caption = str(caption)\n\n        image = cv2.imread(image_path)\n        image = Image.fromarray(image)\n        image = self.transform(image)\n\n        format_instruction = \"Generate caption of this image\"\n        input1 = llama.utils.format_prompt(format_instruction, None)\n        input2 = input1 + caption\n\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image","source_hash":"55d2ae8775c52c638ce4f2a13f0f9bfd9921264010bda2eb46998367cbe47b7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.llama.llama_adapter#L1-L325","kind":"module","name":"llama_adapter_v2_multimodal7b.llama.llama_adapter","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":1,"end_line":325,"context_start_line":1,"context_end_line":325,"code":"import os\nimport json\nfrom pathlib import Path\n\nimport clip\nimport torch\nimport torch.nn as nn\nfrom timm.models.vision_transformer import Block\n\nfrom .llama import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer\nfrom .utils import sample_top_p, _download\n\n\nclass LLaMA_adapter(nn.Module):\n\n    def __init__(self, llama_ckpt_dir, llama_tokenizer,\n                 max_seq_len=512, max_batch_size=1,\n                 clip_model='ViT-L/14',\n                 v_embed_dim=768, v_depth=8,\n                 v_num_heads=16, v_mlp_ratio=4.0,\n                 query_len=10, query_layer=31,\n                 w_bias=False, \n                 w_lora=False, lora_rank=16, \n                 w_new_gate=False,\n                 phase=\"finetune\"):\n        super().__init__()\n\n        # load llama configs\n        with open(os.path.join(llama_ckpt_dir, \"params.json\"), \"r\") as f:\n            params = json.loads(f.read())\n        w_bias = phase == \"finetune\"\n        model_args: ModelArgs = ModelArgs(\n            max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params\n        ) # max_batch_size only affects inferenc\n\n        # 1. clip and clip projector\n        self.clip, self.clip_transform = clip.load(clip_model)\n\n        clip_dim = self.clip.visual.proj.shape[1]\n        self.clip_proj = nn.Linear(clip_dim, v_embed_dim)\n        self.clip_proj_norm = nn.LayerNorm(v_embed_dim)\n\n        self.query_len = query_len\n        self.query_layer = query_layer\n\n        # 2. visual query, blocks and projector\n        self.visual_query = nn.Embedding(query_len, v_embed_dim)\n        self.visual_blocks = nn.ModuleList([\n            Block(v_embed_dim, v_num_heads, v_mlp_ratio, qkv_bias=True)\n            for _ in range(v_depth)])\n        self.visual_proj = nn.Linear(v_embed_dim, model_args.dim)\n        self.visual_proj_norm = nn.LayerNorm(model_args.dim)\n\n        # 3. adapter query\n        self.adapter_query = nn.Embedding(\n            query_len * query_layer, model_args.dim)\n\n        # 4. tokenizer\n        self.tokenizer = Tokenizer(model_path=llama_tokenizer)\n\n        # 5. llama\n        model_args.w_bias = w_bias\n        model_args.w_lora = w_lora\n        model_args.lora_rank = lora_rank\n        model_args.w_new_gate = w_new_gate\n        model_args.vocab_size = self.tokenizer.n_words\n        torch.set_default_tensor_type(torch.cuda.HalfTensor)\n        self.llama = Transformer(model_args)\n        torch.set_default_tensor_type(torch.FloatTensor)\n\n        ckpts = sorted(Path(llama_ckpt_dir).glob(\"*.pth\"))\n        for ckpt in ckpts:\n            ckpt = torch.load(ckpt, map_location='cpu')\n            self.llama.load_state_dict(ckpt, strict=False)\n\n        del self.clip.transformer\n\n         # 6. training criterion\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        # 7. training parameters\n        self.phase = phase\n        self.get_trainable_params(self.phase)\n\n        for name, param in self.named_parameters():\n            if param.requires_grad:\n               print(f\"Trainable param: {name}, {param.shape}, {param.dtype}\")\n\n    def get_trainable_params(self, phase='finetune'):\n        for name, para in self.named_parameters():\n            para.requires_grad = False\n\n        if phase == 'finetune':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'norm' in name or 'bias' in name:\n                        para.data = para.data.float()\n                        para.requires_grad = True\n\n        elif phase == 'pretrain':\n            train_param_name = ['gate', 'clip_proj', 'clip_proj_norm', 'visual_query', 'visual_blocks', 'visual_proj', 'visual_proj_norm', 'adapter_query']\n            for name, para in self.named_parameters():\n                for train_name in train_param_name:\n                    if train_name in name:\n                        para.data = para.data.float()\n                        para.requires_grad = True\n        \n        else:\n            raise ValueError(f\"Unknown model phase: {phase}\")\n        \n    def clip_encode_image(self, x):\n        # modified from CLIP\n        x = self.clip.visual.conv1(x)  # shape = [*, width, grid, grid]\n        # shape = [*, width, grid ** 2]\n        x = x.reshape(x.shape[0], x.shape[1], -1)\n        x = x.permute(0, 2, 1)  # shape = [*, grid ** 2, width]\n        x = torch.cat([self.clip.visual.class_embedding.to(x.dtype) + torch.zeros(x.shape[0], 1,\n                      x.shape[-1], dtype=x.dtype, device=x.device), x], dim=1)  # shape = [*, grid ** 2 + 1, width]\n        x = x + self.clip.visual.positional_embedding.to(x.dtype)\n        x = self.clip.visual.ln_pre(x)\n\n        x = x.permute(1, 0, 2)  # NLD -> LND\n        x = self.clip.visual.transformer(x)\n        x = x.permute(1, 0, 2)  # LND -> NLD\n\n        # preserve all spatial tokens\n        x = self.clip.visual.ln_post(x[:, :, :])\n\n        if self.clip.visual.proj is not None:\n            x = x @ self.clip.visual.proj\n\n        return x\n\n    def forward_visual(self, imgs):\n        clip_feats = self.clip_encode_image(imgs)\n        clip_feats = self.clip_proj_norm(self.clip_proj(clip_feats.float()))\n\n        visual_query = self.visual_query.weight.unsqueeze(\n            0).repeat(len(imgs), 1, 1)\n        visual_query = torch.cat([visual_query, clip_feats], dim=1)\n        for block in self.visual_blocks:\n            visual_query = block(visual_query)\n\n        visual_query = visual_query[:, :self.query_len, :]\n        visual_query = self.visual_proj(visual_query)\n        visual_query = self.visual_proj_norm(visual_query)\n\n        return visual_query\n\n    def forward(self, tokens, labels, imgs):\n        visual_query = self.forward_visual(imgs)\n\n        _bsz, seqlen = tokens.shape\n\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, 0, freqs_cis, mask)\n\n        adapter = self.adapter_query.weight.reshape(self.query_layer, self.query_len, -1).unsqueeze(1)\n        adapter_index = 0\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            dynamic_adapter = adapter[adapter_index].repeat(_bsz, 1, 1)\n            dynamic_adapter = dynamic_adapter + visual_query\n            h = layer(h, 0, freqs_cis, mask, dynamic_adapter)\n            adapter_index = adapter_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n            c_loss = output.mean() * 0\n        else:\n            assert self.llama.vocab_size == 32000\n            c_loss = self.criterion(output.reshape(-1, self.llama.vocab_size), labels.flatten())\n\n        return c_loss, c_loss\n\n    @torch.inference_mode()\n    def forward_inference(self, visual_query, tokens, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[start_pos : start_pos + seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter = self.adapter_query.weight.reshape(self.query_layer, self.query_len, -1).unsqueeze(1)\n        adapter_index = 0\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            dynamic_adapter = adapter[adapter_index].repeat(_bsz, 1, 1)\n            dynamic_adapter = dynamic_adapter + visual_query\n            h = layer(h, start_pos, freqs_cis, mask, dynamic_adapter)\n            adapter_index = adapter_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h[:, -1, :])\n\n        return output.float()\n\n    @torch.inference_mode()\n    def generate(\n        self, imgs, prompts,\n        max_gen_len: int = 256,\n        temperature: float = 0.1,\n        top_p: float = 0.75,\n    ):\n        bsz = len(imgs)\n        params = self.llama.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n        assert len(imgs) == len(prompts)\n\n        with torch.cuda.amp.autocast():\n            visual_query = self.forward_visual(imgs)\n\n        if isinstance(prompts[0], str):\n            prompts = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompts])\n        max_prompt_size = max([len(t) for t in prompts])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n\n        for k, t in enumerate(prompts):\n            tokens[k, : len(t)] = torch.tensor(t).cuda().long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            with torch.cuda.amp.autocast():\n                logits = self.forward_inference(visual_query, tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            # trick: early stop if bsz==1\n            if bsz == 1 and next_token[0] == self.tokenizer.eos_id:\n                break\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n\n            # cut to max gen len\n            t = t[len(prompts[i]): len(prompts[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n\n        return decoded\n\n\n\n_MODELS = {\n    \"BIAS-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/7fa55208379faf2dd862565284101b0e4a2a72114d6490a95e432cf9d9b6c813_BIAS-7B.pth\",\n    \"LORA-BIAS-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/1bcbffc43484332672092e0024a8699a6eb5f558161aebf98a7c6b1db67224d1_LORA-BIAS-7B.pth\",\n    \"CAPTION-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/5088aeb63a89746b90bcfd5cb819e1c7411b2771b267c6d131ce73e250a8abf0_CAPTION-7B.pth\",\n    \"LORA-BIAS-7B-v21\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.1.0/d26d107eec32127ac86ef1997cf7169de1c56a59c539fc1258c6798b969e289c_LORA-BIAS-7B-v21.pth\",\n    # \"LORA16-7B\": \"\",\n    # \"PARTIAL-7B\": \"\"\n}\n\ndef available_models():\n    return list(_MODELS.keys())\n\ndef load(name, llama_dir, llama_type=\"7B\", device=\"cuda\" if torch.cuda.is_available() else \"cpu\", download_root='ckpts', max_seq_len=512,\n        phase=\"finetune\"):\n    if name in _MODELS:\n        model_path = _download(_MODELS[name], download_root)\n    elif os.path.isfile(name):\n        model_path = name\n    else:\n        return RuntimeError(f\"Model {name} not found; available models = {available_models()}\"), None\n\n    # BIAS-7B or https://xxx/sha256_BIAS-7B.pth -> 7B\n    # llama_type = name.split('.')[0].split('-')[-1]\n    llama_ckpt_dir = os.path.join(llama_dir, llama_type)\n    llama_tokenzier_path = os.path.join(llama_dir, 'tokenizer.model')\n\n    # load llama_adapter weights and model_cfg\n    print(f'Loading LLaMA-Adapter from {model_path}')\n    ckpt = torch.load(model_path, map_location='cpu')\n    model_cfg = ckpt.get('config', {})\n\n    model = LLaMA_adapter(\n        llama_ckpt_dir, llama_tokenzier_path,\n        max_seq_len=512, max_batch_size=1,\n        clip_model='ViT-L/14',\n        v_embed_dim=768, v_depth=8,\n        v_num_heads=16, v_mlp_ratio=4.0,\n        query_len=10, query_layer=31,\n        w_bias=model_cfg.get('w_bias', False), \n        w_lora=model_cfg.get('w_lora', False), \n        lora_rank=model_cfg.get('lora_rank', 16),\n        w_new_gate=model_cfg.get('w_lora', False), # for compatibility\n        phase=phase)\n\n    load_result = model.load_state_dict(ckpt['model'], strict=False)\n\n    assert len(load_result.unexpected_keys) == 0, f\"Unexpected keys: {load_result.unexpected_keys}\"\n    return model.to(device), model.clip_transform","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter.LLaMA_adapter","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.llama.llama_adapter.LLaMA_adapter#L15-L274","kind":"class","name":"LLaMA_adapter","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":15,"end_line":274,"context_start_line":1,"context_end_line":294,"code":"import os\nimport json\nfrom pathlib import Path\n\nimport clip\nimport torch\nimport torch.nn as nn\nfrom timm.models.vision_transformer import Block\n\nfrom .llama import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer\nfrom .utils import sample_top_p, _download\n\n\nclass LLaMA_adapter(nn.Module):\n\n    def __init__(self, llama_ckpt_dir, llama_tokenizer,\n                 max_seq_len=512, max_batch_size=1,\n                 clip_model='ViT-L/14',\n                 v_embed_dim=768, v_depth=8,\n                 v_num_heads=16, v_mlp_ratio=4.0,\n                 query_len=10, query_layer=31,\n                 w_bias=False, \n                 w_lora=False, lora_rank=16, \n                 w_new_gate=False,\n                 phase=\"finetune\"):\n        super().__init__()\n\n        # load llama configs\n        with open(os.path.join(llama_ckpt_dir, \"params.json\"), \"r\") as f:\n            params = json.loads(f.read())\n        w_bias = phase == \"finetune\"\n        model_args: ModelArgs = ModelArgs(\n            max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params\n        ) # max_batch_size only affects inferenc\n\n        # 1. clip and clip projector\n        self.clip, self.clip_transform = clip.load(clip_model)\n\n        clip_dim = self.clip.visual.proj.shape[1]\n        self.clip_proj = nn.Linear(clip_dim, v_embed_dim)\n        self.clip_proj_norm = nn.LayerNorm(v_embed_dim)\n\n        self.query_len = query_len\n        self.query_layer = query_layer\n\n        # 2. visual query, blocks and projector\n        self.visual_query = nn.Embedding(query_len, v_embed_dim)\n        self.visual_blocks = nn.ModuleList([\n            Block(v_embed_dim, v_num_heads, v_mlp_ratio, qkv_bias=True)\n            for _ in range(v_depth)])\n        self.visual_proj = nn.Linear(v_embed_dim, model_args.dim)\n        self.visual_proj_norm = nn.LayerNorm(model_args.dim)\n\n        # 3. adapter query\n        self.adapter_query = nn.Embedding(\n            query_len * query_layer, model_args.dim)\n\n        # 4. tokenizer\n        self.tokenizer = Tokenizer(model_path=llama_tokenizer)\n\n        # 5. llama\n        model_args.w_bias = w_bias\n        model_args.w_lora = w_lora\n        model_args.lora_rank = lora_rank\n        model_args.w_new_gate = w_new_gate\n        model_args.vocab_size = self.tokenizer.n_words\n        torch.set_default_tensor_type(torch.cuda.HalfTensor)\n        self.llama = Transformer(model_args)\n        torch.set_default_tensor_type(torch.FloatTensor)\n\n        ckpts = sorted(Path(llama_ckpt_dir).glob(\"*.pth\"))\n        for ckpt in ckpts:\n            ckpt = torch.load(ckpt, map_location='cpu')\n            self.llama.load_state_dict(ckpt, strict=False)\n\n        del self.clip.transformer\n\n         # 6. training criterion\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        # 7. training parameters\n        self.phase = phase\n        self.get_trainable_params(self.phase)\n\n        for name, param in self.named_parameters():\n            if param.requires_grad:\n               print(f\"Trainable param: {name}, {param.shape}, {param.dtype}\")\n\n    def get_trainable_params(self, phase='finetune'):\n        for name, para in self.named_parameters():\n            para.requires_grad = False\n\n        if phase == 'finetune':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'norm' in name or 'bias' in name:\n                        para.data = para.data.float()\n                        para.requires_grad = True\n\n        elif phase == 'pretrain':\n            train_param_name = ['gate', 'clip_proj', 'clip_proj_norm', 'visual_query', 'visual_blocks', 'visual_proj', 'visual_proj_norm', 'adapter_query']\n            for name, para in self.named_parameters():\n                for train_name in train_param_name:\n                    if train_name in name:\n                        para.data = para.data.float()\n                        para.requires_grad = True\n        \n        else:\n            raise ValueError(f\"Unknown model phase: {phase}\")\n        \n    def clip_encode_image(self, x):\n        # modified from CLIP\n        x = self.clip.visual.conv1(x)  # shape = [*, width, grid, grid]\n        # shape = [*, width, grid ** 2]\n        x = x.reshape(x.shape[0], x.shape[1], -1)\n        x = x.permute(0, 2, 1)  # shape = [*, grid ** 2, width]\n        x = torch.cat([self.clip.visual.class_embedding.to(x.dtype) + torch.zeros(x.shape[0], 1,\n                      x.shape[-1], dtype=x.dtype, device=x.device), x], dim=1)  # shape = [*, grid ** 2 + 1, width]\n        x = x + self.clip.visual.positional_embedding.to(x.dtype)\n        x = self.clip.visual.ln_pre(x)\n\n        x = x.permute(1, 0, 2)  # NLD -> LND\n        x = self.clip.visual.transformer(x)\n        x = x.permute(1, 0, 2)  # LND -> NLD\n\n        # preserve all spatial tokens\n        x = self.clip.visual.ln_post(x[:, :, :])\n\n        if self.clip.visual.proj is not None:\n            x = x @ self.clip.visual.proj\n\n        return x\n\n    def forward_visual(self, imgs):\n        clip_feats = self.clip_encode_image(imgs)\n        clip_feats = self.clip_proj_norm(self.clip_proj(clip_feats.float()))\n\n        visual_query = self.visual_query.weight.unsqueeze(\n            0).repeat(len(imgs), 1, 1)\n        visual_query = torch.cat([visual_query, clip_feats], dim=1)\n        for block in self.visual_blocks:\n            visual_query = block(visual_query)\n\n        visual_query = visual_query[:, :self.query_len, :]\n        visual_query = self.visual_proj(visual_query)\n        visual_query = self.visual_proj_norm(visual_query)\n\n        return visual_query\n\n    def forward(self, tokens, labels, imgs):\n        visual_query = self.forward_visual(imgs)\n\n        _bsz, seqlen = tokens.shape\n\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, 0, freqs_cis, mask)\n\n        adapter = self.adapter_query.weight.reshape(self.query_layer, self.query_len, -1).unsqueeze(1)\n        adapter_index = 0\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            dynamic_adapter = adapter[adapter_index].repeat(_bsz, 1, 1)\n            dynamic_adapter = dynamic_adapter + visual_query\n            h = layer(h, 0, freqs_cis, mask, dynamic_adapter)\n            adapter_index = adapter_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n            c_loss = output.mean() * 0\n        else:\n            assert self.llama.vocab_size == 32000\n            c_loss = self.criterion(output.reshape(-1, self.llama.vocab_size), labels.flatten())\n\n        return c_loss, c_loss\n\n    @torch.inference_mode()\n    def forward_inference(self, visual_query, tokens, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[start_pos : start_pos + seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter = self.adapter_query.weight.reshape(self.query_layer, self.query_len, -1).unsqueeze(1)\n        adapter_index = 0\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            dynamic_adapter = adapter[adapter_index].repeat(_bsz, 1, 1)\n            dynamic_adapter = dynamic_adapter + visual_query\n            h = layer(h, start_pos, freqs_cis, mask, dynamic_adapter)\n            adapter_index = adapter_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h[:, -1, :])\n\n        return output.float()\n\n    @torch.inference_mode()\n    def generate(\n        self, imgs, prompts,\n        max_gen_len: int = 256,\n        temperature: float = 0.1,\n        top_p: float = 0.75,\n    ):\n        bsz = len(imgs)\n        params = self.llama.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n        assert len(imgs) == len(prompts)\n\n        with torch.cuda.amp.autocast():\n            visual_query = self.forward_visual(imgs)\n\n        if isinstance(prompts[0], str):\n            prompts = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompts])\n        max_prompt_size = max([len(t) for t in prompts])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n\n        for k, t in enumerate(prompts):\n            tokens[k, : len(t)] = torch.tensor(t).cuda().long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            with torch.cuda.amp.autocast():\n                logits = self.forward_inference(visual_query, tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            # trick: early stop if bsz==1\n            if bsz == 1 and next_token[0] == self.tokenizer.eos_id:\n                break\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n\n            # cut to max gen len\n            t = t[len(prompts[i]): len(prompts[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n\n        return decoded\n\n\n\n_MODELS = {\n    \"BIAS-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/7fa55208379faf2dd862565284101b0e4a2a72114d6490a95e432cf9d9b6c813_BIAS-7B.pth\",\n    \"LORA-BIAS-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/1bcbffc43484332672092e0024a8699a6eb5f558161aebf98a7c6b1db67224d1_LORA-BIAS-7B.pth\",\n    \"CAPTION-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/5088aeb63a89746b90bcfd5cb819e1c7411b2771b267c6d131ce73e250a8abf0_CAPTION-7B.pth\",\n    \"LORA-BIAS-7B-v21\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.1.0/d26d107eec32127ac86ef1997cf7169de1c56a59c539fc1258c6798b969e289c_LORA-BIAS-7B-v21.pth\",\n    # \"LORA16-7B\": \"\",\n    # \"PARTIAL-7B\": \"\"\n}\n\ndef available_models():\n    return list(_MODELS.keys())\n\ndef load(name, llama_dir, llama_type=\"7B\", device=\"cuda\" if torch.cuda.is_available() else \"cpu\", download_root='ckpts', max_seq_len=512,\n        phase=\"finetune\"):\n    if name in _MODELS:\n        model_path = _download(_MODELS[name], download_root)\n    elif os.path.isfile(name):","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter.available_models","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama_adapter.available_models#L287-L288","kind":"function","name":"available_models","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":287,"end_line":288,"context_start_line":267,"context_end_line":308,"code":"            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n\n        return decoded\n\n\n\n_MODELS = {\n    \"BIAS-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/7fa55208379faf2dd862565284101b0e4a2a72114d6490a95e432cf9d9b6c813_BIAS-7B.pth\",\n    \"LORA-BIAS-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/1bcbffc43484332672092e0024a8699a6eb5f558161aebf98a7c6b1db67224d1_LORA-BIAS-7B.pth\",\n    \"CAPTION-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/5088aeb63a89746b90bcfd5cb819e1c7411b2771b267c6d131ce73e250a8abf0_CAPTION-7B.pth\",\n    \"LORA-BIAS-7B-v21\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.1.0/d26d107eec32127ac86ef1997cf7169de1c56a59c539fc1258c6798b969e289c_LORA-BIAS-7B-v21.pth\",\n    # \"LORA16-7B\": \"\",\n    # \"PARTIAL-7B\": \"\"\n}\n\ndef available_models():\n    return list(_MODELS.keys())\n\ndef load(name, llama_dir, llama_type=\"7B\", device=\"cuda\" if torch.cuda.is_available() else \"cpu\", download_root='ckpts', max_seq_len=512,\n        phase=\"finetune\"):\n    if name in _MODELS:\n        model_path = _download(_MODELS[name], download_root)\n    elif os.path.isfile(name):\n        model_path = name\n    else:\n        return RuntimeError(f\"Model {name} not found; available models = {available_models()}\"), None\n\n    # BIAS-7B or https://xxx/sha256_BIAS-7B.pth -> 7B\n    # llama_type = name.split('.')[0].split('-')[-1]\n    llama_ckpt_dir = os.path.join(llama_dir, llama_type)\n    llama_tokenzier_path = os.path.join(llama_dir, 'tokenizer.model')\n\n    # load llama_adapter weights and model_cfg\n    print(f'Loading LLaMA-Adapter from {model_path}')\n    ckpt = torch.load(model_path, map_location='cpu')\n    model_cfg = ckpt.get('config', {})\n","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter.load","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama_adapter.load#L290-L325","kind":"function","name":"load","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":290,"end_line":325,"context_start_line":270,"context_end_line":325,"code":"            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n\n        return decoded\n\n\n\n_MODELS = {\n    \"BIAS-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/7fa55208379faf2dd862565284101b0e4a2a72114d6490a95e432cf9d9b6c813_BIAS-7B.pth\",\n    \"LORA-BIAS-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/1bcbffc43484332672092e0024a8699a6eb5f558161aebf98a7c6b1db67224d1_LORA-BIAS-7B.pth\",\n    \"CAPTION-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/5088aeb63a89746b90bcfd5cb819e1c7411b2771b267c6d131ce73e250a8abf0_CAPTION-7B.pth\",\n    \"LORA-BIAS-7B-v21\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.1.0/d26d107eec32127ac86ef1997cf7169de1c56a59c539fc1258c6798b969e289c_LORA-BIAS-7B-v21.pth\",\n    # \"LORA16-7B\": \"\",\n    # \"PARTIAL-7B\": \"\"\n}\n\ndef available_models():\n    return list(_MODELS.keys())\n\ndef load(name, llama_dir, llama_type=\"7B\", device=\"cuda\" if torch.cuda.is_available() else \"cpu\", download_root='ckpts', max_seq_len=512,\n        phase=\"finetune\"):\n    if name in _MODELS:\n        model_path = _download(_MODELS[name], download_root)\n    elif os.path.isfile(name):\n        model_path = name\n    else:\n        return RuntimeError(f\"Model {name} not found; available models = {available_models()}\"), None\n\n    # BIAS-7B or https://xxx/sha256_BIAS-7B.pth -> 7B\n    # llama_type = name.split('.')[0].split('-')[-1]\n    llama_ckpt_dir = os.path.join(llama_dir, llama_type)\n    llama_tokenzier_path = os.path.join(llama_dir, 'tokenizer.model')\n\n    # load llama_adapter weights and model_cfg\n    print(f'Loading LLaMA-Adapter from {model_path}')\n    ckpt = torch.load(model_path, map_location='cpu')\n    model_cfg = ckpt.get('config', {})\n\n    model = LLaMA_adapter(\n        llama_ckpt_dir, llama_tokenzier_path,\n        max_seq_len=512, max_batch_size=1,\n        clip_model='ViT-L/14',\n        v_embed_dim=768, v_depth=8,\n        v_num_heads=16, v_mlp_ratio=4.0,\n        query_len=10, query_layer=31,\n        w_bias=model_cfg.get('w_bias', False), \n        w_lora=model_cfg.get('w_lora', False), \n        lora_rank=model_cfg.get('lora_rank', 16),\n        w_new_gate=model_cfg.get('w_lora', False), # for compatibility\n        phase=phase)\n\n    load_result = model.load_state_dict(ckpt['model'], strict=False)\n\n    assert len(load_result.unexpected_keys) == 0, f\"Unexpected keys: {load_result.unexpected_keys}\"\n    return model.to(device), model.clip_transform","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter.__init__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama_adapter.__init__#L17-L88","kind":"function","name":"__init__","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":17,"end_line":88,"context_start_line":1,"context_end_line":108,"code":"import os\nimport json\nfrom pathlib import Path\n\nimport clip\nimport torch\nimport torch.nn as nn\nfrom timm.models.vision_transformer import Block\n\nfrom .llama import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer\nfrom .utils import sample_top_p, _download\n\n\nclass LLaMA_adapter(nn.Module):\n\n    def __init__(self, llama_ckpt_dir, llama_tokenizer,\n                 max_seq_len=512, max_batch_size=1,\n                 clip_model='ViT-L/14',\n                 v_embed_dim=768, v_depth=8,\n                 v_num_heads=16, v_mlp_ratio=4.0,\n                 query_len=10, query_layer=31,\n                 w_bias=False, \n                 w_lora=False, lora_rank=16, \n                 w_new_gate=False,\n                 phase=\"finetune\"):\n        super().__init__()\n\n        # load llama configs\n        with open(os.path.join(llama_ckpt_dir, \"params.json\"), \"r\") as f:\n            params = json.loads(f.read())\n        w_bias = phase == \"finetune\"\n        model_args: ModelArgs = ModelArgs(\n            max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params\n        ) # max_batch_size only affects inferenc\n\n        # 1. clip and clip projector\n        self.clip, self.clip_transform = clip.load(clip_model)\n\n        clip_dim = self.clip.visual.proj.shape[1]\n        self.clip_proj = nn.Linear(clip_dim, v_embed_dim)\n        self.clip_proj_norm = nn.LayerNorm(v_embed_dim)\n\n        self.query_len = query_len\n        self.query_layer = query_layer\n\n        # 2. visual query, blocks and projector\n        self.visual_query = nn.Embedding(query_len, v_embed_dim)\n        self.visual_blocks = nn.ModuleList([\n            Block(v_embed_dim, v_num_heads, v_mlp_ratio, qkv_bias=True)\n            for _ in range(v_depth)])\n        self.visual_proj = nn.Linear(v_embed_dim, model_args.dim)\n        self.visual_proj_norm = nn.LayerNorm(model_args.dim)\n\n        # 3. adapter query\n        self.adapter_query = nn.Embedding(\n            query_len * query_layer, model_args.dim)\n\n        # 4. tokenizer\n        self.tokenizer = Tokenizer(model_path=llama_tokenizer)\n\n        # 5. llama\n        model_args.w_bias = w_bias\n        model_args.w_lora = w_lora\n        model_args.lora_rank = lora_rank\n        model_args.w_new_gate = w_new_gate\n        model_args.vocab_size = self.tokenizer.n_words\n        torch.set_default_tensor_type(torch.cuda.HalfTensor)\n        self.llama = Transformer(model_args)\n        torch.set_default_tensor_type(torch.FloatTensor)\n\n        ckpts = sorted(Path(llama_ckpt_dir).glob(\"*.pth\"))\n        for ckpt in ckpts:\n            ckpt = torch.load(ckpt, map_location='cpu')\n            self.llama.load_state_dict(ckpt, strict=False)\n\n        del self.clip.transformer\n\n         # 6. training criterion\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        # 7. training parameters\n        self.phase = phase\n        self.get_trainable_params(self.phase)\n\n        for name, param in self.named_parameters():\n            if param.requires_grad:\n               print(f\"Trainable param: {name}, {param.shape}, {param.dtype}\")\n\n    def get_trainable_params(self, phase='finetune'):\n        for name, para in self.named_parameters():\n            para.requires_grad = False\n\n        if phase == 'finetune':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'norm' in name or 'bias' in name:\n                        para.data = para.data.float()\n                        para.requires_grad = True\n\n        elif phase == 'pretrain':\n            train_param_name = ['gate', 'clip_proj', 'clip_proj_norm', 'visual_query', 'visual_blocks', 'visual_proj', 'visual_proj_norm', 'adapter_query']\n            for name, para in self.named_parameters():\n                for train_name in train_param_name:\n                    if train_name in name:\n                        para.data = para.data.float()\n                        para.requires_grad = True\n        ","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter.get_trainable_params","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama_adapter.get_trainable_params#L90-L110","kind":"function","name":"get_trainable_params","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":90,"end_line":110,"context_start_line":70,"context_end_line":130,"code":"        torch.set_default_tensor_type(torch.FloatTensor)\n\n        ckpts = sorted(Path(llama_ckpt_dir).glob(\"*.pth\"))\n        for ckpt in ckpts:\n            ckpt = torch.load(ckpt, map_location='cpu')\n            self.llama.load_state_dict(ckpt, strict=False)\n\n        del self.clip.transformer\n\n         # 6. training criterion\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        # 7. training parameters\n        self.phase = phase\n        self.get_trainable_params(self.phase)\n\n        for name, param in self.named_parameters():\n            if param.requires_grad:\n               print(f\"Trainable param: {name}, {param.shape}, {param.dtype}\")\n\n    def get_trainable_params(self, phase='finetune'):\n        for name, para in self.named_parameters():\n            para.requires_grad = False\n\n        if phase == 'finetune':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'norm' in name or 'bias' in name:\n                        para.data = para.data.float()\n                        para.requires_grad = True\n\n        elif phase == 'pretrain':\n            train_param_name = ['gate', 'clip_proj', 'clip_proj_norm', 'visual_query', 'visual_blocks', 'visual_proj', 'visual_proj_norm', 'adapter_query']\n            for name, para in self.named_parameters():\n                for train_name in train_param_name:\n                    if train_name in name:\n                        para.data = para.data.float()\n                        para.requires_grad = True\n        \n        else:\n            raise ValueError(f\"Unknown model phase: {phase}\")\n        \n    def clip_encode_image(self, x):\n        # modified from CLIP\n        x = self.clip.visual.conv1(x)  # shape = [*, width, grid, grid]\n        # shape = [*, width, grid ** 2]\n        x = x.reshape(x.shape[0], x.shape[1], -1)\n        x = x.permute(0, 2, 1)  # shape = [*, grid ** 2, width]\n        x = torch.cat([self.clip.visual.class_embedding.to(x.dtype) + torch.zeros(x.shape[0], 1,\n                      x.shape[-1], dtype=x.dtype, device=x.device), x], dim=1)  # shape = [*, grid ** 2 + 1, width]\n        x = x + self.clip.visual.positional_embedding.to(x.dtype)\n        x = self.clip.visual.ln_pre(x)\n\n        x = x.permute(1, 0, 2)  # NLD -> LND\n        x = self.clip.visual.transformer(x)\n        x = x.permute(1, 0, 2)  # LND -> NLD\n\n        # preserve all spatial tokens\n        x = self.clip.visual.ln_post(x[:, :, :])\n\n        if self.clip.visual.proj is not None:","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter.clip_encode_image","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama_adapter.clip_encode_image#L112-L133","kind":"function","name":"clip_encode_image","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":112,"end_line":133,"context_start_line":92,"context_end_line":153,"code":"            para.requires_grad = False\n\n        if phase == 'finetune':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'norm' in name or 'bias' in name:\n                        para.data = para.data.float()\n                        para.requires_grad = True\n\n        elif phase == 'pretrain':\n            train_param_name = ['gate', 'clip_proj', 'clip_proj_norm', 'visual_query', 'visual_blocks', 'visual_proj', 'visual_proj_norm', 'adapter_query']\n            for name, para in self.named_parameters():\n                for train_name in train_param_name:\n                    if train_name in name:\n                        para.data = para.data.float()\n                        para.requires_grad = True\n        \n        else:\n            raise ValueError(f\"Unknown model phase: {phase}\")\n        \n    def clip_encode_image(self, x):\n        # modified from CLIP\n        x = self.clip.visual.conv1(x)  # shape = [*, width, grid, grid]\n        # shape = [*, width, grid ** 2]\n        x = x.reshape(x.shape[0], x.shape[1], -1)\n        x = x.permute(0, 2, 1)  # shape = [*, grid ** 2, width]\n        x = torch.cat([self.clip.visual.class_embedding.to(x.dtype) + torch.zeros(x.shape[0], 1,\n                      x.shape[-1], dtype=x.dtype, device=x.device), x], dim=1)  # shape = [*, grid ** 2 + 1, width]\n        x = x + self.clip.visual.positional_embedding.to(x.dtype)\n        x = self.clip.visual.ln_pre(x)\n\n        x = x.permute(1, 0, 2)  # NLD -> LND\n        x = self.clip.visual.transformer(x)\n        x = x.permute(1, 0, 2)  # LND -> NLD\n\n        # preserve all spatial tokens\n        x = self.clip.visual.ln_post(x[:, :, :])\n\n        if self.clip.visual.proj is not None:\n            x = x @ self.clip.visual.proj\n\n        return x\n\n    def forward_visual(self, imgs):\n        clip_feats = self.clip_encode_image(imgs)\n        clip_feats = self.clip_proj_norm(self.clip_proj(clip_feats.float()))\n\n        visual_query = self.visual_query.weight.unsqueeze(\n            0).repeat(len(imgs), 1, 1)\n        visual_query = torch.cat([visual_query, clip_feats], dim=1)\n        for block in self.visual_blocks:\n            visual_query = block(visual_query)\n\n        visual_query = visual_query[:, :self.query_len, :]\n        visual_query = self.visual_proj(visual_query)\n        visual_query = self.visual_proj_norm(visual_query)\n\n        return visual_query\n\n    def forward(self, tokens, labels, imgs):\n        visual_query = self.forward_visual(imgs)\n","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter.forward_visual","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama_adapter.forward_visual#L135-L149","kind":"function","name":"forward_visual","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":135,"end_line":149,"context_start_line":115,"context_end_line":169,"code":"        # shape = [*, width, grid ** 2]\n        x = x.reshape(x.shape[0], x.shape[1], -1)\n        x = x.permute(0, 2, 1)  # shape = [*, grid ** 2, width]\n        x = torch.cat([self.clip.visual.class_embedding.to(x.dtype) + torch.zeros(x.shape[0], 1,\n                      x.shape[-1], dtype=x.dtype, device=x.device), x], dim=1)  # shape = [*, grid ** 2 + 1, width]\n        x = x + self.clip.visual.positional_embedding.to(x.dtype)\n        x = self.clip.visual.ln_pre(x)\n\n        x = x.permute(1, 0, 2)  # NLD -> LND\n        x = self.clip.visual.transformer(x)\n        x = x.permute(1, 0, 2)  # LND -> NLD\n\n        # preserve all spatial tokens\n        x = self.clip.visual.ln_post(x[:, :, :])\n\n        if self.clip.visual.proj is not None:\n            x = x @ self.clip.visual.proj\n\n        return x\n\n    def forward_visual(self, imgs):\n        clip_feats = self.clip_encode_image(imgs)\n        clip_feats = self.clip_proj_norm(self.clip_proj(clip_feats.float()))\n\n        visual_query = self.visual_query.weight.unsqueeze(\n            0).repeat(len(imgs), 1, 1)\n        visual_query = torch.cat([visual_query, clip_feats], dim=1)\n        for block in self.visual_blocks:\n            visual_query = block(visual_query)\n\n        visual_query = visual_query[:, :self.query_len, :]\n        visual_query = self.visual_proj(visual_query)\n        visual_query = self.visual_proj_norm(visual_query)\n\n        return visual_query\n\n    def forward(self, tokens, labels, imgs):\n        visual_query = self.forward_visual(imgs)\n\n        _bsz, seqlen = tokens.shape\n\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, 0, freqs_cis, mask)\n\n        adapter = self.adapter_query.weight.reshape(self.query_layer, self.query_len, -1).unsqueeze(1)\n        adapter_index = 0\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            dynamic_adapter = adapter[adapter_index].repeat(_bsz, 1, 1)","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter.forward","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama_adapter.forward#L151-L185","kind":"function","name":"forward","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":151,"end_line":185,"context_start_line":131,"context_end_line":205,"code":"            x = x @ self.clip.visual.proj\n\n        return x\n\n    def forward_visual(self, imgs):\n        clip_feats = self.clip_encode_image(imgs)\n        clip_feats = self.clip_proj_norm(self.clip_proj(clip_feats.float()))\n\n        visual_query = self.visual_query.weight.unsqueeze(\n            0).repeat(len(imgs), 1, 1)\n        visual_query = torch.cat([visual_query, clip_feats], dim=1)\n        for block in self.visual_blocks:\n            visual_query = block(visual_query)\n\n        visual_query = visual_query[:, :self.query_len, :]\n        visual_query = self.visual_proj(visual_query)\n        visual_query = self.visual_proj_norm(visual_query)\n\n        return visual_query\n\n    def forward(self, tokens, labels, imgs):\n        visual_query = self.forward_visual(imgs)\n\n        _bsz, seqlen = tokens.shape\n\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, 0, freqs_cis, mask)\n\n        adapter = self.adapter_query.weight.reshape(self.query_layer, self.query_len, -1).unsqueeze(1)\n        adapter_index = 0\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            dynamic_adapter = adapter[adapter_index].repeat(_bsz, 1, 1)\n            dynamic_adapter = dynamic_adapter + visual_query\n            h = layer(h, 0, freqs_cis, mask, dynamic_adapter)\n            adapter_index = adapter_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n            c_loss = output.mean() * 0\n        else:\n            assert self.llama.vocab_size == 32000\n            c_loss = self.criterion(output.reshape(-1, self.llama.vocab_size), labels.flatten())\n\n        return c_loss, c_loss\n\n    @torch.inference_mode()\n    def forward_inference(self, visual_query, tokens, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[start_pos : start_pos + seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter = self.adapter_query.weight.reshape(self.query_layer, self.query_len, -1).unsqueeze(1)\n        adapter_index = 0\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            dynamic_adapter = adapter[adapter_index].repeat(_bsz, 1, 1)\n            dynamic_adapter = dynamic_adapter + visual_query\n            h = layer(h, start_pos, freqs_cis, mask, dynamic_adapter)","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter.forward_inference","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama_adapter.forward_inference#L188-L211","kind":"function","name":"forward_inference","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":188,"end_line":211,"context_start_line":168,"context_end_line":231,"code":"        for layer in self.llama.layers[-1 * self.query_layer:]:\n            dynamic_adapter = adapter[adapter_index].repeat(_bsz, 1, 1)\n            dynamic_adapter = dynamic_adapter + visual_query\n            h = layer(h, 0, freqs_cis, mask, dynamic_adapter)\n            adapter_index = adapter_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n            c_loss = output.mean() * 0\n        else:\n            assert self.llama.vocab_size == 32000\n            c_loss = self.criterion(output.reshape(-1, self.llama.vocab_size), labels.flatten())\n\n        return c_loss, c_loss\n\n    @torch.inference_mode()\n    def forward_inference(self, visual_query, tokens, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[start_pos : start_pos + seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter = self.adapter_query.weight.reshape(self.query_layer, self.query_len, -1).unsqueeze(1)\n        adapter_index = 0\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            dynamic_adapter = adapter[adapter_index].repeat(_bsz, 1, 1)\n            dynamic_adapter = dynamic_adapter + visual_query\n            h = layer(h, start_pos, freqs_cis, mask, dynamic_adapter)\n            adapter_index = adapter_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h[:, -1, :])\n\n        return output.float()\n\n    @torch.inference_mode()\n    def generate(\n        self, imgs, prompts,\n        max_gen_len: int = 256,\n        temperature: float = 0.1,\n        top_p: float = 0.75,\n    ):\n        bsz = len(imgs)\n        params = self.llama.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n        assert len(imgs) == len(prompts)\n\n        with torch.cuda.amp.autocast():\n            visual_query = self.forward_visual(imgs)\n\n        if isinstance(prompts[0], str):\n            prompts = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompts])","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama_adapter.generate","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama_adapter.generate#L214-L274","kind":"function","name":"generate","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":214,"end_line":274,"context_start_line":194,"context_end_line":294,"code":"        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter = self.adapter_query.weight.reshape(self.query_layer, self.query_len, -1).unsqueeze(1)\n        adapter_index = 0\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            dynamic_adapter = adapter[adapter_index].repeat(_bsz, 1, 1)\n            dynamic_adapter = dynamic_adapter + visual_query\n            h = layer(h, start_pos, freqs_cis, mask, dynamic_adapter)\n            adapter_index = adapter_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h[:, -1, :])\n\n        return output.float()\n\n    @torch.inference_mode()\n    def generate(\n        self, imgs, prompts,\n        max_gen_len: int = 256,\n        temperature: float = 0.1,\n        top_p: float = 0.75,\n    ):\n        bsz = len(imgs)\n        params = self.llama.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n        assert len(imgs) == len(prompts)\n\n        with torch.cuda.amp.autocast():\n            visual_query = self.forward_visual(imgs)\n\n        if isinstance(prompts[0], str):\n            prompts = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompts])\n        max_prompt_size = max([len(t) for t in prompts])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n\n        for k, t in enumerate(prompts):\n            tokens[k, : len(t)] = torch.tensor(t).cuda().long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            with torch.cuda.amp.autocast():\n                logits = self.forward_inference(visual_query, tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            # trick: early stop if bsz==1\n            if bsz == 1 and next_token[0] == self.tokenizer.eos_id:\n                break\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n\n            # cut to max gen len\n            t = t[len(prompts[i]): len(prompts[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n\n        return decoded\n\n\n\n_MODELS = {\n    \"BIAS-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/7fa55208379faf2dd862565284101b0e4a2a72114d6490a95e432cf9d9b6c813_BIAS-7B.pth\",\n    \"LORA-BIAS-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/1bcbffc43484332672092e0024a8699a6eb5f558161aebf98a7c6b1db67224d1_LORA-BIAS-7B.pth\",\n    \"CAPTION-7B\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.0.0/5088aeb63a89746b90bcfd5cb819e1c7411b2771b267c6d131ce73e250a8abf0_CAPTION-7B.pth\",\n    \"LORA-BIAS-7B-v21\": \"https://github.com/OpenGVLab/LLaMA-Adapter/releases/download/v.2.1.0/d26d107eec32127ac86ef1997cf7169de1c56a59c539fc1258c6798b969e289c_LORA-BIAS-7B-v21.pth\",\n    # \"LORA16-7B\": \"\",\n    # \"PARTIAL-7B\": \"\"\n}\n\ndef available_models():\n    return list(_MODELS.keys())\n\ndef load(name, llama_dir, llama_type=\"7B\", device=\"cuda\" if torch.cuda.is_available() else \"cpu\", download_root='ckpts', max_seq_len=512,\n        phase=\"finetune\"):\n    if name in _MODELS:\n        model_path = _download(_MODELS[name], download_root)\n    elif os.path.isfile(name):","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.llama.llama#L1-L324","kind":"module","name":"llama_adapter_v2_multimodal7b.llama.llama","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":1,"end_line":324,"context_start_line":1,"context_end_line":324,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\n\nimport torch\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\nimport torch.nn.functional as F\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    w_bias: bool = False # use bias tuning\n    w_lora: bool = False # use lora tuning\n    lora_rank: int = 16\n    w_new_gate: bool = False # for compatibility\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n        self.args = args\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=args.w_bias\n        )\n        self.wk = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wv = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wo = Linear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=args.w_bias\n        )\n        if args.w_bias:\n            nn.init.constant_(self.wq.bias.data, 0)\n            nn.init.constant_(self.wo.bias.data, 0)\n\n        self.w_lora = args.w_lora\n        if args.w_lora:\n           self.lora_wq_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wq_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wk_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wk_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wv_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wv_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wo_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wo_l2 = Linear(args.lora_rank, args.dim, bias=False)\n           nn.init.constant_(self.lora_wq_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wk_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wv_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wo_l2.weight.data, 0)\n\n        self.cache_k = None\n        self.cache_v = None\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n        \n        self.w_new_gate = args.w_new_gate\n        if args.w_new_gate:\n            self.new_gate = torch.nn.Parameter(torch.ones(1, 1, 1, 1))\n\n\n    def train(self, mode: bool = True):\n        if mode:\n            self.cache_k = None\n            self.cache_v = None\n        else:\n            self.cache_k = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n            self.cache_v = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n        return super().train(mode)\n\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n        if self.w_lora:\n           xq = xq + self.lora_wq_l2(self.lora_wq_l1(x))\n           xk = xk + self.lora_wk_l2(self.lora_wk_l1(x))\n           xv = xv + self.lora_wv_l2(self.lora_wv_l1(x))\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if not self.training:\n            self.cache_k = self.cache_k.to(xq)\n            self.cache_v = self.cache_v.to(xq)\n\n            self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk\n            self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n            keys = self.cache_k[:bsz, : start_pos + seqlen]\n            values = self.cache_v[:bsz, : start_pos + seqlen]\n        else:\n            assert start_pos==0\n            keys = xk\n            values = xv\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_v = self.wv(adapter).view(bsz, adapter_len, self.n_local_heads, self.head_dim)\n            adapter_v = adapter_v.transpose(1, 2)\n\n            if adapter_len > 1:\n                adapter_k = self.wk(adapter).view(bsz, adapter_len, self.n_local_heads, self.head_dim)\n                adapter_k = adapter_k.transpose(1, 2)\n\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n\n        if adapter is not None:\n            if adapter_len > 1:\n                adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n                adapter_scores = self.gate.tanh() * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n                if self.w_new_gate:\n                    adapter_scores = self.new_gate * adapter_scores\n                output = output + torch.matmul(adapter_scores, adapter_v)\n            else:\n                output = output + self.gate.tanh() * adapter_v\n\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        if self.w_lora:\n           return self.wo(output) + self.lora_wo_l2(self.lora_wo_l1(output))\n        else:\n           return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n        args: ModelArgs\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=args.w_bias\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=args.w_bias\n        )\n        self.w3 = Linear(\n            dim, hidden_dim, bias=args.w_bias\n        )\n        if args.w_bias:\n            nn.init.constant_(self.w1.bias.data, 0)\n            nn.init.constant_(self.w2.bias.data, 0)\n            nn.init.constant_(self.w3.bias.data, 0)\n\n        self.w_lora = args.w_lora\n        if args.w_lora:\n           self.lora_w1_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w1_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           self.lora_w2_l1 = Linear(hidden_dim, args.lora_rank, bias=False)\n           self.lora_w2_l2 = Linear(args.lora_rank, dim, bias=False)\n           self.lora_w3_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w3_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           nn.init.constant_(self.lora_w1_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w2_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w3_l2.weight.data, 0)\n\n    def forward(self, x):\n        if self.w_lora:\n           out = F.silu(self.w1(x) + self.lora_w1_l2(self.lora_w1_l1(x))) * (self.w3(x) + self.lora_w3_l2(self.lora_w3_l1(x)))\n           return self.w2(out) + self.lora_w2_l2(self.lora_w2_l1(out))\n        else:\n           return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, args=args\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(\n            params.dim, params.vocab_size, bias=False\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.ModelArgs","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.llama.llama.ModelArgs#L15-L29","kind":"class","name":"ModelArgs","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":15,"end_line":29,"context_start_line":1,"context_end_line":49,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\n\nimport torch\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\nimport torch.nn.functional as F\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    w_bias: bool = False # use bias tuning\n    w_lora: bool = False # use lora tuning\n    lora_rank: int = 16\n    w_new_gate: bool = False # for compatibility\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.RMSNorm","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.llama.llama.RMSNorm#L32-L43","kind":"class","name":"RMSNorm","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":32,"end_line":43,"context_start_line":12,"context_end_line":63,"code":"\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    w_bias: bool = False # use bias tuning\n    w_lora: bool = False # use lora tuning\n    lora_rank: int = 16\n    w_new_gate: bool = False # for compatibility\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.precompute_freqs_cis","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama.precompute_freqs_cis#L46-L51","kind":"function","name":"precompute_freqs_cis","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":46,"end_line":51,"context_start_line":26,"context_end_line":71,"code":"    w_bias: bool = False # use bias tuning\n    w_lora: bool = False # use lora tuning\n    lora_rank: int = 16\n    w_new_gate: bool = False # for compatibility\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.reshape_for_broadcast","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama.reshape_for_broadcast#L54-L59","kind":"function","name":"reshape_for_broadcast","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":54,"end_line":59,"context_start_line":34,"context_end_line":79,"code":"        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n        self.args = args\n","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.apply_rotary_emb","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama.apply_rotary_emb#L62-L72","kind":"function","name":"apply_rotary_emb","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":62,"end_line":72,"context_start_line":42,"context_end_line":92,"code":"        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n        self.args = args\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=args.w_bias\n        )\n        self.wk = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.Attention","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.llama.llama.Attention#L75-L215","kind":"class","name":"Attention","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":75,"end_line":215,"context_start_line":55,"context_end_line":235,"code":"    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n        self.args = args\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=args.w_bias\n        )\n        self.wk = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wv = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wo = Linear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=args.w_bias\n        )\n        if args.w_bias:\n            nn.init.constant_(self.wq.bias.data, 0)\n            nn.init.constant_(self.wo.bias.data, 0)\n\n        self.w_lora = args.w_lora\n        if args.w_lora:\n           self.lora_wq_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wq_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wk_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wk_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wv_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wv_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wo_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wo_l2 = Linear(args.lora_rank, args.dim, bias=False)\n           nn.init.constant_(self.lora_wq_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wk_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wv_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wo_l2.weight.data, 0)\n\n        self.cache_k = None\n        self.cache_v = None\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n        \n        self.w_new_gate = args.w_new_gate\n        if args.w_new_gate:\n            self.new_gate = torch.nn.Parameter(torch.ones(1, 1, 1, 1))\n\n\n    def train(self, mode: bool = True):\n        if mode:\n            self.cache_k = None\n            self.cache_v = None\n        else:\n            self.cache_k = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n            self.cache_v = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n        return super().train(mode)\n\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n        if self.w_lora:\n           xq = xq + self.lora_wq_l2(self.lora_wq_l1(x))\n           xk = xk + self.lora_wk_l2(self.lora_wk_l1(x))\n           xv = xv + self.lora_wv_l2(self.lora_wv_l1(x))\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if not self.training:\n            self.cache_k = self.cache_k.to(xq)\n            self.cache_v = self.cache_v.to(xq)\n\n            self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk\n            self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n            keys = self.cache_k[:bsz, : start_pos + seqlen]\n            values = self.cache_v[:bsz, : start_pos + seqlen]\n        else:\n            assert start_pos==0\n            keys = xk\n            values = xv\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_v = self.wv(adapter).view(bsz, adapter_len, self.n_local_heads, self.head_dim)\n            adapter_v = adapter_v.transpose(1, 2)\n\n            if adapter_len > 1:\n                adapter_k = self.wk(adapter).view(bsz, adapter_len, self.n_local_heads, self.head_dim)\n                adapter_k = adapter_k.transpose(1, 2)\n\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n\n        if adapter is not None:\n            if adapter_len > 1:\n                adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n                adapter_scores = self.gate.tanh() * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n                if self.w_new_gate:\n                    adapter_scores = self.new_gate * adapter_scores\n                output = output + torch.matmul(adapter_scores, adapter_v)\n            else:\n                output = output + self.gate.tanh() * adapter_v\n\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        if self.w_lora:\n           return self.wo(output) + self.lora_wo_l2(self.lora_wo_l1(output))\n        else:\n           return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n        args: ModelArgs\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=args.w_bias\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=args.w_bias\n        )","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.FeedForward","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.llama.llama.FeedForward#L218-L261","kind":"class","name":"FeedForward","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":218,"end_line":261,"context_start_line":198,"context_end_line":281,"code":"        if adapter is not None:\n            if adapter_len > 1:\n                adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n                adapter_scores = self.gate.tanh() * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n                if self.w_new_gate:\n                    adapter_scores = self.new_gate * adapter_scores\n                output = output + torch.matmul(adapter_scores, adapter_v)\n            else:\n                output = output + self.gate.tanh() * adapter_v\n\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        if self.w_lora:\n           return self.wo(output) + self.lora_wo_l2(self.lora_wo_l1(output))\n        else:\n           return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n        args: ModelArgs\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=args.w_bias\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=args.w_bias\n        )\n        self.w3 = Linear(\n            dim, hidden_dim, bias=args.w_bias\n        )\n        if args.w_bias:\n            nn.init.constant_(self.w1.bias.data, 0)\n            nn.init.constant_(self.w2.bias.data, 0)\n            nn.init.constant_(self.w3.bias.data, 0)\n\n        self.w_lora = args.w_lora\n        if args.w_lora:\n           self.lora_w1_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w1_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           self.lora_w2_l1 = Linear(hidden_dim, args.lora_rank, bias=False)\n           self.lora_w2_l2 = Linear(args.lora_rank, dim, bias=False)\n           self.lora_w3_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w3_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           nn.init.constant_(self.lora_w1_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w2_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w3_l2.weight.data, 0)\n\n    def forward(self, x):\n        if self.w_lora:\n           out = F.silu(self.w1(x) + self.lora_w1_l2(self.lora_w1_l1(x))) * (self.w3(x) + self.lora_w3_l2(self.lora_w3_l1(x)))\n           return self.w2(out) + self.lora_w2_l2(self.lora_w2_l1(out))\n        else:\n           return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, args=args\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.TransformerBlock","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.llama.llama.TransformerBlock#L264-L282","kind":"class","name":"TransformerBlock","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":264,"end_line":282,"context_start_line":244,"context_end_line":302,"code":"        self.w_lora = args.w_lora\n        if args.w_lora:\n           self.lora_w1_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w1_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           self.lora_w2_l1 = Linear(hidden_dim, args.lora_rank, bias=False)\n           self.lora_w2_l2 = Linear(args.lora_rank, dim, bias=False)\n           self.lora_w3_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w3_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           nn.init.constant_(self.lora_w1_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w2_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w3_l2.weight.data, 0)\n\n    def forward(self, x):\n        if self.w_lora:\n           out = F.silu(self.w1(x) + self.lora_w1_l2(self.lora_w1_l1(x))) * (self.w3(x) + self.lora_w3_l2(self.lora_w3_l1(x)))\n           return self.w2(out) + self.lora_w2_l2(self.lora_w2_l1(out))\n        else:\n           return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, args=args\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(\n            params.dim, params.vocab_size, bias=False\n        )","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.Transformer","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.llama.llama.Transformer#L285-L324","kind":"class","name":"Transformer","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":285,"end_line":324,"context_start_line":265,"context_end_line":324,"code":"    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, args=args\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(\n            params.dim, params.vocab_size, bias=False\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.__init__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama.__init__#L286-L306","kind":"function","name":"__init__","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":286,"end_line":306,"context_start_line":266,"context_end_line":324,"code":"        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, args=args\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(\n            params.dim, params.vocab_size, bias=False\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama._norm","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama._norm#L38-L39","kind":"function","name":"_norm","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":38,"end_line":39,"context_start_line":18,"context_end_line":59,"code":"    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    w_bias: bool = False # use bias tuning\n    w_lora: bool = False # use lora tuning\n    lora_rank: int = 16\n    w_new_gate: bool = False # for compatibility\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.forward","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama.forward#L309-L324","kind":"function","name":"forward","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":309,"end_line":324,"context_start_line":289,"context_end_line":324,"code":"        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(\n            params.dim, params.vocab_size, bias=False\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.llama.train","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.llama.train#L135-L146","kind":"function","name":"train","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":135,"end_line":146,"context_start_line":115,"context_end_line":166,"code":"           self.lora_wv_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wv_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wo_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wo_l2 = Linear(args.lora_rank, args.dim, bias=False)\n           nn.init.constant_(self.lora_wq_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wk_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wv_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wo_l2.weight.data, 0)\n\n        self.cache_k = None\n        self.cache_v = None\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n        \n        self.w_new_gate = args.w_new_gate\n        if args.w_new_gate:\n            self.new_gate = torch.nn.Parameter(torch.ones(1, 1, 1, 1))\n\n\n    def train(self, mode: bool = True):\n        if mode:\n            self.cache_k = None\n            self.cache_v = None\n        else:\n            self.cache_k = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n            self.cache_v = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n        return super().train(mode)\n\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n        if self.w_lora:\n           xq = xq + self.lora_wq_l2(self.lora_wq_l1(x))\n           xk = xk + self.lora_wk_l2(self.lora_wk_l1(x))\n           xv = xv + self.lora_wv_l2(self.lora_wv_l1(x))\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if not self.training:\n            self.cache_k = self.cache_k.to(xq)\n            self.cache_v = self.cache_v.to(xq)\n","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.utils","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.llama.utils#L1-L71","kind":"module","name":"llama_adapter_v2_multimodal7b.llama.utils","path":"llama_adapter_v2_multimodal7b/llama/utils.py","language":"python","start_line":1,"end_line":71,"context_start_line":1,"context_end_line":71,"code":"import os\nimport urllib\nimport hashlib\nimport warnings\n\nfrom tqdm import tqdm\nimport torch\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token\n\n\ndef format_prompt(instruction, input=None):\n\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None:\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\n\ndef _download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    # assume the url is https://some/path/sha256_model.pth\n    expected_sha256 = url.split(\"/\")[-1].split('_')[0]\n    # expected_sha256 = url.split(\"/\")[-2]\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        if hashlib.sha256(open(download_target, \"rb\").read()).hexdigest() == expected_sha256:\n            return download_target\n        else:\n            warnings.warn(f\"{download_target} exists, but the SHA256 checksum does not match; re-downloading the file\")\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:\n                    break\n\n                output.write(buffer)\n                loop.update(len(buffer))\n\n    if hashlib.sha256(open(download_target, \"rb\").read()).hexdigest() != expected_sha256:\n        raise RuntimeError(\"Model has been downloaded but the SHA256 checksum does not not match\")\n\n    return download_target","source_hash":"72a4d5400ca4d22079d19105a2a641bbf30f652005b5e7c6888b7106f87aae4c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.utils.sample_top_p","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.utils.sample_top_p#L10-L18","kind":"function","name":"sample_top_p","path":"llama_adapter_v2_multimodal7b/llama/utils.py","language":"python","start_line":10,"end_line":18,"context_start_line":1,"context_end_line":38,"code":"import os\nimport urllib\nimport hashlib\nimport warnings\n\nfrom tqdm import tqdm\nimport torch\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token\n\n\ndef format_prompt(instruction, input=None):\n\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None:\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})","source_hash":"72a4d5400ca4d22079d19105a2a641bbf30f652005b5e7c6888b7106f87aae4c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.utils.format_prompt","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.utils.format_prompt#L21-L38","kind":"function","name":"format_prompt","path":"llama_adapter_v2_multimodal7b/llama/utils.py","language":"python","start_line":21,"end_line":38,"context_start_line":1,"context_end_line":58,"code":"import os\nimport urllib\nimport hashlib\nimport warnings\n\nfrom tqdm import tqdm\nimport torch\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token\n\n\ndef format_prompt(instruction, input=None):\n\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None:\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\n\ndef _download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    # assume the url is https://some/path/sha256_model.pth\n    expected_sha256 = url.split(\"/\")[-1].split('_')[0]\n    # expected_sha256 = url.split(\"/\")[-2]\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        if hashlib.sha256(open(download_target, \"rb\").read()).hexdigest() == expected_sha256:\n            return download_target\n        else:\n            warnings.warn(f\"{download_target} exists, but the SHA256 checksum does not match; re-downloading the file\")\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:","source_hash":"72a4d5400ca4d22079d19105a2a641bbf30f652005b5e7c6888b7106f87aae4c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.utils._download","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.utils._download#L41-L71","kind":"function","name":"_download","path":"llama_adapter_v2_multimodal7b/llama/utils.py","language":"python","start_line":41,"end_line":71,"context_start_line":21,"context_end_line":71,"code":"def format_prompt(instruction, input=None):\n\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None:\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\n\ndef _download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    # assume the url is https://some/path/sha256_model.pth\n    expected_sha256 = url.split(\"/\")[-1].split('_')[0]\n    # expected_sha256 = url.split(\"/\")[-2]\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        if hashlib.sha256(open(download_target, \"rb\").read()).hexdigest() == expected_sha256:\n            return download_target\n        else:\n            warnings.warn(f\"{download_target} exists, but the SHA256 checksum does not match; re-downloading the file\")\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:\n                    break\n\n                output.write(buffer)\n                loop.update(len(buffer))\n\n    if hashlib.sha256(open(download_target, \"rb\").read()).hexdigest() != expected_sha256:\n        raise RuntimeError(\"Model has been downloaded but the SHA256 checksum does not not match\")\n\n    return download_target","source_hash":"72a4d5400ca4d22079d19105a2a641bbf30f652005b5e7c6888b7106f87aae4c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.tokenizer","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_multimodal7b.llama.tokenizer#L1-L40","kind":"module","name":"llama_adapter_v2_multimodal7b.llama.tokenizer","path":"llama_adapter_v2_multimodal7b/llama/tokenizer.py","language":"python","start_line":1,"end_line":40,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.tokenizer.Tokenizer","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_multimodal7b.llama.tokenizer.Tokenizer#L13-L40","kind":"class","name":"Tokenizer","path":"llama_adapter_v2_multimodal7b/llama/tokenizer.py","language":"python","start_line":13,"end_line":40,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.tokenizer.__init__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.tokenizer.__init__#L14-L28","kind":"function","name":"__init__","path":"llama_adapter_v2_multimodal7b/llama/tokenizer.py","language":"python","start_line":14,"end_line":28,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.tokenizer.encode","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.tokenizer.encode#L30-L37","kind":"function","name":"encode","path":"llama_adapter_v2_multimodal7b/llama/tokenizer.py","language":"python","start_line":30,"end_line":37,"context_start_line":10,"context_end_line":40,"code":"logger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_multimodal7b.llama.tokenizer.decode","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_multimodal7b.llama.tokenizer.decode#L39-L40","kind":"function","name":"decode","path":"llama_adapter_v2_multimodal7b/llama/tokenizer.py","language":"python","start_line":39,"end_line":40,"context_start_line":19,"context_end_line":40,"code":"\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.extract_adapter_from_checkpoint","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.extract_adapter_from_checkpoint#L1-L15","kind":"module","name":"alpaca_finetuning_v1.extract_adapter_from_checkpoint","path":"alpaca_finetuning_v1/extract_adapter_from_checkpoint.py","language":"python","start_line":1,"end_line":15,"context_start_line":1,"context_end_line":15,"code":"import torch\n\nmodel = torch.load(\"./checkpoint/checkpoint-4.pth\", map_location=\"cpu\")\nnew_model = dict()\nweight_list = [\"layers.\" + str(i) + \".attention.gate\" for i in range(32)]\nold_weight_list = [\"layers.\" + str(i) + \".attention.gate\" for i in range(32)]\nweight_list = weight_list + [\"adapter_query.weight\"]\n\nprint(weight_list)\nprint(model[\"model\"][\"adapter_query.weight\"].shape)\n\nfor i in range(len(weight_list)):\n    new_model[weight_list[i]] = model[\"model\"][weight_list[i]]\n\ntorch.save(new_model, \"adapter_adapter_len10_layer30_epoch5.pth\")","source_hash":"9adf0123c67c13105f937d04e39a03742783601f02d8483d772abaf3d222ffbc","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.models_llama_adapter","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.models_llama_adapter#L1-L50","kind":"module","name":"alpaca_finetuning_v1.models_llama_adapter","path":"alpaca_finetuning_v1/models_llama_adapter.py","language":"python","start_line":1,"end_line":50,"context_start_line":1,"context_end_line":50,"code":"import json\n\nimport torch\n\nfrom llama import ModelArgs, Tokenizer, Transformer\n\n\ndef Llama7B_adapter(args, **kwargs):\n\n    llama_model_path = args.llama_model_path\n    model_name = \"7B\"\n\n    checkpoint = torch.load(llama_model_path + model_name + \"/consolidated.00.pth\", map_location=\"cpu\")\n    print(llama_model_path + model_name + \"/consolidated.00.pth\")\n\n    with open(llama_model_path + model_name + \"/params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=args.max_seq_len,\n        max_batch_size=32,\n        adapter_len=args.adapter_len,\n        adapter_layer=args.adapter_layer,\n        **params\n    )\n    tokenizer = Tokenizer(model_path=llama_model_path + \"/tokenizer.model\")\n\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model_llama_adapter = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    model_llama_adapter.load_state_dict(checkpoint, strict=False)\n\n    for name, param in model_llama_adapter.named_parameters():\n        if \"adapter\" not in name:\n            param.requires_grad = False\n        else:\n            param.requires_grad = True\n            param.data = param.data.float()\n\n    for name, param in model_llama_adapter.layers[-1 * args.adapter_layer :].named_parameters():\n        if \"gate\" in name or \"adapter\" in name:\n            param.data = param.data.float()\n            param.requires_grad = True\n\n    return model_llama_adapter\n\n\n# set recommended archs\nLlama7B_adapter = Llama7B_adapter","source_hash":"0890bdd4d67183d4e58df19168966750005c5cc195ba3cfb50884e31dcc24d7a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.models_llama_adapter.Llama7B_adapter","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.models_llama_adapter.Llama7B_adapter#L8-L46","kind":"function","name":"Llama7B_adapter","path":"alpaca_finetuning_v1/models_llama_adapter.py","language":"python","start_line":8,"end_line":46,"context_start_line":1,"context_end_line":50,"code":"import json\n\nimport torch\n\nfrom llama import ModelArgs, Tokenizer, Transformer\n\n\ndef Llama7B_adapter(args, **kwargs):\n\n    llama_model_path = args.llama_model_path\n    model_name = \"7B\"\n\n    checkpoint = torch.load(llama_model_path + model_name + \"/consolidated.00.pth\", map_location=\"cpu\")\n    print(llama_model_path + model_name + \"/consolidated.00.pth\")\n\n    with open(llama_model_path + model_name + \"/params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=args.max_seq_len,\n        max_batch_size=32,\n        adapter_len=args.adapter_len,\n        adapter_layer=args.adapter_layer,\n        **params\n    )\n    tokenizer = Tokenizer(model_path=llama_model_path + \"/tokenizer.model\")\n\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model_llama_adapter = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    model_llama_adapter.load_state_dict(checkpoint, strict=False)\n\n    for name, param in model_llama_adapter.named_parameters():\n        if \"adapter\" not in name:\n            param.requires_grad = False\n        else:\n            param.requires_grad = True\n            param.data = param.data.float()\n\n    for name, param in model_llama_adapter.layers[-1 * args.adapter_layer :].named_parameters():\n        if \"gate\" in name or \"adapter\" in name:\n            param.data = param.data.float()\n            param.requires_grad = True\n\n    return model_llama_adapter\n\n\n# set recommended archs\nLlama7B_adapter = Llama7B_adapter","source_hash":"0890bdd4d67183d4e58df19168966750005c5cc195ba3cfb50884e31dcc24d7a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.engine_finetuning","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.engine_finetuning#L1-L132","kind":"module","name":"alpaca_finetuning_v1.engine_finetuning","path":"alpaca_finetuning_v1/engine_finetuning.py","language":"python","start_line":1,"end_line":132,"context_start_line":1,"context_end_line":132,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\nimport util.lr_sched as lr_sched\nimport util.misc as misc\n\n\ndef train_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n\n    model.train(True)\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter(\"lr\", misc.SmoothedValue(window_size=1, fmt=\"{value:.6f}\"))\n    header = \"Epoch: [{}]\".format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print(\"log_dir: {}\".format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header)\n    ):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        c_loss = model(examples, labels)\n        loss = c_loss\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n\n        loss_scaler(loss, optimizer, parameters=model.parameters(), update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\"We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar(\"c_train_loss\", c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar(\"lr\", lr, epoch_1000x)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n\ndef val_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter(\"lr\", misc.SmoothedValue(window_size=1, fmt=\"{value:.6f}\"))\n    header = \"Epoch: [{}]\".format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    if log_writer is not None:\n        print(\"log_dir: {}\".format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header)\n    ):\n\n        with torch.no_grad():\n            c_loss = model(examples, labels)\n        loss = c_loss\n        loss_value = loss.item()\n\n        c_loss_value = c_loss.item()\n\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        metric_logger.update(closs=c_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\"We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar(\"c_train_loss\", c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar(\"lr\", lr, epoch_1000x)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"e90ffdf70e0a5bc1217c32533638a96c91df1ec263f2805e7eecde3789ceb5a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.engine_finetuning.train_one_epoch","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.engine_finetuning.train_one_epoch#L10-L76","kind":"function","name":"train_one_epoch","path":"alpaca_finetuning_v1/engine_finetuning.py","language":"python","start_line":10,"end_line":76,"context_start_line":1,"context_end_line":96,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\nimport util.lr_sched as lr_sched\nimport util.misc as misc\n\n\ndef train_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n\n    model.train(True)\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter(\"lr\", misc.SmoothedValue(window_size=1, fmt=\"{value:.6f}\"))\n    header = \"Epoch: [{}]\".format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print(\"log_dir: {}\".format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header)\n    ):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        c_loss = model(examples, labels)\n        loss = c_loss\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n\n        loss_scaler(loss, optimizer, parameters=model.parameters(), update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\"We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar(\"c_train_loss\", c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar(\"lr\", lr, epoch_1000x)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n\ndef val_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter(\"lr\", misc.SmoothedValue(window_size=1, fmt=\"{value:.6f}\"))\n    header = \"Epoch: [{}]\".format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n","source_hash":"e90ffdf70e0a5bc1217c32533638a96c91df1ec263f2805e7eecde3789ceb5a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.engine_finetuning.val_one_epoch","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.engine_finetuning.val_one_epoch#L79-L132","kind":"function","name":"val_one_epoch","path":"alpaca_finetuning_v1/engine_finetuning.py","language":"python","start_line":79,"end_line":132,"context_start_line":59,"context_end_line":132,"code":"        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\"We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar(\"c_train_loss\", c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar(\"lr\", lr, epoch_1000x)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n\ndef val_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter(\"lr\", misc.SmoothedValue(window_size=1, fmt=\"{value:.6f}\"))\n    header = \"Epoch: [{}]\".format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    if log_writer is not None:\n        print(\"log_dir: {}\".format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header)\n    ):\n\n        with torch.no_grad():\n            c_loss = model(examples, labels)\n        loss = c_loss\n        loss_value = loss.item()\n\n        c_loss_value = c_loss.item()\n\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        metric_logger.update(closs=c_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\"We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar(\"c_train_loss\", c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar(\"lr\", lr, epoch_1000x)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"e90ffdf70e0a5bc1217c32533638a96c91df1ec263f2805e7eecde3789ceb5a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.finetuning","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.finetuning#L1-L294","kind":"module","name":"alpaca_finetuning_v1.finetuning","path":"alpaca_finetuning_v1/finetuning.py","language":"python","start_line":1,"end_line":294,"context_start_line":1,"context_end_line":294,"code":"import argparse\nimport copy\nimport datetime\nimport json\nimport os\nimport time\nfrom pathlib import Path\n\nimport models_llama_adapter\nimport numpy as np\nimport timm.optim.optim_factory as optim_factory\nimport torch\nimport torch.backends.cudnn as cudnn\nimport util.misc as misc\nfrom engine_finetuning import train_one_epoch, val_one_epoch\nfrom torch.utils.data import Dataset\nfrom torch.utils.tensorboard import SummaryWriter\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\n\nfrom llama import Tokenizer\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\nclass InstructionDataset(Dataset):\n    def __init__(self, data_path, model_path, max_words=30, partition=\"train\"):\n        self.ann = json.load(open(data_path))\n        if partition == \"train\":\n            self.ann = self.ann\n        else:\n            self.ann = self.ann[:200]\n\n        self.max_words = max_words\n        tokenizer = Tokenizer(model_path=model_path + \"./tokenizer.model\")\n        self.tokenizer1 = tokenizer\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n\n        ann = self.ann[index]\n        if ann.get(\"input\", \"\") == \"\":\n            prompt = PROMPT_DICT[\"prompt_no_input\"].format_map(ann)\n        else:\n            prompt = PROMPT_DICT[\"prompt_input\"].format_map(ann)\n        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)\n        labels[: len(prompt)] = -1\n        example_mask = example.ge(0)\n        label_mask = labels.ge(0)\n        example[~example_mask] = 0\n        labels[~label_mask] = 0\n        example_mask = example_mask.float()\n        label_mask = label_mask.float()\n\n        return example, labels, example_mask\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser(\"MAE pre-training\", add_help=False)\n    parser.add_argument(\n        \"--batch_size\",\n        default=64,\n        type=int,\n        help=\"Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus\",\n    )\n    parser.add_argument(\"--epochs\", default=400, type=int)\n    parser.add_argument(\n        \"--accum_iter\",\n        default=1,\n        type=int,\n        help=\"Accumulate gradient iterations (for increasing the effective batch size under memory constraints)\",\n    )\n\n    # Model parameters\n    parser.add_argument(\"--llama_model_path\", default=\"./llama\", type=str, help=\"path of llama model\")\n    parser.add_argument(\"--model\", default=\"llama7B_adapter\", type=str, metavar=\"MODEL\", help=\"Name of model to train\")\n\n    parser.add_argument(\"--adapter_layer\", type=int, default=30, metavar=\"LENGTH\", help=\"the number of adapter layer\")\n\n    parser.add_argument(\"--adapter_len\", type=int, default=10, metavar=\"LENGTH\", help=\"the adapter length\")\n\n    parser.add_argument(\"--max_seq_len\", type=int, default=512, metavar=\"LENGTH\", help=\"the maximum sequence length\")\n\n    # Optimizer parameters\n    parser.add_argument(\"--weight_decay\", type=float, default=0.05, help=\"weight decay (default: 0.05)\")\n\n    parser.add_argument(\"--lr\", type=float, default=None, metavar=\"LR\", help=\"learning rate (absolute lr)\")\n    parser.add_argument(\n        \"--blr\",\n        type=float,\n        default=1e-3,\n        metavar=\"LR\",\n        help=\"base learning rate: absolute_lr = base_lr * total_batch_size / 256\",\n    )\n    parser.add_argument(\n        \"--min_lr\", type=float, default=0.0, metavar=\"LR\", help=\"lower lr bound for cyclic schedulers that hit 0\"\n    )\n\n    parser.add_argument(\"--warmup_epochs\", type=int, default=40, metavar=\"N\", help=\"epochs to warmup LR\")\n\n    # Dataset parameters\n    parser.add_argument(\"--data_path\", default=\"/instruction_dataset/\", type=str, help=\"dataset path\")\n\n    parser.add_argument(\"--output_dir\", default=\"./output_dir\", help=\"path where to save, empty for no saving\")\n    parser.add_argument(\"--log_dir\", default=\"./output_dir\", help=\"path where to tensorboard log\")\n    parser.add_argument(\"--device\", default=\"cuda\", help=\"device to use for training / testing\")\n    parser.add_argument(\"--seed\", default=0, type=int)\n    parser.add_argument(\"--resume\", default=\"\", help=\"resume from checkpoint\")\n\n    parser.add_argument(\"--start_epoch\", default=0, type=int, metavar=\"N\", help=\"start epoch\")\n    parser.add_argument(\"--num_workers\", default=10, type=int)\n    parser.add_argument(\n        \"--pin_mem\",\n        action=\"store_true\",\n        help=\"Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.\",\n    )\n    parser.add_argument(\"--no_pin_mem\", action=\"store_false\", dest=\"pin_mem\")\n    parser.set_defaults(pin_mem=True)\n\n    # distributed training parameters\n    parser.add_argument(\"--world_size\", default=1, type=int, help=\"number of distributed processes\")\n    parser.add_argument(\"--local_rank\", default=-1, type=int)\n    parser.add_argument(\"--dist_on_itp\", action=\"store_true\")\n    parser.add_argument(\"--dist_url\", default=\"env://\", help=\"url used to set up distributed training\")\n\n    return parser\n\n\ndef main(args):\n\n    misc.init_distributed_mode(args)\n\n    print(\"job dir: {}\".format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(\", \", \",\\n\"))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n\n    cudnn.benchmark = True\n\n    dataset_train = InstructionDataset(\n        data_path=args.data_path, model_path=args.llama_model_path, max_words=args.max_seq_len, partition=\"train\"\n    )\n    dataset_val = InstructionDataset(\n        data_path=args.data_path, model_path=args.llama_model_path, max_words=args.max_seq_len, partition=\"val\"\n    )\n\n    print(dataset_train)\n    print(dataset_val)\n\n    if True:  # args.distributed:\n        num_tasks = misc.get_world_size()\n        global_rank = misc.get_rank()\n        sampler_train = torch.utils.data.DistributedSampler(\n            dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True\n        )\n\n        sampler_val = torch.utils.data.DistributedSampler(\n            dataset_val, num_replicas=num_tasks, rank=global_rank, shuffle=True\n        )\n\n        print(\"Sampler_train = %s\" % str(sampler_train))\n    else:\n        sampler_train = torch.utils.data.RandomSampler(dataset_train)\n\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train,\n        sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    data_loader_val = torch.utils.data.DataLoader(\n        dataset_val,\n        sampler=sampler_val,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # define the model\n    model = models_llama_adapter.__dict__[args.model](args)\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = optim_factory.param_groups_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n    misc.load_model(args=args, model_without_ddp=model_without_ddp, optimizer=optimizer, loss_scaler=loss_scaler)\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n            data_loader_val.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train, optimizer, device, epoch, loss_scaler, log_writer=log_writer, args=args\n        )\n\n        val_stats = val_one_epoch(\n            model, data_loader_val, optimizer, device, epoch, loss_scaler, log_writer=log_writer, args=args\n        )\n\n        if args.output_dir and (epoch % 8 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args,\n                model=model,\n                model_without_ddp=model_without_ddp,\n                optimizer=optimizer,\n                loss_scaler=loss_scaler,\n                epoch=epoch,\n            )\n\n        log_stats = {\n            **{f\"train_{k}\": v for k, v in train_stats.items()},\n            \"epoch\": epoch,\n            **{f\"val_{k}\": v for k, v in val_stats.items()},\n        }\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print(\"Training time {}\".format(total_time_str))\n\n\nif __name__ == \"__main__\":\n\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"a0ec94f11bf4623533eb76da6c0ad77e1e0697d4cd35193d37bc7f4e445185d1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.finetuning.InstructionDataset","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.finetuning.InstructionDataset#L36-L75","kind":"class","name":"InstructionDataset","path":"alpaca_finetuning_v1/finetuning.py","language":"python","start_line":36,"end_line":75,"context_start_line":16,"context_end_line":95,"code":"from torch.utils.data import Dataset\nfrom torch.utils.tensorboard import SummaryWriter\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\n\nfrom llama import Tokenizer\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\nclass InstructionDataset(Dataset):\n    def __init__(self, data_path, model_path, max_words=30, partition=\"train\"):\n        self.ann = json.load(open(data_path))\n        if partition == \"train\":\n            self.ann = self.ann\n        else:\n            self.ann = self.ann[:200]\n\n        self.max_words = max_words\n        tokenizer = Tokenizer(model_path=model_path + \"./tokenizer.model\")\n        self.tokenizer1 = tokenizer\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n\n        ann = self.ann[index]\n        if ann.get(\"input\", \"\") == \"\":\n            prompt = PROMPT_DICT[\"prompt_no_input\"].format_map(ann)\n        else:\n            prompt = PROMPT_DICT[\"prompt_input\"].format_map(ann)\n        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)\n        labels[: len(prompt)] = -1\n        example_mask = example.ge(0)\n        label_mask = labels.ge(0)\n        example[~example_mask] = 0\n        labels[~label_mask] = 0\n        example_mask = example_mask.float()\n        label_mask = label_mask.float()\n\n        return example, labels, example_mask\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser(\"MAE pre-training\", add_help=False)\n    parser.add_argument(\n        \"--batch_size\",\n        default=64,\n        type=int,\n        help=\"Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus\",\n    )\n    parser.add_argument(\"--epochs\", default=400, type=int)\n    parser.add_argument(\n        \"--accum_iter\",\n        default=1,\n        type=int,\n        help=\"Accumulate gradient iterations (for increasing the effective batch size under memory constraints)\",\n    )\n\n    # Model parameters\n    parser.add_argument(\"--llama_model_path\", default=\"./llama\", type=str, help=\"path of llama model\")","source_hash":"a0ec94f11bf4623533eb76da6c0ad77e1e0697d4cd35193d37bc7f4e445185d1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.finetuning.get_args_parser","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.finetuning.get_args_parser#L78-L146","kind":"function","name":"get_args_parser","path":"alpaca_finetuning_v1/finetuning.py","language":"python","start_line":78,"end_line":146,"context_start_line":58,"context_end_line":166,"code":"        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)\n        labels[: len(prompt)] = -1\n        example_mask = example.ge(0)\n        label_mask = labels.ge(0)\n        example[~example_mask] = 0\n        labels[~label_mask] = 0\n        example_mask = example_mask.float()\n        label_mask = label_mask.float()\n\n        return example, labels, example_mask\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser(\"MAE pre-training\", add_help=False)\n    parser.add_argument(\n        \"--batch_size\",\n        default=64,\n        type=int,\n        help=\"Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus\",\n    )\n    parser.add_argument(\"--epochs\", default=400, type=int)\n    parser.add_argument(\n        \"--accum_iter\",\n        default=1,\n        type=int,\n        help=\"Accumulate gradient iterations (for increasing the effective batch size under memory constraints)\",\n    )\n\n    # Model parameters\n    parser.add_argument(\"--llama_model_path\", default=\"./llama\", type=str, help=\"path of llama model\")\n    parser.add_argument(\"--model\", default=\"llama7B_adapter\", type=str, metavar=\"MODEL\", help=\"Name of model to train\")\n\n    parser.add_argument(\"--adapter_layer\", type=int, default=30, metavar=\"LENGTH\", help=\"the number of adapter layer\")\n\n    parser.add_argument(\"--adapter_len\", type=int, default=10, metavar=\"LENGTH\", help=\"the adapter length\")\n\n    parser.add_argument(\"--max_seq_len\", type=int, default=512, metavar=\"LENGTH\", help=\"the maximum sequence length\")\n\n    # Optimizer parameters\n    parser.add_argument(\"--weight_decay\", type=float, default=0.05, help=\"weight decay (default: 0.05)\")\n\n    parser.add_argument(\"--lr\", type=float, default=None, metavar=\"LR\", help=\"learning rate (absolute lr)\")\n    parser.add_argument(\n        \"--blr\",\n        type=float,\n        default=1e-3,\n        metavar=\"LR\",\n        help=\"base learning rate: absolute_lr = base_lr * total_batch_size / 256\",\n    )\n    parser.add_argument(\n        \"--min_lr\", type=float, default=0.0, metavar=\"LR\", help=\"lower lr bound for cyclic schedulers that hit 0\"\n    )\n\n    parser.add_argument(\"--warmup_epochs\", type=int, default=40, metavar=\"N\", help=\"epochs to warmup LR\")\n\n    # Dataset parameters\n    parser.add_argument(\"--data_path\", default=\"/instruction_dataset/\", type=str, help=\"dataset path\")\n\n    parser.add_argument(\"--output_dir\", default=\"./output_dir\", help=\"path where to save, empty for no saving\")\n    parser.add_argument(\"--log_dir\", default=\"./output_dir\", help=\"path where to tensorboard log\")\n    parser.add_argument(\"--device\", default=\"cuda\", help=\"device to use for training / testing\")\n    parser.add_argument(\"--seed\", default=0, type=int)\n    parser.add_argument(\"--resume\", default=\"\", help=\"resume from checkpoint\")\n\n    parser.add_argument(\"--start_epoch\", default=0, type=int, metavar=\"N\", help=\"start epoch\")\n    parser.add_argument(\"--num_workers\", default=10, type=int)\n    parser.add_argument(\n        \"--pin_mem\",\n        action=\"store_true\",\n        help=\"Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.\",\n    )\n    parser.add_argument(\"--no_pin_mem\", action=\"store_false\", dest=\"pin_mem\")\n    parser.set_defaults(pin_mem=True)\n\n    # distributed training parameters\n    parser.add_argument(\"--world_size\", default=1, type=int, help=\"number of distributed processes\")\n    parser.add_argument(\"--local_rank\", default=-1, type=int)\n    parser.add_argument(\"--dist_on_itp\", action=\"store_true\")\n    parser.add_argument(\"--dist_url\", default=\"env://\", help=\"url used to set up distributed training\")\n\n    return parser\n\n\ndef main(args):\n\n    misc.init_distributed_mode(args)\n\n    print(\"job dir: {}\".format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(\", \", \",\\n\"))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n\n    cudnn.benchmark = True\n\n    dataset_train = InstructionDataset(\n        data_path=args.data_path, model_path=args.llama_model_path, max_words=args.max_seq_len, partition=\"train\"","source_hash":"a0ec94f11bf4623533eb76da6c0ad77e1e0697d4cd35193d37bc7f4e445185d1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.finetuning.main","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.finetuning.main#L149-L285","kind":"function","name":"main","path":"alpaca_finetuning_v1/finetuning.py","language":"python","start_line":149,"end_line":285,"context_start_line":129,"context_end_line":294,"code":"\n    parser.add_argument(\"--start_epoch\", default=0, type=int, metavar=\"N\", help=\"start epoch\")\n    parser.add_argument(\"--num_workers\", default=10, type=int)\n    parser.add_argument(\n        \"--pin_mem\",\n        action=\"store_true\",\n        help=\"Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.\",\n    )\n    parser.add_argument(\"--no_pin_mem\", action=\"store_false\", dest=\"pin_mem\")\n    parser.set_defaults(pin_mem=True)\n\n    # distributed training parameters\n    parser.add_argument(\"--world_size\", default=1, type=int, help=\"number of distributed processes\")\n    parser.add_argument(\"--local_rank\", default=-1, type=int)\n    parser.add_argument(\"--dist_on_itp\", action=\"store_true\")\n    parser.add_argument(\"--dist_url\", default=\"env://\", help=\"url used to set up distributed training\")\n\n    return parser\n\n\ndef main(args):\n\n    misc.init_distributed_mode(args)\n\n    print(\"job dir: {}\".format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(\", \", \",\\n\"))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n\n    cudnn.benchmark = True\n\n    dataset_train = InstructionDataset(\n        data_path=args.data_path, model_path=args.llama_model_path, max_words=args.max_seq_len, partition=\"train\"\n    )\n    dataset_val = InstructionDataset(\n        data_path=args.data_path, model_path=args.llama_model_path, max_words=args.max_seq_len, partition=\"val\"\n    )\n\n    print(dataset_train)\n    print(dataset_val)\n\n    if True:  # args.distributed:\n        num_tasks = misc.get_world_size()\n        global_rank = misc.get_rank()\n        sampler_train = torch.utils.data.DistributedSampler(\n            dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True\n        )\n\n        sampler_val = torch.utils.data.DistributedSampler(\n            dataset_val, num_replicas=num_tasks, rank=global_rank, shuffle=True\n        )\n\n        print(\"Sampler_train = %s\" % str(sampler_train))\n    else:\n        sampler_train = torch.utils.data.RandomSampler(dataset_train)\n\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train,\n        sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    data_loader_val = torch.utils.data.DataLoader(\n        dataset_val,\n        sampler=sampler_val,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # define the model\n    model = models_llama_adapter.__dict__[args.model](args)\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = optim_factory.param_groups_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n    misc.load_model(args=args, model_without_ddp=model_without_ddp, optimizer=optimizer, loss_scaler=loss_scaler)\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n            data_loader_val.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train, optimizer, device, epoch, loss_scaler, log_writer=log_writer, args=args\n        )\n\n        val_stats = val_one_epoch(\n            model, data_loader_val, optimizer, device, epoch, loss_scaler, log_writer=log_writer, args=args\n        )\n\n        if args.output_dir and (epoch % 8 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args,\n                model=model,\n                model_without_ddp=model_without_ddp,\n                optimizer=optimizer,\n                loss_scaler=loss_scaler,\n                epoch=epoch,\n            )\n\n        log_stats = {\n            **{f\"train_{k}\": v for k, v in train_stats.items()},\n            \"epoch\": epoch,\n            **{f\"val_{k}\": v for k, v in val_stats.items()},\n        }\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print(\"Training time {}\".format(total_time_str))\n\n\nif __name__ == \"__main__\":\n\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"a0ec94f11bf4623533eb76da6c0ad77e1e0697d4cd35193d37bc7f4e445185d1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.finetuning.__init__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.finetuning.__init__#L37-L46","kind":"function","name":"__init__","path":"alpaca_finetuning_v1/finetuning.py","language":"python","start_line":37,"end_line":46,"context_start_line":17,"context_end_line":66,"code":"from torch.utils.tensorboard import SummaryWriter\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\n\nfrom llama import Tokenizer\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\nclass InstructionDataset(Dataset):\n    def __init__(self, data_path, model_path, max_words=30, partition=\"train\"):\n        self.ann = json.load(open(data_path))\n        if partition == \"train\":\n            self.ann = self.ann\n        else:\n            self.ann = self.ann[:200]\n\n        self.max_words = max_words\n        tokenizer = Tokenizer(model_path=model_path + \"./tokenizer.model\")\n        self.tokenizer1 = tokenizer\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n\n        ann = self.ann[index]\n        if ann.get(\"input\", \"\") == \"\":\n            prompt = PROMPT_DICT[\"prompt_no_input\"].format_map(ann)\n        else:\n            prompt = PROMPT_DICT[\"prompt_input\"].format_map(ann)\n        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)","source_hash":"a0ec94f11bf4623533eb76da6c0ad77e1e0697d4cd35193d37bc7f4e445185d1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.finetuning.__len__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.finetuning.__len__#L48-L49","kind":"function","name":"__len__","path":"alpaca_finetuning_v1/finetuning.py","language":"python","start_line":48,"end_line":49,"context_start_line":28,"context_end_line":69,"code":"    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\nclass InstructionDataset(Dataset):\n    def __init__(self, data_path, model_path, max_words=30, partition=\"train\"):\n        self.ann = json.load(open(data_path))\n        if partition == \"train\":\n            self.ann = self.ann\n        else:\n            self.ann = self.ann[:200]\n\n        self.max_words = max_words\n        tokenizer = Tokenizer(model_path=model_path + \"./tokenizer.model\")\n        self.tokenizer1 = tokenizer\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n\n        ann = self.ann[index]\n        if ann.get(\"input\", \"\") == \"\":\n            prompt = PROMPT_DICT[\"prompt_no_input\"].format_map(ann)\n        else:\n            prompt = PROMPT_DICT[\"prompt_input\"].format_map(ann)\n        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)\n        labels[: len(prompt)] = -1\n        example_mask = example.ge(0)\n        label_mask = labels.ge(0)","source_hash":"a0ec94f11bf4623533eb76da6c0ad77e1e0697d4cd35193d37bc7f4e445185d1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.finetuning.__getitem__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.finetuning.__getitem__#L51-L75","kind":"function","name":"__getitem__","path":"alpaca_finetuning_v1/finetuning.py","language":"python","start_line":51,"end_line":75,"context_start_line":31,"context_end_line":95,"code":"        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\nclass InstructionDataset(Dataset):\n    def __init__(self, data_path, model_path, max_words=30, partition=\"train\"):\n        self.ann = json.load(open(data_path))\n        if partition == \"train\":\n            self.ann = self.ann\n        else:\n            self.ann = self.ann[:200]\n\n        self.max_words = max_words\n        tokenizer = Tokenizer(model_path=model_path + \"./tokenizer.model\")\n        self.tokenizer1 = tokenizer\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n\n        ann = self.ann[index]\n        if ann.get(\"input\", \"\") == \"\":\n            prompt = PROMPT_DICT[\"prompt_no_input\"].format_map(ann)\n        else:\n            prompt = PROMPT_DICT[\"prompt_input\"].format_map(ann)\n        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)\n        labels[: len(prompt)] = -1\n        example_mask = example.ge(0)\n        label_mask = labels.ge(0)\n        example[~example_mask] = 0\n        labels[~label_mask] = 0\n        example_mask = example_mask.float()\n        label_mask = label_mask.float()\n\n        return example, labels, example_mask\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser(\"MAE pre-training\", add_help=False)\n    parser.add_argument(\n        \"--batch_size\",\n        default=64,\n        type=int,\n        help=\"Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus\",\n    )\n    parser.add_argument(\"--epochs\", default=400, type=int)\n    parser.add_argument(\n        \"--accum_iter\",\n        default=1,\n        type=int,\n        help=\"Accumulate gradient iterations (for increasing the effective batch size under memory constraints)\",\n    )\n\n    # Model parameters\n    parser.add_argument(\"--llama_model_path\", default=\"./llama\", type=str, help=\"path of llama model\")","source_hash":"a0ec94f11bf4623533eb76da6c0ad77e1e0697d4cd35193d37bc7f4e445185d1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.util.misc#L1-L340","kind":"module","name":"alpaca_finetuning_v1.util.misc","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":1,"end_line":340,"context_start_line":1,"context_end_line":340,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\n\nimport torch\nimport torch.distributed as dist\nfrom torch import inf\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.resume:\n        if args.resume.startswith('https'):\n            checkpoint = torch.hub.load_state_dict_from_url(\n                args.resume, map_location='cpu', check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location='cpu')\n        model_without_ddp.load_state_dict(checkpoint['model'])\n        print(\"Resume checkpoint %s\" % args.resume)\n        if 'optimizer' in checkpoint and 'epoch' in checkpoint and not (hasattr(args, 'eval') and args.eval):\n            optimizer.load_state_dict(checkpoint['optimizer'])\n            args.start_epoch = checkpoint['epoch'] + 1\n            if 'scaler' in checkpoint:\n                loss_scaler.load_state_dict(checkpoint['scaler'])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.SmoothedValue","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.util.misc.SmoothedValue#L24-L83","kind":"class","name":"SmoothedValue","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":24,"end_line":83,"context_start_line":4,"context_end_line":103,"code":"# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\n\nimport torch\nimport torch.distributed as dist\nfrom torch import inf\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.MetricLogger","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.util.misc.MetricLogger#L86-L167","kind":"class","name":"MetricLogger","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":86,"end_line":167,"context_start_line":66,"context_end_line":187,"code":"    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.setup_for_distributed","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.setup_for_distributed#L170-L184","kind":"function","name":"setup_for_distributed","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":170,"end_line":184,"context_start_line":150,"context_end_line":204,"code":"                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.is_dist_avail_and_initialized","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.is_dist_avail_and_initialized#L187-L192","kind":"function","name":"is_dist_avail_and_initialized","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":187,"end_line":192,"context_start_line":167,"context_end_line":212,"code":"            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.get_world_size","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.get_world_size#L195-L198","kind":"function","name":"get_world_size","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":195,"end_line":198,"context_start_line":175,"context_end_line":218,"code":"\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.get_rank","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.get_rank#L201-L204","kind":"function","name":"get_rank","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":201,"end_line":204,"context_start_line":181,"context_end_line":224,"code":"            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.is_main_process","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.is_main_process#L207-L208","kind":"function","name":"is_main_process","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":207,"end_line":208,"context_start_line":187,"context_end_line":228,"code":"def is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.save_on_master","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.save_on_master#L211-L213","kind":"function","name":"save_on_master","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":211,"end_line":213,"context_start_line":191,"context_end_line":233,"code":"        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.init_distributed_mode","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.init_distributed_mode#L216-L248","kind":"function","name":"init_distributed_mode","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":216,"end_line":248,"context_start_line":196,"context_end_line":268,"code":"    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.NativeScalerWithGradNormCount","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.util.misc.NativeScalerWithGradNormCount#L251-L277","kind":"class","name":"NativeScalerWithGradNormCount","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":251,"end_line":277,"context_start_line":231,"context_end_line":297,"code":"        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.get_grad_norm_","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.get_grad_norm_#L280-L292","kind":"function","name":"get_grad_norm_","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":280,"end_line":292,"context_start_line":260,"context_end_line":312,"code":"            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.save_model","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.save_model#L295-L312","kind":"function","name":"save_model","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":295,"end_line":312,"context_start_line":275,"context_end_line":332,"code":"\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.resume:\n        if args.resume.startswith('https'):\n            checkpoint = torch.hub.load_state_dict_from_url(\n                args.resume, map_location='cpu', check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location='cpu')\n        model_without_ddp.load_state_dict(checkpoint['model'])\n        print(\"Resume checkpoint %s\" % args.resume)\n        if 'optimizer' in checkpoint and 'epoch' in checkpoint and not (hasattr(args, 'eval') and args.eval):\n            optimizer.load_state_dict(checkpoint['optimizer'])\n            args.start_epoch = checkpoint['epoch'] + 1\n            if 'scaler' in checkpoint:\n                loss_scaler.load_state_dict(checkpoint['scaler'])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.load_model","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.load_model#L315-L329","kind":"function","name":"load_model","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":315,"end_line":329,"context_start_line":295,"context_end_line":340,"code":"def save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.resume:\n        if args.resume.startswith('https'):\n            checkpoint = torch.hub.load_state_dict_from_url(\n                args.resume, map_location='cpu', check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location='cpu')\n        model_without_ddp.load_state_dict(checkpoint['model'])\n        print(\"Resume checkpoint %s\" % args.resume)\n        if 'optimizer' in checkpoint and 'epoch' in checkpoint and not (hasattr(args, 'eval') and args.eval):\n            optimizer.load_state_dict(checkpoint['optimizer'])\n            args.start_epoch = checkpoint['epoch'] + 1\n            if 'scaler' in checkpoint:\n                loss_scaler.load_state_dict(checkpoint['scaler'])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.all_reduce_mean","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.all_reduce_mean#L332-L340","kind":"function","name":"all_reduce_mean","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":332,"end_line":340,"context_start_line":312,"context_end_line":340,"code":"        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.resume:\n        if args.resume.startswith('https'):\n            checkpoint = torch.hub.load_state_dict_from_url(\n                args.resume, map_location='cpu', check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location='cpu')\n        model_without_ddp.load_state_dict(checkpoint['model'])\n        print(\"Resume checkpoint %s\" % args.resume)\n        if 'optimizer' in checkpoint and 'epoch' in checkpoint and not (hasattr(args, 'eval') and args.eval):\n            optimizer.load_state_dict(checkpoint['optimizer'])\n            args.start_epoch = checkpoint['epoch'] + 1\n            if 'scaler' in checkpoint:\n                loss_scaler.load_state_dict(checkpoint['scaler'])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.__init__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.__init__#L254-L255","kind":"function","name":"__init__","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":254,"end_line":255,"context_start_line":234,"context_end_line":275,"code":"        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.update","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.update#L91-L98","kind":"function","name":"update","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":91,"end_line":98,"context_start_line":71,"context_end_line":118,"code":"        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.synchronize_between_processes","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.synchronize_between_processes#L116-L118","kind":"function","name":"synchronize_between_processes","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":116,"end_line":118,"context_start_line":96,"context_end_line":138,"code":"                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.median","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.median#L56-L58","kind":"function","name":"median","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":56,"end_line":58,"context_start_line":36,"context_end_line":78,"code":"\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.avg","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.avg#L61-L63","kind":"function","name":"avg","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":61,"end_line":63,"context_start_line":41,"context_end_line":83,"code":"\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.global_avg","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.global_avg#L66-L67","kind":"function","name":"global_avg","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":66,"end_line":67,"context_start_line":46,"context_end_line":87,"code":"        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.max","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.max#L70-L71","kind":"function","name":"max","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":70,"end_line":71,"context_start_line":50,"context_end_line":91,"code":"        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.value","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.value#L74-L75","kind":"function","name":"value","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":74,"end_line":75,"context_start_line":54,"context_end_line":95,"code":"\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.__str__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.__str__#L108-L114","kind":"function","name":"__str__","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":108,"end_line":114,"context_start_line":88,"context_end_line":134,"code":"        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.__getattr__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.__getattr__#L100-L106","kind":"function","name":"__getattr__","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":100,"end_line":106,"context_start_line":80,"context_end_line":126,"code":"            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.add_meter","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.add_meter#L120-L121","kind":"function","name":"add_meter","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":120,"end_line":121,"context_start_line":100,"context_end_line":141,"code":"    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.log_every","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.log_every#L123-L167","kind":"function","name":"log_every","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":123,"end_line":167,"context_start_line":103,"context_end_line":187,"code":"        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.print","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.print#L176-L182","kind":"function","name":"print","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":176,"end_line":182,"context_start_line":156,"context_end_line":202,"code":"                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.__call__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.__call__#L257-L271","kind":"function","name":"__call__","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":257,"end_line":271,"context_start_line":237,"context_end_line":291,"code":"        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.state_dict","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.state_dict#L273-L274","kind":"function","name":"state_dict","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":273,"end_line":274,"context_start_line":253,"context_end_line":294,"code":"\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.misc.load_state_dict","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.misc.load_state_dict#L276-L277","kind":"function","name":"load_state_dict","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":276,"end_line":277,"context_start_line":256,"context_end_line":297,"code":"\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.datasets","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.util.datasets#L1-L64","kind":"module","name":"alpaca_finetuning_v1.util.datasets","path":"alpaca_finetuning_v1/util/datasets.py","language":"python","start_line":1,"end_line":64,"context_start_line":1,"context_end_line":64,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\n\nimport os\n\nimport PIL\nfrom timm.data import create_transform\nfrom timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD\nfrom torchvision import datasets, transforms\n\n\ndef build_dataset(is_train, args):\n    transform = build_transform(is_train, args)\n\n    root = os.path.join(args.data_path, \"train\" if is_train else \"val\")\n    dataset = datasets.ImageFolder(root, transform=transform)\n\n    print(dataset)\n\n    return dataset\n\n\ndef build_transform(is_train, args):\n    mean = IMAGENET_DEFAULT_MEAN\n    std = IMAGENET_DEFAULT_STD\n    # train transform\n    if is_train:\n        # this should always dispatch to transforms_imagenet_train\n        transform = create_transform(\n            input_size=args.input_size,\n            is_training=True,\n            color_jitter=args.color_jitter,\n            auto_augment=args.aa,\n            interpolation=\"bicubic\",\n            re_prob=args.reprob,\n            re_mode=args.remode,\n            re_count=args.recount,\n            mean=mean,\n            std=std,\n        )\n        return transform\n\n    # eval transform\n    t = []\n    if args.input_size <= 224:\n        crop_pct = 224 / 256\n    else:\n        crop_pct = 1.0\n    size = int(args.input_size / crop_pct)\n    t.append(\n        transforms.Resize(size, interpolation=PIL.Image.BICUBIC),  # to maintain same ratio w.r.t. 224 images\n    )\n    t.append(transforms.CenterCrop(args.input_size))\n\n    t.append(transforms.ToTensor())\n    t.append(transforms.Normalize(mean, std))\n    return transforms.Compose(t)","source_hash":"b40d68d3bfae8ebdf0b0c35884f609d3d8a46dc73c5e6397d8f537a389ee9b1d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.datasets.build_dataset","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.datasets.build_dataset#L19-L27","kind":"function","name":"build_dataset","path":"alpaca_finetuning_v1/util/datasets.py","language":"python","start_line":19,"end_line":27,"context_start_line":1,"context_end_line":47,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\n\nimport os\n\nimport PIL\nfrom timm.data import create_transform\nfrom timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD\nfrom torchvision import datasets, transforms\n\n\ndef build_dataset(is_train, args):\n    transform = build_transform(is_train, args)\n\n    root = os.path.join(args.data_path, \"train\" if is_train else \"val\")\n    dataset = datasets.ImageFolder(root, transform=transform)\n\n    print(dataset)\n\n    return dataset\n\n\ndef build_transform(is_train, args):\n    mean = IMAGENET_DEFAULT_MEAN\n    std = IMAGENET_DEFAULT_STD\n    # train transform\n    if is_train:\n        # this should always dispatch to transforms_imagenet_train\n        transform = create_transform(\n            input_size=args.input_size,\n            is_training=True,\n            color_jitter=args.color_jitter,\n            auto_augment=args.aa,\n            interpolation=\"bicubic\",\n            re_prob=args.reprob,\n            re_mode=args.remode,\n            re_count=args.recount,\n            mean=mean,\n            std=std,\n        )","source_hash":"b40d68d3bfae8ebdf0b0c35884f609d3d8a46dc73c5e6397d8f537a389ee9b1d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.datasets.build_transform","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.datasets.build_transform#L30-L64","kind":"function","name":"build_transform","path":"alpaca_finetuning_v1/util/datasets.py","language":"python","start_line":30,"end_line":64,"context_start_line":10,"context_end_line":64,"code":"\nimport os\n\nimport PIL\nfrom timm.data import create_transform\nfrom timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD\nfrom torchvision import datasets, transforms\n\n\ndef build_dataset(is_train, args):\n    transform = build_transform(is_train, args)\n\n    root = os.path.join(args.data_path, \"train\" if is_train else \"val\")\n    dataset = datasets.ImageFolder(root, transform=transform)\n\n    print(dataset)\n\n    return dataset\n\n\ndef build_transform(is_train, args):\n    mean = IMAGENET_DEFAULT_MEAN\n    std = IMAGENET_DEFAULT_STD\n    # train transform\n    if is_train:\n        # this should always dispatch to transforms_imagenet_train\n        transform = create_transform(\n            input_size=args.input_size,\n            is_training=True,\n            color_jitter=args.color_jitter,\n            auto_augment=args.aa,\n            interpolation=\"bicubic\",\n            re_prob=args.reprob,\n            re_mode=args.remode,\n            re_count=args.recount,\n            mean=mean,\n            std=std,\n        )\n        return transform\n\n    # eval transform\n    t = []\n    if args.input_size <= 224:\n        crop_pct = 224 / 256\n    else:\n        crop_pct = 1.0\n    size = int(args.input_size / crop_pct)\n    t.append(\n        transforms.Resize(size, interpolation=PIL.Image.BICUBIC),  # to maintain same ratio w.r.t. 224 images\n    )\n    t.append(transforms.CenterCrop(args.input_size))\n\n    t.append(transforms.ToTensor())\n    t.append(transforms.Normalize(mean, std))\n    return transforms.Compose(t)","source_hash":"b40d68d3bfae8ebdf0b0c35884f609d3d8a46dc73c5e6397d8f537a389ee9b1d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.lr_decay","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.util.lr_decay#L1-L76","kind":"module","name":"alpaca_finetuning_v1.util.lr_decay","path":"alpaca_finetuning_v1/util/lr_decay.py","language":"python","start_line":1,"end_line":76,"context_start_line":1,"context_end_line":76,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# ELECTRA https://github.com/google-research/electra\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport json\n\n\ndef param_groups_lrd(model, weight_decay=0.05, no_weight_decay_list=[], layer_decay=.75):\n    \"\"\"\n    Parameter groups for layer-wise lr decay\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L58\n    \"\"\"\n    param_group_names = {}\n    param_groups = {}\n\n    num_layers = len(model.blocks) + 1\n\n    layer_scales = list(layer_decay ** (num_layers - i) for i in range(num_layers + 1))\n\n    for n, p in model.named_parameters():\n        if not p.requires_grad:\n            continue\n\n        # no decay: all 1D parameters and model specific ones\n        if p.ndim == 1 or n in no_weight_decay_list:\n            g_decay = \"no_decay\"\n            this_decay = 0.\n        else:\n            g_decay = \"decay\"\n            this_decay = weight_decay\n            \n        layer_id = get_layer_id_for_vit(n, num_layers)\n        group_name = \"layer_%d_%s\" % (layer_id, g_decay)\n\n        if group_name not in param_group_names:\n            this_scale = layer_scales[layer_id]\n\n            param_group_names[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n            param_groups[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n\n        param_group_names[group_name][\"params\"].append(n)\n        param_groups[group_name][\"params\"].append(p)\n\n    # print(\"parameter groups: \\n%s\" % json.dumps(param_group_names, indent=2))\n\n    return list(param_groups.values())\n\n\ndef get_layer_id_for_vit(name, num_layers):\n    \"\"\"\n    Assign a parameter with its layer id\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L33\n    \"\"\"\n    if name in ['cls_token', 'pos_embed']:\n        return 0\n    elif name.startswith('patch_embed'):\n        return 0\n    elif name.startswith('blocks'):\n        return int(name.split('.')[1]) + 1\n    else:\n        return num_layers","source_hash":"c21f7ac4070e7dc3ef9fe3a904b9ab59b80082d136b70d99cf7853c5ceb62dba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.lr_decay.param_groups_lrd","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.lr_decay.param_groups_lrd#L15-L61","kind":"function","name":"param_groups_lrd","path":"alpaca_finetuning_v1/util/lr_decay.py","language":"python","start_line":15,"end_line":61,"context_start_line":1,"context_end_line":76,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# ELECTRA https://github.com/google-research/electra\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport json\n\n\ndef param_groups_lrd(model, weight_decay=0.05, no_weight_decay_list=[], layer_decay=.75):\n    \"\"\"\n    Parameter groups for layer-wise lr decay\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L58\n    \"\"\"\n    param_group_names = {}\n    param_groups = {}\n\n    num_layers = len(model.blocks) + 1\n\n    layer_scales = list(layer_decay ** (num_layers - i) for i in range(num_layers + 1))\n\n    for n, p in model.named_parameters():\n        if not p.requires_grad:\n            continue\n\n        # no decay: all 1D parameters and model specific ones\n        if p.ndim == 1 or n in no_weight_decay_list:\n            g_decay = \"no_decay\"\n            this_decay = 0.\n        else:\n            g_decay = \"decay\"\n            this_decay = weight_decay\n            \n        layer_id = get_layer_id_for_vit(n, num_layers)\n        group_name = \"layer_%d_%s\" % (layer_id, g_decay)\n\n        if group_name not in param_group_names:\n            this_scale = layer_scales[layer_id]\n\n            param_group_names[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n            param_groups[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n\n        param_group_names[group_name][\"params\"].append(n)\n        param_groups[group_name][\"params\"].append(p)\n\n    # print(\"parameter groups: \\n%s\" % json.dumps(param_group_names, indent=2))\n\n    return list(param_groups.values())\n\n\ndef get_layer_id_for_vit(name, num_layers):\n    \"\"\"\n    Assign a parameter with its layer id\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L33\n    \"\"\"\n    if name in ['cls_token', 'pos_embed']:\n        return 0\n    elif name.startswith('patch_embed'):\n        return 0\n    elif name.startswith('blocks'):\n        return int(name.split('.')[1]) + 1\n    else:\n        return num_layers","source_hash":"c21f7ac4070e7dc3ef9fe3a904b9ab59b80082d136b70d99cf7853c5ceb62dba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.lr_decay.get_layer_id_for_vit","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.lr_decay.get_layer_id_for_vit#L64-L76","kind":"function","name":"get_layer_id_for_vit","path":"alpaca_finetuning_v1/util/lr_decay.py","language":"python","start_line":64,"end_line":76,"context_start_line":44,"context_end_line":76,"code":"\n            param_group_names[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n            param_groups[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n\n        param_group_names[group_name][\"params\"].append(n)\n        param_groups[group_name][\"params\"].append(p)\n\n    # print(\"parameter groups: \\n%s\" % json.dumps(param_group_names, indent=2))\n\n    return list(param_groups.values())\n\n\ndef get_layer_id_for_vit(name, num_layers):\n    \"\"\"\n    Assign a parameter with its layer id\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L33\n    \"\"\"\n    if name in ['cls_token', 'pos_embed']:\n        return 0\n    elif name.startswith('patch_embed'):\n        return 0\n    elif name.startswith('blocks'):\n        return int(name.split('.')[1]) + 1\n    else:\n        return num_layers","source_hash":"c21f7ac4070e7dc3ef9fe3a904b9ab59b80082d136b70d99cf7853c5ceb62dba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.lr_sched","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.util.lr_sched#L1-L23","kind":"module","name":"alpaca_finetuning_v1.util.lr_sched","path":"alpaca_finetuning_v1/util/lr_sched.py","language":"python","start_line":1,"end_line":23,"context_start_line":1,"context_end_line":23,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs\n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * (\n            1.0 + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs))\n        )\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"136a8dba00b53af7934e7377cc6a0dce1e84fef604e55673ce8460b109aa6d0d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.lr_sched.adjust_learning_rate","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.lr_sched.adjust_learning_rate#L10-L23","kind":"function","name":"adjust_learning_rate","path":"alpaca_finetuning_v1/util/lr_sched.py","language":"python","start_line":10,"end_line":23,"context_start_line":1,"context_end_line":23,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs\n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * (\n            1.0 + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs))\n        )\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"136a8dba00b53af7934e7377cc6a0dce1e84fef604e55673ce8460b109aa6d0d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.pos_embed","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.util.pos_embed#L1-L97","kind":"module","name":"alpaca_finetuning_v1.util.pos_embed","path":"alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":1,"end_line":97,"context_start_line":1,"context_end_line":97,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# Position embedding utils\n# --------------------------------------------------------\n\nimport numpy as np\nimport torch\n\n\n# --------------------------------------------------------\n# 2D sine-cosine position embedding\n# References:\n# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py\n# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"\n    grid_size: int of the grid height and width\n    return:\n    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)\n    \"\"\"\n    grid_h = np.arange(grid_size, dtype=np.float32)\n    grid_w = np.arange(grid_size, dtype=np.float32)\n    grid = np.meshgrid(grid_w, grid_h)  # here w goes first\n    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.0\n    omega = 1.0 / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum(\"m,d->md\", pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out)  # (M, D/2)\n    emb_cos = np.cos(out)  # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)\n    return emb\n\n\n# --------------------------------------------------------\n# Interpolate position embeddings for high-resolution\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\ndef interpolate_pos_embed(model, checkpoint_model):\n    if \"pos_embed\" in checkpoint_model:\n        pos_embed_checkpoint = checkpoint_model[\"pos_embed\"]\n        embedding_size = pos_embed_checkpoint.shape[-1]\n        num_patches = model.patch_embed.num_patches\n        num_extra_tokens = model.pos_embed.shape[-2] - num_patches\n        # height (== width) for the checkpoint position embedding\n        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)\n        # height (== width) for the new position embedding\n        new_size = int(num_patches**0.5)\n        # class_token and dist_token are kept unchanged\n        if orig_size != new_size:\n            print(\"Position interpolate from %dx%d to %dx%d\" % (orig_size, orig_size, new_size, new_size))\n            extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]\n            # only the position tokens are interpolated\n            pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:]\n            pos_tokens = pos_tokens.reshape(-1, orig_size, orig_size, embedding_size).permute(0, 3, 1, 2)\n            pos_tokens = torch.nn.functional.interpolate(\n                pos_tokens, size=(new_size, new_size), mode=\"bicubic\", align_corners=False\n            )\n            pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)\n            new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)\n            checkpoint_model[\"pos_embed\"] = new_pos_embed","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.pos_embed.get_2d_sincos_pos_embed","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.pos_embed.get_2d_sincos_pos_embed#L20-L35","kind":"function","name":"get_2d_sincos_pos_embed","path":"alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":20,"end_line":35,"context_start_line":1,"context_end_line":55,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# Position embedding utils\n# --------------------------------------------------------\n\nimport numpy as np\nimport torch\n\n\n# --------------------------------------------------------\n# 2D sine-cosine position embedding\n# References:\n# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py\n# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"\n    grid_size: int of the grid height and width\n    return:\n    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)\n    \"\"\"\n    grid_h = np.arange(grid_size, dtype=np.float32)\n    grid_w = np.arange(grid_size, dtype=np.float32)\n    grid = np.meshgrid(grid_w, grid_h)  # here w goes first\n    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.pos_embed.get_2d_sincos_pos_embed_from_grid","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.pos_embed.get_2d_sincos_pos_embed_from_grid#L38-L46","kind":"function","name":"get_2d_sincos_pos_embed_from_grid","path":"alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":38,"end_line":46,"context_start_line":18,"context_end_line":66,"code":"# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"\n    grid_size: int of the grid height and width\n    return:\n    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)\n    \"\"\"\n    grid_h = np.arange(grid_size, dtype=np.float32)\n    grid_w = np.arange(grid_size, dtype=np.float32)\n    grid = np.meshgrid(grid_w, grid_h)  # here w goes first\n    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.0\n    omega = 1.0 / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum(\"m,d->md\", pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out)  # (M, D/2)\n    emb_cos = np.cos(out)  # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.pos_embed.get_1d_sincos_pos_embed_from_grid","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.pos_embed.get_1d_sincos_pos_embed_from_grid#L49-L67","kind":"function","name":"get_1d_sincos_pos_embed_from_grid","path":"alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":49,"end_line":67,"context_start_line":29,"context_end_line":87,"code":"    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.0\n    omega = 1.0 / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum(\"m,d->md\", pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out)  # (M, D/2)\n    emb_cos = np.cos(out)  # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)\n    return emb\n\n\n# --------------------------------------------------------\n# Interpolate position embeddings for high-resolution\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\ndef interpolate_pos_embed(model, checkpoint_model):\n    if \"pos_embed\" in checkpoint_model:\n        pos_embed_checkpoint = checkpoint_model[\"pos_embed\"]\n        embedding_size = pos_embed_checkpoint.shape[-1]\n        num_patches = model.patch_embed.num_patches\n        num_extra_tokens = model.pos_embed.shape[-2] - num_patches\n        # height (== width) for the checkpoint position embedding\n        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)\n        # height (== width) for the new position embedding\n        new_size = int(num_patches**0.5)\n        # class_token and dist_token are kept unchanged\n        if orig_size != new_size:\n            print(\"Position interpolate from %dx%d to %dx%d\" % (orig_size, orig_size, new_size, new_size))","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.pos_embed.interpolate_pos_embed","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.pos_embed.interpolate_pos_embed#L75-L97","kind":"function","name":"interpolate_pos_embed","path":"alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":75,"end_line":97,"context_start_line":55,"context_end_line":97,"code":"    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.0\n    omega = 1.0 / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum(\"m,d->md\", pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out)  # (M, D/2)\n    emb_cos = np.cos(out)  # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)\n    return emb\n\n\n# --------------------------------------------------------\n# Interpolate position embeddings for high-resolution\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\ndef interpolate_pos_embed(model, checkpoint_model):\n    if \"pos_embed\" in checkpoint_model:\n        pos_embed_checkpoint = checkpoint_model[\"pos_embed\"]\n        embedding_size = pos_embed_checkpoint.shape[-1]\n        num_patches = model.patch_embed.num_patches\n        num_extra_tokens = model.pos_embed.shape[-2] - num_patches\n        # height (== width) for the checkpoint position embedding\n        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)\n        # height (== width) for the new position embedding\n        new_size = int(num_patches**0.5)\n        # class_token and dist_token are kept unchanged\n        if orig_size != new_size:\n            print(\"Position interpolate from %dx%d to %dx%d\" % (orig_size, orig_size, new_size, new_size))\n            extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]\n            # only the position tokens are interpolated\n            pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:]\n            pos_tokens = pos_tokens.reshape(-1, orig_size, orig_size, embedding_size).permute(0, 3, 1, 2)\n            pos_tokens = torch.nn.functional.interpolate(\n                pos_tokens, size=(new_size, new_size), mode=\"bicubic\", align_corners=False\n            )\n            pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)\n            new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)\n            checkpoint_model[\"pos_embed\"] = new_pos_embed","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.lars","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.util.lars#L1-L47","kind":"module","name":"alpaca_finetuning_v1.util.lars","path":"alpaca_finetuning_v1/util/lars.py","language":"python","start_line":1,"end_line":47,"context_start_line":1,"context_end_line":47,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# LARS optimizer, implementation from MoCo v3:\n# https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\n\nimport torch\n\n\nclass LARS(torch.optim.Optimizer):\n    \"\"\"\n    LARS optimizer, no rate scaling or weight decay for parameters <= 1D.\n    \"\"\"\n    def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, trust_coefficient=0.001):\n        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, trust_coefficient=trust_coefficient)\n        super().__init__(params, defaults)\n\n    @torch.no_grad()\n    def step(self):\n        for g in self.param_groups:\n            for p in g['params']:\n                dp = p.grad\n\n                if dp is None:\n                    continue\n\n                if p.ndim > 1: # if not normalization gamma/beta or bias\n                    dp = dp.add(p, alpha=g['weight_decay'])\n                    param_norm = torch.norm(p)\n                    update_norm = torch.norm(dp)\n                    one = torch.ones_like(param_norm)\n                    q = torch.where(param_norm > 0.,\n                                    torch.where(update_norm > 0,\n                                    (g['trust_coefficient'] * param_norm / update_norm), one),\n                                    one)\n                    dp = dp.mul(q)\n\n                param_state = self.state[p]\n                if 'mu' not in param_state:\n                    param_state['mu'] = torch.zeros_like(p)\n                mu = param_state['mu']\n                mu.mul_(g['momentum']).add_(dp)\n                p.add_(mu, alpha=-g['lr'])","source_hash":"1bb75085eecf93b8d7ed7972c507495586e71a8f483bcdcd4f37c354f88a0fc0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.lars.LARS","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.util.lars.LARS#L14-L47","kind":"class","name":"LARS","path":"alpaca_finetuning_v1/util/lars.py","language":"python","start_line":14,"end_line":47,"context_start_line":1,"context_end_line":47,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# LARS optimizer, implementation from MoCo v3:\n# https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\n\nimport torch\n\n\nclass LARS(torch.optim.Optimizer):\n    \"\"\"\n    LARS optimizer, no rate scaling or weight decay for parameters <= 1D.\n    \"\"\"\n    def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, trust_coefficient=0.001):\n        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, trust_coefficient=trust_coefficient)\n        super().__init__(params, defaults)\n\n    @torch.no_grad()\n    def step(self):\n        for g in self.param_groups:\n            for p in g['params']:\n                dp = p.grad\n\n                if dp is None:\n                    continue\n\n                if p.ndim > 1: # if not normalization gamma/beta or bias\n                    dp = dp.add(p, alpha=g['weight_decay'])\n                    param_norm = torch.norm(p)\n                    update_norm = torch.norm(dp)\n                    one = torch.ones_like(param_norm)\n                    q = torch.where(param_norm > 0.,\n                                    torch.where(update_norm > 0,\n                                    (g['trust_coefficient'] * param_norm / update_norm), one),\n                                    one)\n                    dp = dp.mul(q)\n\n                param_state = self.state[p]\n                if 'mu' not in param_state:\n                    param_state['mu'] = torch.zeros_like(p)\n                mu = param_state['mu']\n                mu.mul_(g['momentum']).add_(dp)\n                p.add_(mu, alpha=-g['lr'])","source_hash":"1bb75085eecf93b8d7ed7972c507495586e71a8f483bcdcd4f37c354f88a0fc0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.lars.__init__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.lars.__init__#L18-L20","kind":"function","name":"__init__","path":"alpaca_finetuning_v1/util/lars.py","language":"python","start_line":18,"end_line":20,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# LARS optimizer, implementation from MoCo v3:\n# https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\n\nimport torch\n\n\nclass LARS(torch.optim.Optimizer):\n    \"\"\"\n    LARS optimizer, no rate scaling or weight decay for parameters <= 1D.\n    \"\"\"\n    def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, trust_coefficient=0.001):\n        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, trust_coefficient=trust_coefficient)\n        super().__init__(params, defaults)\n\n    @torch.no_grad()\n    def step(self):\n        for g in self.param_groups:\n            for p in g['params']:\n                dp = p.grad\n\n                if dp is None:\n                    continue\n\n                if p.ndim > 1: # if not normalization gamma/beta or bias\n                    dp = dp.add(p, alpha=g['weight_decay'])\n                    param_norm = torch.norm(p)\n                    update_norm = torch.norm(dp)\n                    one = torch.ones_like(param_norm)\n                    q = torch.where(param_norm > 0.,\n                                    torch.where(update_norm > 0,\n                                    (g['trust_coefficient'] * param_norm / update_norm), one),\n                                    one)\n                    dp = dp.mul(q)","source_hash":"1bb75085eecf93b8d7ed7972c507495586e71a8f483bcdcd4f37c354f88a0fc0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.util.lars.step","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.util.lars.step#L23-L47","kind":"function","name":"step","path":"alpaca_finetuning_v1/util/lars.py","language":"python","start_line":23,"end_line":47,"context_start_line":3,"context_end_line":47,"code":"\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# LARS optimizer, implementation from MoCo v3:\n# https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\n\nimport torch\n\n\nclass LARS(torch.optim.Optimizer):\n    \"\"\"\n    LARS optimizer, no rate scaling or weight decay for parameters <= 1D.\n    \"\"\"\n    def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, trust_coefficient=0.001):\n        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, trust_coefficient=trust_coefficient)\n        super().__init__(params, defaults)\n\n    @torch.no_grad()\n    def step(self):\n        for g in self.param_groups:\n            for p in g['params']:\n                dp = p.grad\n\n                if dp is None:\n                    continue\n\n                if p.ndim > 1: # if not normalization gamma/beta or bias\n                    dp = dp.add(p, alpha=g['weight_decay'])\n                    param_norm = torch.norm(p)\n                    update_norm = torch.norm(dp)\n                    one = torch.ones_like(param_norm)\n                    q = torch.where(param_norm > 0.,\n                                    torch.where(update_norm > 0,\n                                    (g['trust_coefficient'] * param_norm / update_norm), one),\n                                    one)\n                    dp = dp.mul(q)\n\n                param_state = self.state[p]\n                if 'mu' not in param_state:\n                    param_state['mu'] = torch.zeros_like(p)\n                mu = param_state['mu']\n                mu.mul_(g['momentum']).add_(dp)\n                p.add_(mu, alpha=-g['lr'])","source_hash":"1bb75085eecf93b8d7ed7972c507495586e71a8f483bcdcd4f37c354f88a0fc0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.generation","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.llama.generation#L1-L75","kind":"module","name":"alpaca_finetuning_v1.llama.generation","path":"alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":1,"end_line":75,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward_only(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.generation.LLaMA","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.llama.generation.LLaMA#L12-L64","kind":"class","name":"LLaMA","path":"alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":12,"end_line":64,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward_only(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.generation.sample_top_p","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.generation.sample_top_p#L67-L75","kind":"function","name":"sample_top_p","path":"alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":67,"end_line":75,"context_start_line":47,"context_end_line":75,"code":"                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.generation.__init__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.generation.__init__#L13-L15","kind":"function","name":"__init__","path":"alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":13,"end_line":15,"context_start_line":1,"context_end_line":35,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.generation.generate","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.generation.generate#L17-L64","kind":"function","name":"generate","path":"alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":17,"end_line":64,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward_only(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.llama.model#L1-L222","kind":"module","name":"alpaca_finetuning_v1.llama.model","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":1,"end_line":222,"context_start_line":1,"context_end_line":222,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport torch\nimport torch.nn.functional as F\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n    adapter_len: int = 10\n    adapter_layer: int = 30\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wk = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wv = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wo = Linear(args.n_heads * self.head_dim, args.dim, bias=False)\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_k = self.wk(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_v = self.wv(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            xk = torch.cat([adapter_k, xk], dim=1)\n            xv = torch.cat([adapter_v, xv], dim=1)\n            extra_mask = torch.zeros(1, 1, seqlen, adapter_len).to(mask)\n            mask = torch.cat([extra_mask, mask], dim=-1)\n        keys = xk\n        values = xv\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        if adapter is not None:\n            scores = torch.cat(\n                [\n                    self.gate.tanh().half() * F.softmax(scores[:, :, :, :adapter_len].float(), dim=-1).type_as(xq),\n                    F.softmax(scores[:, :, :, adapter_len:].float(), dim=-1).type_as(xq),\n                ],\n                dim=-1,\n            )\n        else:\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(dim, hidden_dim, bias=False)\n        self.w2 = Linear(hidden_dim, dim, bias=False)\n        self.w3 = Linear(dim, hidden_dim, bias=False)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(params.vocab_size, params.dim)\n\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n\n        _bsz, seqlen = examples.shape\n\n        with torch.no_grad():\n            h = self.tok_embeddings(examples)\n            freqs_cis = self.freqs_cis.to(h.device)\n            freqs_cis = freqs_cis[:seqlen]\n            mask = None\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n            mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n            start_pos = 0\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter_index = 0\n        adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, 4096).unsqueeze(1)\n        for layer in self.layers[-1 * self.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n            adapter_index = adapter_index + 1\n\n        h = self.norm(h)\n        output = self.output(h)\n        output = output[:, :-1, :].reshape(-1, self.vocab_size)\n        labels = labels[:, 1:].flatten()\n\n        c_loss = self.criterion(output, labels)\n        return c_loss","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.ModelArgs","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.llama.model.ModelArgs#L15-L26","kind":"class","name":"ModelArgs","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":15,"end_line":26,"context_start_line":1,"context_end_line":46,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport torch\nimport torch.nn.functional as F\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n    adapter_len: int = 10\n    adapter_layer: int = 30\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.RMSNorm","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.llama.model.RMSNorm#L29-L40","kind":"class","name":"RMSNorm","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":29,"end_line":40,"context_start_line":9,"context_end_line":60,"code":"import torch.nn.functional as F\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n    adapter_len: int = 10\n    adapter_layer: int = 30\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.precompute_freqs_cis","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.model.precompute_freqs_cis#L43-L48","kind":"function","name":"precompute_freqs_cis","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":43,"end_line":48,"context_start_line":23,"context_end_line":68,"code":"    max_batch_size: int = 32\n    max_seq_len: int = 2048\n    adapter_len: int = 10\n    adapter_layer: int = 30\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.reshape_for_broadcast","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.model.reshape_for_broadcast#L51-L56","kind":"function","name":"reshape_for_broadcast","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":51,"end_line":56,"context_start_line":31,"context_end_line":76,"code":"        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.apply_rotary_emb","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.model.apply_rotary_emb#L59-L69","kind":"function","name":"apply_rotary_emb","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":59,"end_line":69,"context_start_line":39,"context_end_line":89,"code":"        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wk = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wv = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wo = Linear(args.n_heads * self.head_dim, args.dim, bias=False)\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.Attention","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.llama.model.Attention#L72-L129","kind":"class","name":"Attention","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":72,"end_line":129,"context_start_line":52,"context_end_line":149,"code":"    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wk = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wv = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wo = Linear(args.n_heads * self.head_dim, args.dim, bias=False)\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_k = self.wk(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_v = self.wv(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            xk = torch.cat([adapter_k, xk], dim=1)\n            xv = torch.cat([adapter_v, xv], dim=1)\n            extra_mask = torch.zeros(1, 1, seqlen, adapter_len).to(mask)\n            mask = torch.cat([extra_mask, mask], dim=-1)\n        keys = xk\n        values = xv\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        if adapter is not None:\n            scores = torch.cat(\n                [\n                    self.gate.tanh().half() * F.softmax(scores[:, :, :, :adapter_len].float(), dim=-1).type_as(xq),\n                    F.softmax(scores[:, :, :, adapter_len:].float(), dim=-1).type_as(xq),\n                ],\n                dim=-1,\n            )\n        else:\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(dim, hidden_dim, bias=False)\n        self.w2 = Linear(hidden_dim, dim, bias=False)\n        self.w3 = Linear(dim, hidden_dim, bias=False)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.FeedForward","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.llama.model.FeedForward#L132-L148","kind":"class","name":"FeedForward","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":132,"end_line":148,"context_start_line":112,"context_end_line":168,"code":"        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        if adapter is not None:\n            scores = torch.cat(\n                [\n                    self.gate.tanh().half() * F.softmax(scores[:, :, :, :adapter_len].float(), dim=-1).type_as(xq),\n                    F.softmax(scores[:, :, :, adapter_len:].float(), dim=-1).type_as(xq),\n                ],\n                dim=-1,\n            )\n        else:\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(dim, hidden_dim, bias=False)\n        self.w2 = Linear(hidden_dim, dim, bias=False)\n        self.w3 = Linear(dim, hidden_dim, bias=False)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.TransformerBlock","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.llama.model.TransformerBlock#L151-L169","kind":"class","name":"TransformerBlock","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":151,"end_line":169,"context_start_line":131,"context_end_line":189,"code":"\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(dim, hidden_dim, bias=False)\n        self.w2 = Linear(hidden_dim, dim, bias=False)\n        self.w3 = Linear(dim, hidden_dim, bias=False)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(params.vocab_size, params.dim)\n\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.Transformer","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.llama.model.Transformer#L172-L222","kind":"class","name":"Transformer","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":172,"end_line":222,"context_start_line":152,"context_end_line":222,"code":"    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(params.vocab_size, params.dim)\n\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n\n        _bsz, seqlen = examples.shape\n\n        with torch.no_grad():\n            h = self.tok_embeddings(examples)\n            freqs_cis = self.freqs_cis.to(h.device)\n            freqs_cis = freqs_cis[:seqlen]\n            mask = None\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n            mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n            start_pos = 0\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter_index = 0\n        adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, 4096).unsqueeze(1)\n        for layer in self.layers[-1 * self.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n            adapter_index = adapter_index + 1\n\n        h = self.norm(h)\n        output = self.output(h)\n        output = output[:, :-1, :].reshape(-1, self.vocab_size)\n        labels = labels[:, 1:].flatten()\n\n        c_loss = self.criterion(output, labels)\n        return c_loss","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.__init__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.model.__init__#L173-L193","kind":"function","name":"__init__","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":173,"end_line":193,"context_start_line":153,"context_end_line":213,"code":"        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(params.vocab_size, params.dim)\n\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n\n        _bsz, seqlen = examples.shape\n\n        with torch.no_grad():\n            h = self.tok_embeddings(examples)\n            freqs_cis = self.freqs_cis.to(h.device)\n            freqs_cis = freqs_cis[:seqlen]\n            mask = None\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n            mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n            start_pos = 0\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter_index = 0\n        adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, 4096).unsqueeze(1)\n        for layer in self.layers[-1 * self.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model._norm","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.model._norm#L35-L36","kind":"function","name":"_norm","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":35,"end_line":36,"context_start_line":15,"context_end_line":56,"code":"class ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n    adapter_len: int = 10\n    adapter_layer: int = 30\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.model.forward","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.model.forward#L195-L222","kind":"function","name":"forward","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":195,"end_line":222,"context_start_line":175,"context_end_line":222,"code":"        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(params.vocab_size, params.dim)\n\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n\n        _bsz, seqlen = examples.shape\n\n        with torch.no_grad():\n            h = self.tok_embeddings(examples)\n            freqs_cis = self.freqs_cis.to(h.device)\n            freqs_cis = freqs_cis[:seqlen]\n            mask = None\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n            mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n            start_pos = 0\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter_index = 0\n        adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, 4096).unsqueeze(1)\n        for layer in self.layers[-1 * self.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n            adapter_index = adapter_index + 1\n\n        h = self.norm(h)\n        output = self.output(h)\n        output = output[:, :-1, :].reshape(-1, self.vocab_size)\n        labels = labels[:, 1:].flatten()\n\n        c_loss = self.criterion(output, labels)\n        return c_loss","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.tokenizer","uri":"program://LLaMA-Adapter/module/alpaca_finetuning_v1.llama.tokenizer#L1-L38","kind":"module","name":"alpaca_finetuning_v1.llama.tokenizer","path":"alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":1,"end_line":38,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.tokenizer.Tokenizer","uri":"program://LLaMA-Adapter/class/alpaca_finetuning_v1.llama.tokenizer.Tokenizer#L13-L38","kind":"class","name":"Tokenizer","path":"alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":13,"end_line":38,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.tokenizer.__init__","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.tokenizer.__init__#L14-L26","kind":"function","name":"__init__","path":"alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":14,"end_line":26,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.tokenizer.encode","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.tokenizer.encode#L28-L35","kind":"function","name":"encode","path":"alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":28,"end_line":35,"context_start_line":8,"context_end_line":38,"code":"from sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:alpaca_finetuning_v1.llama.tokenizer.decode","uri":"program://LLaMA-Adapter/function/alpaca_finetuning_v1.llama.tokenizer.decode#L37-L38","kind":"function","name":"decode","path":"alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":37,"end_line":38,"context_start_line":17,"context_end_line":38,"code":"        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.generation","uri":"program://LLaMA-Adapter/module/llama.generation#L1-L75","kind":"module","name":"llama.generation","path":"llama/generation.py","language":"python","start_line":1,"end_line":75,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.generation.LLaMA","uri":"program://LLaMA-Adapter/class/llama.generation.LLaMA#L12-L64","kind":"class","name":"LLaMA","path":"llama/generation.py","language":"python","start_line":12,"end_line":64,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.generation.sample_top_p","uri":"program://LLaMA-Adapter/function/llama.generation.sample_top_p#L67-L75","kind":"function","name":"sample_top_p","path":"llama/generation.py","language":"python","start_line":67,"end_line":75,"context_start_line":47,"context_end_line":75,"code":"                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.generation.__init__","uri":"program://LLaMA-Adapter/function/llama.generation.__init__#L13-L15","kind":"function","name":"__init__","path":"llama/generation.py","language":"python","start_line":13,"end_line":15,"context_start_line":1,"context_end_line":35,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.generation.generate","uri":"program://LLaMA-Adapter/function/llama.generation.generate#L17-L64","kind":"function","name":"generate","path":"llama/generation.py","language":"python","start_line":17,"end_line":64,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model","uri":"program://LLaMA-Adapter/module/llama.model#L1-L240","kind":"module","name":"llama.model","path":"llama/model.py","language":"python","start_line":1,"end_line":240,"context_start_line":1,"context_end_line":240,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.nn.functional as F\nfrom fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    adapter_len: int = 10\n    adapter_layer: int = 8\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wv = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wo = RowParallelLinear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=False,\n            input_is_parallel=True,\n            init_method=lambda x: x,\n        )\n\n        self.cache_k = torch.zeros((args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)).cuda()\n        self.cache_v = torch.zeros((args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)).cuda()\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        self.cache_k = self.cache_k.to(xq)\n        self.cache_v = self.cache_v.to(xq)\n\n        self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk\n        self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n        keys = self.cache_k[:bsz, : start_pos + seqlen]\n        values = self.cache_v[:bsz, : start_pos + seqlen]\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_k = self.wk(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_v = self.wv(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_k = adapter_k.transpose(1, 2)\n            adapter_v = adapter_v.transpose(1, 2)\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        if adapter is not None:\n            adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n            adapter_scores = self.gate * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n            output = output + torch.matmul(adapter_scores, adapter_v)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n        self.w2 = RowParallelLinear(hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x)\n        self.w3 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        prompt = self.adapter_query.weight.reshape(\n            self.params.adapter_layer, self.params.adapter_len, self.params.dim\n        ).unsqueeze(1)\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers[: -1 * self.params.adapter_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        layer_index = 0\n        for layer in self.layers[-1 * self.params.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, prompt[layer_index])\n            layer_index = layer_index + 1\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.ModelArgs","uri":"program://LLaMA-Adapter/class/llama.model.ModelArgs#L16-L28","kind":"class","name":"ModelArgs","path":"llama/model.py","language":"python","start_line":16,"end_line":28,"context_start_line":1,"context_end_line":48,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.nn.functional as F\nfrom fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    adapter_len: int = 10\n    adapter_layer: int = 8\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.RMSNorm","uri":"program://LLaMA-Adapter/class/llama.model.RMSNorm#L31-L42","kind":"class","name":"RMSNorm","path":"llama/model.py","language":"python","start_line":31,"end_line":42,"context_start_line":11,"context_end_line":62,"code":"from fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    adapter_len: int = 10\n    adapter_layer: int = 8\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.precompute_freqs_cis","uri":"program://LLaMA-Adapter/function/llama.model.precompute_freqs_cis#L45-L50","kind":"function","name":"precompute_freqs_cis","path":"llama/model.py","language":"python","start_line":45,"end_line":50,"context_start_line":25,"context_end_line":70,"code":"    max_seq_len: int = 2048\n\n    adapter_len: int = 10\n    adapter_layer: int = 8\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.reshape_for_broadcast","uri":"program://LLaMA-Adapter/function/llama.model.reshape_for_broadcast#L53-L58","kind":"function","name":"reshape_for_broadcast","path":"llama/model.py","language":"python","start_line":53,"end_line":58,"context_start_line":33,"context_end_line":78,"code":"        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.apply_rotary_emb","uri":"program://LLaMA-Adapter/function/llama.model.apply_rotary_emb#L61-L71","kind":"function","name":"apply_rotary_emb","path":"llama/model.py","language":"python","start_line":61,"end_line":71,"context_start_line":41,"context_end_line":91,"code":"        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.Attention","uri":"program://LLaMA-Adapter/class/llama.model.Attention#L74-L155","kind":"class","name":"Attention","path":"llama/model.py","language":"python","start_line":74,"end_line":155,"context_start_line":54,"context_end_line":175,"code":"    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wv = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wo = RowParallelLinear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=False,\n            input_is_parallel=True,\n            init_method=lambda x: x,\n        )\n\n        self.cache_k = torch.zeros((args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)).cuda()\n        self.cache_v = torch.zeros((args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)).cuda()\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        self.cache_k = self.cache_k.to(xq)\n        self.cache_v = self.cache_v.to(xq)\n\n        self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk\n        self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n        keys = self.cache_k[:bsz, : start_pos + seqlen]\n        values = self.cache_v[:bsz, : start_pos + seqlen]\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_k = self.wk(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_v = self.wv(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_k = adapter_k.transpose(1, 2)\n            adapter_v = adapter_v.transpose(1, 2)\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        if adapter is not None:\n            adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n            adapter_scores = self.gate * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n            output = output + torch.matmul(adapter_scores, adapter_v)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n        self.w2 = RowParallelLinear(hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x)\n        self.w3 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.FeedForward","uri":"program://LLaMA-Adapter/class/llama.model.FeedForward#L158-L174","kind":"class","name":"FeedForward","path":"llama/model.py","language":"python","start_line":158,"end_line":174,"context_start_line":138,"context_end_line":194,"code":"            adapter_v = self.wv(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_k = adapter_k.transpose(1, 2)\n            adapter_v = adapter_v.transpose(1, 2)\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        if adapter is not None:\n            adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n            adapter_scores = self.gate * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n            output = output + torch.matmul(adapter_scores, adapter_v)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n        self.w2 = RowParallelLinear(hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x)\n        self.w3 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.TransformerBlock","uri":"program://LLaMA-Adapter/class/llama.model.TransformerBlock#L177-L194","kind":"class","name":"TransformerBlock","path":"llama/model.py","language":"python","start_line":177,"end_line":194,"context_start_line":157,"context_end_line":214,"code":"\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n        self.w2 = RowParallelLinear(hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x)\n        self.w3 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.Transformer","uri":"program://LLaMA-Adapter/class/llama.model.Transformer#L197-L240","kind":"class","name":"Transformer","path":"llama/model.py","language":"python","start_line":197,"end_line":240,"context_start_line":177,"context_end_line":240,"code":"class TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        prompt = self.adapter_query.weight.reshape(\n            self.params.adapter_layer, self.params.adapter_len, self.params.dim\n        ).unsqueeze(1)\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers[: -1 * self.params.adapter_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        layer_index = 0\n        for layer in self.layers[-1 * self.params.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, prompt[layer_index])\n            layer_index = layer_index + 1\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.__init__","uri":"program://LLaMA-Adapter/function/llama.model.__init__#L198-L216","kind":"function","name":"__init__","path":"llama/model.py","language":"python","start_line":198,"end_line":216,"context_start_line":178,"context_end_line":236,"code":"    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        prompt = self.adapter_query.weight.reshape(\n            self.params.adapter_layer, self.params.adapter_len, self.params.dim\n        ).unsqueeze(1)\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers[: -1 * self.params.adapter_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        layer_index = 0\n        for layer in self.layers[-1 * self.params.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, prompt[layer_index])","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model._norm","uri":"program://LLaMA-Adapter/function/llama.model._norm#L37-L38","kind":"function","name":"_norm","path":"llama/model.py","language":"python","start_line":37,"end_line":38,"context_start_line":17,"context_end_line":58,"code":"    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    adapter_len: int = 10\n    adapter_layer: int = 8\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.model.forward","uri":"program://LLaMA-Adapter/function/llama.model.forward#L219-L240","kind":"function","name":"forward","path":"llama/model.py","language":"python","start_line":219,"end_line":240,"context_start_line":199,"context_end_line":240,"code":"        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        prompt = self.adapter_query.weight.reshape(\n            self.params.adapter_layer, self.params.adapter_len, self.params.dim\n        ).unsqueeze(1)\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers[: -1 * self.params.adapter_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        layer_index = 0\n        for layer in self.layers[-1 * self.params.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, prompt[layer_index])\n            layer_index = layer_index + 1\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.tokenizer","uri":"program://LLaMA-Adapter/module/llama.tokenizer#L1-L38","kind":"module","name":"llama.tokenizer","path":"llama/tokenizer.py","language":"python","start_line":1,"end_line":38,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.tokenizer.Tokenizer","uri":"program://LLaMA-Adapter/class/llama.tokenizer.Tokenizer#L13-L38","kind":"class","name":"Tokenizer","path":"llama/tokenizer.py","language":"python","start_line":13,"end_line":38,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.tokenizer.__init__","uri":"program://LLaMA-Adapter/function/llama.tokenizer.__init__#L14-L26","kind":"function","name":"__init__","path":"llama/tokenizer.py","language":"python","start_line":14,"end_line":26,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.tokenizer.encode","uri":"program://LLaMA-Adapter/function/llama.tokenizer.encode#L28-L35","kind":"function","name":"encode","path":"llama/tokenizer.py","language":"python","start_line":28,"end_line":35,"context_start_line":8,"context_end_line":38,"code":"from sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama.tokenizer.decode","uri":"program://LLaMA-Adapter/function/llama.tokenizer.decode#L37-L38","kind":"function","name":"decode","path":"llama/tokenizer.py","language":"python","start_line":37,"end_line":38,"context_start_line":17,"context_end_line":38,"code":"        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.demo","uri":"program://LLaMA-Adapter/module/gorilla.finetune.demo#L1-L137","kind":"module","name":"gorilla.finetune.demo","path":"gorilla/finetune/demo.py","language":"python","start_line":1,"end_line":137,"context_start_line":1,"context_end_line":137,"code":"import json\nimport os\nimport sys\nimport time\nfrom pathlib import Path\nfrom typing import List, Tuple\n\nimport fire\nimport torch\nfrom model.tokenizer import Tokenizer\nfrom model.meta import MetaModel\n\n\nclass LLaMA:\n    def __init__(self, model, tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.llma.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(\n            x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full(\n            (bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.llma.forward_inference(tokens[:, :cur_pos], 0)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = self.sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\n    def sample_top_p(self, probs, p):\n        probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n        probs_sum = torch.cumsum(probs_sort, dim=-1)\n        mask = probs_sum - probs_sort > p\n        probs_sort[mask] = 0.0\n        probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n        next_token = torch.multinomial(probs_sort, num_samples=1)\n        next_token = torch.gather(probs_idx, -1, next_token)\n        return next_token\n\n\ndef main(\n    ckpt_path: str,\n    tokenizer_path: str,\n    llama_type: str = \"llama\",\n    reversible_grad: bool = False,\n    temperature: float = 0.8,\n    top_p: float = 0.95,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n):\n    ckpt = torch.load(ckpt_path, map_location='cpu')\n    model = MetaModel(llama_type, reversible_grad).cuda()\n    model.load_state_dict(ckpt['model'])\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    generator = LLaMA(model, tokenizer)\n\n    prompts = [\n        # For these prompts, the expected answer is the natural continuation of the prompt\n        \"I believe the meaning of life is\",\n        \"Simply put, the theory of relativity states that \",\n        \"Building a website can be done in 10 simple steps:\\n\",\n        # Few shot prompts: https://huggingface.co/blog/few-shot-learning-gpt-neo-and-inference-api\n        \"\"\"Tweet: \"I hate it when my phone battery dies.\"\nSentiment: Negative\n###\nTweet: \"My day has been 👍\"\nSentiment: Positive\n###\nTweet: \"This is the link to the article\"\nSentiment: Neutral\n###\nTweet: \"This new music video was incredibile\"\nSentiment:\"\"\",\n        \"\"\"Translate English to French:\n\nsea otter => loutre de mer\n\npeppermint => menthe poivrée\n\nplush girafe => girafe peluche\n\ncheese =>\"\"\",\n    ]\n    results = generator.generate(\n        prompts, max_gen_len=256, temperature=temperature, top_p=top_p\n    )\n\n    for result in results:\n        print(result)\n        print(\"\\n==================================\\n\")\n\n\nif __name__ == \"__main__\":\n    fire.Fire(main)","source_hash":"96f9e8f2bc294b12c1871e05e7ad362d902383cf881a3efc503a3ea7c865f265","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.demo.LLaMA","uri":"program://LLaMA-Adapter/class/gorilla.finetune.demo.LLaMA#L14-L81","kind":"class","name":"LLaMA","path":"gorilla/finetune/demo.py","language":"python","start_line":14,"end_line":81,"context_start_line":1,"context_end_line":101,"code":"import json\nimport os\nimport sys\nimport time\nfrom pathlib import Path\nfrom typing import List, Tuple\n\nimport fire\nimport torch\nfrom model.tokenizer import Tokenizer\nfrom model.meta import MetaModel\n\n\nclass LLaMA:\n    def __init__(self, model, tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.llma.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(\n            x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full(\n            (bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.llma.forward_inference(tokens[:, :cur_pos], 0)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = self.sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\n    def sample_top_p(self, probs, p):\n        probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n        probs_sum = torch.cumsum(probs_sort, dim=-1)\n        mask = probs_sum - probs_sort > p\n        probs_sort[mask] = 0.0\n        probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n        next_token = torch.multinomial(probs_sort, num_samples=1)\n        next_token = torch.gather(probs_idx, -1, next_token)\n        return next_token\n\n\ndef main(\n    ckpt_path: str,\n    tokenizer_path: str,\n    llama_type: str = \"llama\",\n    reversible_grad: bool = False,\n    temperature: float = 0.8,\n    top_p: float = 0.95,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n):\n    ckpt = torch.load(ckpt_path, map_location='cpu')\n    model = MetaModel(llama_type, reversible_grad).cuda()\n    model.load_state_dict(ckpt['model'])\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    generator = LLaMA(model, tokenizer)\n\n    prompts = [\n        # For these prompts, the expected answer is the natural continuation of the prompt","source_hash":"96f9e8f2bc294b12c1871e05e7ad362d902383cf881a3efc503a3ea7c865f265","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.demo.main","uri":"program://LLaMA-Adapter/function/gorilla.finetune.demo.main#L84-L133","kind":"function","name":"main","path":"gorilla/finetune/demo.py","language":"python","start_line":84,"end_line":133,"context_start_line":64,"context_end_line":137,"code":"            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\n    def sample_top_p(self, probs, p):\n        probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n        probs_sum = torch.cumsum(probs_sort, dim=-1)\n        mask = probs_sum - probs_sort > p\n        probs_sort[mask] = 0.0\n        probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n        next_token = torch.multinomial(probs_sort, num_samples=1)\n        next_token = torch.gather(probs_idx, -1, next_token)\n        return next_token\n\n\ndef main(\n    ckpt_path: str,\n    tokenizer_path: str,\n    llama_type: str = \"llama\",\n    reversible_grad: bool = False,\n    temperature: float = 0.8,\n    top_p: float = 0.95,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n):\n    ckpt = torch.load(ckpt_path, map_location='cpu')\n    model = MetaModel(llama_type, reversible_grad).cuda()\n    model.load_state_dict(ckpt['model'])\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    generator = LLaMA(model, tokenizer)\n\n    prompts = [\n        # For these prompts, the expected answer is the natural continuation of the prompt\n        \"I believe the meaning of life is\",\n        \"Simply put, the theory of relativity states that \",\n        \"Building a website can be done in 10 simple steps:\\n\",\n        # Few shot prompts: https://huggingface.co/blog/few-shot-learning-gpt-neo-and-inference-api\n        \"\"\"Tweet: \"I hate it when my phone battery dies.\"\nSentiment: Negative\n###\nTweet: \"My day has been 👍\"\nSentiment: Positive\n###\nTweet: \"This is the link to the article\"\nSentiment: Neutral\n###\nTweet: \"This new music video was incredibile\"\nSentiment:\"\"\",\n        \"\"\"Translate English to French:\n\nsea otter => loutre de mer\n\npeppermint => menthe poivrée\n\nplush girafe => girafe peluche\n\ncheese =>\"\"\",\n    ]\n    results = generator.generate(\n        prompts, max_gen_len=256, temperature=temperature, top_p=top_p\n    )\n\n    for result in results:\n        print(result)\n        print(\"\\n==================================\\n\")\n\n\nif __name__ == \"__main__\":\n    fire.Fire(main)","source_hash":"96f9e8f2bc294b12c1871e05e7ad362d902383cf881a3efc503a3ea7c865f265","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.demo.__init__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.demo.__init__#L15-L17","kind":"function","name":"__init__","path":"gorilla/finetune/demo.py","language":"python","start_line":15,"end_line":17,"context_start_line":1,"context_end_line":37,"code":"import json\nimport os\nimport sys\nimport time\nfrom pathlib import Path\nfrom typing import List, Tuple\n\nimport fire\nimport torch\nfrom model.tokenizer import Tokenizer\nfrom model.meta import MetaModel\n\n\nclass LLaMA:\n    def __init__(self, model, tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.llma.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(\n            x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n","source_hash":"96f9e8f2bc294b12c1871e05e7ad362d902383cf881a3efc503a3ea7c865f265","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.demo.generate","uri":"program://LLaMA-Adapter/function/gorilla.finetune.demo.generate#L19-L70","kind":"function","name":"generate","path":"gorilla/finetune/demo.py","language":"python","start_line":19,"end_line":70,"context_start_line":1,"context_end_line":90,"code":"import json\nimport os\nimport sys\nimport time\nfrom pathlib import Path\nfrom typing import List, Tuple\n\nimport fire\nimport torch\nfrom model.tokenizer import Tokenizer\nfrom model.meta import MetaModel\n\n\nclass LLaMA:\n    def __init__(self, model, tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.llma.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(\n            x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full(\n            (bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.llma.forward_inference(tokens[:, :cur_pos], 0)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = self.sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\n    def sample_top_p(self, probs, p):\n        probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n        probs_sum = torch.cumsum(probs_sort, dim=-1)\n        mask = probs_sum - probs_sort > p\n        probs_sort[mask] = 0.0\n        probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n        next_token = torch.multinomial(probs_sort, num_samples=1)\n        next_token = torch.gather(probs_idx, -1, next_token)\n        return next_token\n\n\ndef main(\n    ckpt_path: str,\n    tokenizer_path: str,\n    llama_type: str = \"llama\",\n    reversible_grad: bool = False,\n    temperature: float = 0.8,\n    top_p: float = 0.95,","source_hash":"96f9e8f2bc294b12c1871e05e7ad362d902383cf881a3efc503a3ea7c865f265","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.demo.sample_top_p","uri":"program://LLaMA-Adapter/function/gorilla.finetune.demo.sample_top_p#L73-L81","kind":"function","name":"sample_top_p","path":"gorilla/finetune/demo.py","language":"python","start_line":73,"end_line":81,"context_start_line":53,"context_end_line":101,"code":"            # only replace token if prompt has already been generated\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\n    def sample_top_p(self, probs, p):\n        probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n        probs_sum = torch.cumsum(probs_sort, dim=-1)\n        mask = probs_sum - probs_sort > p\n        probs_sort[mask] = 0.0\n        probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n        next_token = torch.multinomial(probs_sort, num_samples=1)\n        next_token = torch.gather(probs_idx, -1, next_token)\n        return next_token\n\n\ndef main(\n    ckpt_path: str,\n    tokenizer_path: str,\n    llama_type: str = \"llama\",\n    reversible_grad: bool = False,\n    temperature: float = 0.8,\n    top_p: float = 0.95,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n):\n    ckpt = torch.load(ckpt_path, map_location='cpu')\n    model = MetaModel(llama_type, reversible_grad).cuda()\n    model.load_state_dict(ckpt['model'])\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    generator = LLaMA(model, tokenizer)\n\n    prompts = [\n        # For these prompts, the expected answer is the natural continuation of the prompt","source_hash":"96f9e8f2bc294b12c1871e05e7ad362d902383cf881a3efc503a3ea7c865f265","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.main_finetune","uri":"program://LLaMA-Adapter/module/gorilla.finetune.main_finetune#L1-L302","kind":"module","name":"gorilla.finetune.main_finetune","path":"gorilla/finetune/main_finetune.py","language":"python","start_line":1,"end_line":302,"context_start_line":1,"context_end_line":302,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\nimport functools\nfrom functools import partial\n\nimport torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.distributed.fsdp import (\n    FullyShardedDataParallel as FSDP,\n    MixedPrecision,\n    ShardingStrategy,\n)\nfrom torch.distributed.algorithms._checkpoint.checkpoint_wrapper import (\n   checkpoint_wrapper,\n   CheckpointImpl,\n   apply_activation_checkpointing,\n)\nfrom torch.distributed.fsdp.wrap import (\n    transformer_auto_wrap_policy,\n)\n\nfrom fairscale.nn.model_parallel import initialize as fs_init\n\nfrom apex.optimizers import FusedAdam\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom model.meta import MetaModel\nfrom model.LLM.llama import Attention, FeedForward\nfrom engine_finetune import train_one_epoch, val_one_epoch\nfrom torch.utils.data import Dataset\nfrom data.alpaca import FinetuneDataset, transform_train, transform_val\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('MAE pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=16, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--epochs', default=400, type=int)\n    parser.add_argument('--accum_iter', default=4, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='llama', type=str, metavar='MODEL', choices=['llama', 'revllama'],\n                        help='Name of model to train')\n\n    parser.add_argument('--llama_config', default='params.json', type=str,\n                        help='Path to llama model config')\n\n    parser.add_argument('--llama_tokenizer_path', default='/mnt/petrelfs/share_data/llm_llama/tokenizer.model', type=str,\n                        help='Path to llama tokenizer')\n\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='path to checkpoint from pretrain stage')\n    parser.add_argument('--pretrained_type', type=str, choices=['sharded', 'consolidated', 'meta_ori'],\n                        help='pretrained checkpoint save format')\n\n    parser.add_argument('--reversible_grad', action='store_true', default=False,\n                        help='Whether to use reversible grad')\n\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.02,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=0.001, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--min_lr', type=float, default=0.0001, metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_epochs', type=int, default=20000, metavar='N',\n                        help='iterations to warmup LR')\n\n    parser.add_argument('--clip_grad', type=int, default=-1,\n                        help='grad clipping norm')\n\n    # Dataset parameters\n    parser.add_argument('--max_words', default=1024, type=int,\n                        help='dataset path')\n    parser.add_argument('--data_config', default='/path/to/data/config/yaml', type=str,\n                        help='data config path')\n\n    parser.add_argument('--output_dir', default='./output_dir',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output_dir',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n    parser.add_argument('--resume', default='',\n                        help='resume from checkpoint')\n\n\n    parser.add_argument('--num_workers', default=5, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--model_parallel_size', type=int, default=1)\n    parser.add_argument('--data_parallel', type=str, choices=['ddp', 'sdp', 'fsdp'], default='sdp')\n    parser.add_argument('--precision', type=str, choices=['fp16', 'bf16', 'tf32'], default='bf16')\n    parser.add_argument('--save_freq', type=int, default=5000)\n    parser.add_argument('--save_consolidated', action=\"store_true\",\n                        help=\"save consolidated model weights along with regular checkpoints \"\n                             \"used to resume training. useful for convenient deployment but \"\n                             \"will occupy some additional disk space.\")\n    parser.add_argument('--checkpointing', action=\"store_true\", default=False,\n                        help=\"enable gradient checkopointing\")\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n    fs_init.initialize_model_parallel(args.model_parallel_size)\n    if args.precision == \"tf32\":\n        torch.backends.cuda.matmul.allow_tf32 = True\n        torch.backends.cudnn.allow_tf32 = True\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    global_rank = misc.get_rank()\n    mp_rank = fs_init.get_model_parallel_rank()\n    dp_rank = fs_init.get_data_parallel_rank()\n    dp_world_size = fs_init.get_data_parallel_world_size()\n    dp_group = fs_init.get_data_parallel_group()\n\n    dataset_train = FinetuneDataset(args.data_config, transform_train,\n                                    max_words=args.max_words, tokenizer_path=args.llama_tokenizer_path)\n    dataset_val = FinetuneDataset(args.data_config, transform_val,\n                                  max_words=args.max_words, tokenizer_path=args.llama_tokenizer_path)\n    print(dataset_train)\n\n\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n    sampler_train = torch.utils.data.DistributedSampler(\n        dataset_train, num_replicas=dp_world_size, rank=dp_rank, shuffle=True\n    )\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        sampler=sampler_train,\n        drop_last=True,\n    )\n\n    sampler_val = torch.utils.data.DistributedSampler(\n        dataset_val, num_replicas=dp_world_size, rank=dp_rank, shuffle=False\n    )\n    data_loader_val = torch.utils.data.DataLoader(\n        dataset_val,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        sampler=sampler_val,\n        drop_last=True,\n    )\n\n    \n    # define the model\n    model = MetaModel(args.llama_type, args.reversible_grad, args.llama_config)\n    print(f\"load pretrained from {args.pretrained_path}\")\n    misc.load_pretrained(args.pretrained_path, args, model)\n    print(\"Unwrapped Model = %s\" % str(model))\n\n    mixed_precision_dtype = {\n        \"fp16\": torch.float16,\n        \"bf16\": torch.bfloat16,\n        \"tf32\": torch.float32,\n    }[args.precision]\n    model = FSDP(\n        model,\n        process_group=fs_init.get_data_parallel_group(),\n        auto_wrap_policy=functools.partial(\n            transformer_auto_wrap_policy,\n            transformer_layer_cls=[Attention, FeedForward],\n        ),\n        limit_all_gathers=True,\n        use_orig_params=True,\n        sync_module_states=True,\n        mixed_precision=MixedPrecision(\n            param_dtype=mixed_precision_dtype,\n            reduce_dtype=mixed_precision_dtype,\n            buffer_dtype=mixed_precision_dtype,\n        ),\n        sharding_strategy={\n            \"sdp\": ShardingStrategy.SHARD_GRAD_OP,\n            \"ddp\": ShardingStrategy.NO_SHARD,\n            \"fsdp\": ShardingStrategy.FULL_SHARD,\n        }[args.data_parallel],\n        device_id=device\n    )\n\n    # gradient checkpointing\n    if args.checkpointing:\n        print(\"apply gradient checkpointing\")\n        non_reentrant_wrapper = partial(\n            checkpoint_wrapper,\n            offload_to_cpu=False,\n            checkpoint_impl=CheckpointImpl.NO_REENTRANT,\n        )\n        check_fn = lambda submodule: isinstance(submodule, (Attention, FeedForward))\n        apply_activation_checkpointing(model, checkpoint_wrapper_fn=non_reentrant_wrapper, check_fn=check_fn)\n\n    print(\"Model = %s\" % str(model))\n\n    eff_batch_size = args.batch_size * args.accum_iter * fs_init.get_data_parallel_world_size()\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model, args.weight_decay)\n    optimizer = FusedAdam(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler(args)\n\n    start_epoch = 0\n    if args.resume:\n        start_epoch, _ = misc.load_model(args=args, model=model, optimizer=optimizer, loss_scaler=loss_scaler, dataset_train=dataset_train)\n\n    print(f\"Start training\")\n    start_time = time.time()\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(start_epoch, args.epochs): # todo start epoch\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train,\n            optimizer, epoch, loss_scaler,\n            log_writer=log_writer,\n            args=args\n        )\n\n        if args.output_dir and (epoch % 5 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                output_dir=args.output_dir,\n                args=args, epoch=epoch, iteration=None, model=model, optimizer=optimizer,\n                loss_scaler=loss_scaler, dataset_state=None)\n\n        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},\n                     'epoch': epoch,\n                     **{f'val_{k}': v for k, v in train_stats.items()}}\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"719bf3b77c2305b815d1b3a47187e66b12c8034226c996c5995e3ff628d3204d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.main_finetune.get_args_parser","uri":"program://LLaMA-Adapter/function/gorilla.finetune.main_finetune.get_args_parser#L51-L135","kind":"function","name":"get_args_parser","path":"gorilla/finetune/main_finetune.py","language":"python","start_line":51,"end_line":135,"context_start_line":31,"context_end_line":155,"code":"   CheckpointImpl,\n   apply_activation_checkpointing,\n)\nfrom torch.distributed.fsdp.wrap import (\n    transformer_auto_wrap_policy,\n)\n\nfrom fairscale.nn.model_parallel import initialize as fs_init\n\nfrom apex.optimizers import FusedAdam\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom model.meta import MetaModel\nfrom model.LLM.llama import Attention, FeedForward\nfrom engine_finetune import train_one_epoch, val_one_epoch\nfrom torch.utils.data import Dataset\nfrom data.alpaca import FinetuneDataset, transform_train, transform_val\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('MAE pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=16, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--epochs', default=400, type=int)\n    parser.add_argument('--accum_iter', default=4, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='llama', type=str, metavar='MODEL', choices=['llama', 'revllama'],\n                        help='Name of model to train')\n\n    parser.add_argument('--llama_config', default='params.json', type=str,\n                        help='Path to llama model config')\n\n    parser.add_argument('--llama_tokenizer_path', default='/mnt/petrelfs/share_data/llm_llama/tokenizer.model', type=str,\n                        help='Path to llama tokenizer')\n\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='path to checkpoint from pretrain stage')\n    parser.add_argument('--pretrained_type', type=str, choices=['sharded', 'consolidated', 'meta_ori'],\n                        help='pretrained checkpoint save format')\n\n    parser.add_argument('--reversible_grad', action='store_true', default=False,\n                        help='Whether to use reversible grad')\n\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.02,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=0.001, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--min_lr', type=float, default=0.0001, metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_epochs', type=int, default=20000, metavar='N',\n                        help='iterations to warmup LR')\n\n    parser.add_argument('--clip_grad', type=int, default=-1,\n                        help='grad clipping norm')\n\n    # Dataset parameters\n    parser.add_argument('--max_words', default=1024, type=int,\n                        help='dataset path')\n    parser.add_argument('--data_config', default='/path/to/data/config/yaml', type=str,\n                        help='data config path')\n\n    parser.add_argument('--output_dir', default='./output_dir',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output_dir',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n    parser.add_argument('--resume', default='',\n                        help='resume from checkpoint')\n\n\n    parser.add_argument('--num_workers', default=5, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--model_parallel_size', type=int, default=1)\n    parser.add_argument('--data_parallel', type=str, choices=['ddp', 'sdp', 'fsdp'], default='sdp')\n    parser.add_argument('--precision', type=str, choices=['fp16', 'bf16', 'tf32'], default='bf16')\n    parser.add_argument('--save_freq', type=int, default=5000)\n    parser.add_argument('--save_consolidated', action=\"store_true\",\n                        help=\"save consolidated model weights along with regular checkpoints \"\n                             \"used to resume training. useful for convenient deployment but \"\n                             \"will occupy some additional disk space.\")\n    parser.add_argument('--checkpointing', action=\"store_true\", default=False,\n                        help=\"enable gradient checkopointing\")\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n    fs_init.initialize_model_parallel(args.model_parallel_size)\n    if args.precision == \"tf32\":\n        torch.backends.cuda.matmul.allow_tf32 = True\n        torch.backends.cudnn.allow_tf32 = True\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n","source_hash":"719bf3b77c2305b815d1b3a47187e66b12c8034226c996c5995e3ff628d3204d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.main_finetune.main","uri":"program://LLaMA-Adapter/function/gorilla.finetune.main_finetune.main#L138-L294","kind":"function","name":"main","path":"gorilla/finetune/main_finetune.py","language":"python","start_line":138,"end_line":294,"context_start_line":118,"context_end_line":302,"code":"                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--model_parallel_size', type=int, default=1)\n    parser.add_argument('--data_parallel', type=str, choices=['ddp', 'sdp', 'fsdp'], default='sdp')\n    parser.add_argument('--precision', type=str, choices=['fp16', 'bf16', 'tf32'], default='bf16')\n    parser.add_argument('--save_freq', type=int, default=5000)\n    parser.add_argument('--save_consolidated', action=\"store_true\",\n                        help=\"save consolidated model weights along with regular checkpoints \"\n                             \"used to resume training. useful for convenient deployment but \"\n                             \"will occupy some additional disk space.\")\n    parser.add_argument('--checkpointing', action=\"store_true\", default=False,\n                        help=\"enable gradient checkopointing\")\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n    fs_init.initialize_model_parallel(args.model_parallel_size)\n    if args.precision == \"tf32\":\n        torch.backends.cuda.matmul.allow_tf32 = True\n        torch.backends.cudnn.allow_tf32 = True\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    global_rank = misc.get_rank()\n    mp_rank = fs_init.get_model_parallel_rank()\n    dp_rank = fs_init.get_data_parallel_rank()\n    dp_world_size = fs_init.get_data_parallel_world_size()\n    dp_group = fs_init.get_data_parallel_group()\n\n    dataset_train = FinetuneDataset(args.data_config, transform_train,\n                                    max_words=args.max_words, tokenizer_path=args.llama_tokenizer_path)\n    dataset_val = FinetuneDataset(args.data_config, transform_val,\n                                  max_words=args.max_words, tokenizer_path=args.llama_tokenizer_path)\n    print(dataset_train)\n\n\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n    sampler_train = torch.utils.data.DistributedSampler(\n        dataset_train, num_replicas=dp_world_size, rank=dp_rank, shuffle=True\n    )\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        sampler=sampler_train,\n        drop_last=True,\n    )\n\n    sampler_val = torch.utils.data.DistributedSampler(\n        dataset_val, num_replicas=dp_world_size, rank=dp_rank, shuffle=False\n    )\n    data_loader_val = torch.utils.data.DataLoader(\n        dataset_val,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        sampler=sampler_val,\n        drop_last=True,\n    )\n\n    \n    # define the model\n    model = MetaModel(args.llama_type, args.reversible_grad, args.llama_config)\n    print(f\"load pretrained from {args.pretrained_path}\")\n    misc.load_pretrained(args.pretrained_path, args, model)\n    print(\"Unwrapped Model = %s\" % str(model))\n\n    mixed_precision_dtype = {\n        \"fp16\": torch.float16,\n        \"bf16\": torch.bfloat16,\n        \"tf32\": torch.float32,\n    }[args.precision]\n    model = FSDP(\n        model,\n        process_group=fs_init.get_data_parallel_group(),\n        auto_wrap_policy=functools.partial(\n            transformer_auto_wrap_policy,\n            transformer_layer_cls=[Attention, FeedForward],\n        ),\n        limit_all_gathers=True,\n        use_orig_params=True,\n        sync_module_states=True,\n        mixed_precision=MixedPrecision(\n            param_dtype=mixed_precision_dtype,\n            reduce_dtype=mixed_precision_dtype,\n            buffer_dtype=mixed_precision_dtype,\n        ),\n        sharding_strategy={\n            \"sdp\": ShardingStrategy.SHARD_GRAD_OP,\n            \"ddp\": ShardingStrategy.NO_SHARD,\n            \"fsdp\": ShardingStrategy.FULL_SHARD,\n        }[args.data_parallel],\n        device_id=device\n    )\n\n    # gradient checkpointing\n    if args.checkpointing:\n        print(\"apply gradient checkpointing\")\n        non_reentrant_wrapper = partial(\n            checkpoint_wrapper,\n            offload_to_cpu=False,\n            checkpoint_impl=CheckpointImpl.NO_REENTRANT,\n        )\n        check_fn = lambda submodule: isinstance(submodule, (Attention, FeedForward))\n        apply_activation_checkpointing(model, checkpoint_wrapper_fn=non_reentrant_wrapper, check_fn=check_fn)\n\n    print(\"Model = %s\" % str(model))\n\n    eff_batch_size = args.batch_size * args.accum_iter * fs_init.get_data_parallel_world_size()\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model, args.weight_decay)\n    optimizer = FusedAdam(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler(args)\n\n    start_epoch = 0\n    if args.resume:\n        start_epoch, _ = misc.load_model(args=args, model=model, optimizer=optimizer, loss_scaler=loss_scaler, dataset_train=dataset_train)\n\n    print(f\"Start training\")\n    start_time = time.time()\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(start_epoch, args.epochs): # todo start epoch\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train,\n            optimizer, epoch, loss_scaler,\n            log_writer=log_writer,\n            args=args\n        )\n\n        if args.output_dir and (epoch % 5 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                output_dir=args.output_dir,\n                args=args, epoch=epoch, iteration=None, model=model, optimizer=optimizer,\n                loss_scaler=loss_scaler, dataset_state=None)\n\n        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},\n                     'epoch': epoch,\n                     **{f'val_{k}': v for k, v in train_stats.items()}}\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"719bf3b77c2305b815d1b3a47187e66b12c8034226c996c5995e3ff628d3204d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer","uri":"program://LLaMA-Adapter/module/gorilla.finetune.transformer#L1-L299","kind":"module","name":"gorilla.finetune.transformer","path":"gorilla/finetune/transformer.py","language":"python","start_line":1,"end_line":299,"context_start_line":1,"context_end_line":299,"code":"# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved\n\"\"\"\nDETR Transformer class.\nCopy-paste from torch.nn.Transformer with modifications:\n    * positional encodings are passed in MHattention\n    * extra LN at the end of encoder is removed\n    * decoder returns a stack of activations from all decoding layers\n\"\"\"\nimport copy\nfrom typing import Optional, List\n\nimport torch\nimport torch.nn.functional as F\nfrom torch import nn, Tensor\nimport pdb\n\nclass Transformer(nn.Module):\n\n    def __init__(self, d_model=512, nhead=8, num_encoder_layers=6,\n                 num_decoder_layers=6, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False,\n                 return_intermediate_dec=False):\n        super().__init__()\n\n        encoder_layer = TransformerEncoderLayer(d_model, nhead, dim_feedforward,\n                                                dropout, activation, normalize_before)\n        encoder_norm = nn.LayerNorm(d_model) if normalize_before else None\n        self.encoder = TransformerEncoder(encoder_layer, num_encoder_layers, encoder_norm)\n\n        decoder_layer = TransformerDecoderLayer(d_model, nhead, dim_feedforward,\n                                                dropout, activation, normalize_before)\n        decoder_norm = nn.LayerNorm(d_model)\n        self.decoder = TransformerDecoder(decoder_layer, num_decoder_layers, decoder_norm,\n                                          return_intermediate=return_intermediate_dec)\n\n        self._reset_parameters()\n\n        self.d_model = d_model\n        self.nhead = nhead\n\n    def _reset_parameters(self):\n        for p in self.parameters():\n            if p.dim() > 1:\n                nn.init.xavier_uniform_(p)\n\n    def forward(self, src, mask, query_embed, pos_embed):\n        # flatten NxCxHxW to HWxNxC\n        bs, l, c = src.shape\n        src = src.permute(1, 0, 2)\n        pos_embed = pos_embed.permute(1, 0, 2)\n        query_embed = query_embed.unsqueeze(1).repeat(1, bs, 1)\n        pos_embed = pos_embed.repeat(1, bs, 1)\n        tgt = torch.zeros_like(query_embed)\n        memory = src\n#        memory = self.encoder(src, src_key_padding_mask=mask, pos=pos_embed)\n        hs = self.decoder(tgt, memory, memory_key_padding_mask=mask,\n                          pos=pos_embed, query_pos=query_embed)\n        return hs[-1].permute(1, 0, 2), memory\n\n\nclass TransformerEncoder(nn.Module):\n\n    def __init__(self, encoder_layer, num_layers, norm=None):\n        super().__init__()\n        self.layers = _get_clones(encoder_layer, num_layers)\n        self.num_layers = num_layers\n        self.norm = norm\n\n    def forward(self, src,\n                mask: Optional[Tensor] = None,\n                src_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None):\n        output = src\n\n        for layer in self.layers:\n            output = layer(output, src_mask=mask,\n                           src_key_padding_mask=src_key_padding_mask, pos=pos)\n\n        if self.norm is not None:\n            output = self.norm(output)\n\n        return output\n\n\nclass TransformerDecoder(nn.Module):\n\n    def __init__(self, decoder_layer, num_layers, norm=None, return_intermediate=False):\n        super().__init__()\n        self.layers = _get_clones(decoder_layer, num_layers)\n        self.num_layers = num_layers\n        self.norm = norm\n        self.return_intermediate = return_intermediate\n\n    def forward(self, tgt, memory,\n                tgt_mask: Optional[Tensor] = None,\n                memory_mask: Optional[Tensor] = None,\n                tgt_key_padding_mask: Optional[Tensor] = None,\n                memory_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None,\n                query_pos: Optional[Tensor] = None):\n        output = tgt\n\n        intermediate = []\n\n        for layer in self.layers:\n            output = layer(output, memory, tgt_mask=tgt_mask,\n                           memory_mask=memory_mask,\n                           tgt_key_padding_mask=tgt_key_padding_mask,\n                           memory_key_padding_mask=memory_key_padding_mask,\n                           pos=pos, query_pos=query_pos)\n            if self.return_intermediate:\n                intermediate.append(self.norm(output))\n\n        if self.norm is not None:\n            output = self.norm(output)\n            if self.return_intermediate:\n                intermediate.pop()\n                intermediate.append(output)\n\n        if self.return_intermediate:\n            return torch.stack(intermediate)\n\n        return output.unsqueeze(0)\n\n\nclass TransformerEncoderLayer(nn.Module):\n\n    def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False):\n        super().__init__()\n        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        # Implementation of Feedforward model\n        self.linear1 = nn.Linear(d_model, dim_feedforward)\n        self.dropout = nn.Dropout(dropout)\n        self.linear2 = nn.Linear(dim_feedforward, d_model)\n\n        self.norm1 = nn.LayerNorm(d_model)\n        self.norm2 = nn.LayerNorm(d_model)\n        self.dropout1 = nn.Dropout(dropout)\n        self.dropout2 = nn.Dropout(dropout)\n\n        self.activation = _get_activation_fn(activation)\n        self.normalize_before = normalize_before\n\n    def with_pos_embed(self, tensor, pos: Optional[Tensor]):\n        return tensor if pos is None else tensor + pos\n\n    def forward_post(self,\n                     src,\n                     src_mask: Optional[Tensor] = None,\n                     src_key_padding_mask: Optional[Tensor] = None,\n                     pos: Optional[Tensor] = None):\n        q = k = self.with_pos_embed(src, pos)\n        src2 = self.self_attn(q, k, value=src, attn_mask=src_mask,\n                              key_padding_mask=src_key_padding_mask)[0]\n        src = src + self.dropout1(src2)\n        src = self.norm1(src)\n        src2 = self.linear2(self.dropout(self.activation(self.linear1(src))))\n        src = src + self.dropout2(src2)\n        src = self.norm2(src)\n        return src\n\n    def forward_pre(self, src,\n                    src_mask: Optional[Tensor] = None,\n                    src_key_padding_mask: Optional[Tensor] = None,\n                    pos: Optional[Tensor] = None):\n        src2 = self.norm1(src)\n        q = k = self.with_pos_embed(src2, pos)\n        src2 = self.self_attn(q, k, value=src2, attn_mask=src_mask,\n                              key_padding_mask=src_key_padding_mask)[0]\n        src = src + self.dropout1(src2)\n        src2 = self.norm2(src)\n        src2 = self.linear2(self.dropout(self.activation(self.linear1(src2))))\n        src = src + self.dropout2(src2)\n        return src\n\n    def forward(self, src,\n                src_mask: Optional[Tensor] = None,\n                src_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None):\n        if self.normalize_before:\n            return self.forward_pre(src, src_mask, src_key_padding_mask, pos)\n        return self.forward_post(src, src_mask, src_key_padding_mask, pos)\n\n\nclass TransformerDecoderLayer(nn.Module):\n\n    def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False):\n        super().__init__()\n        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        self.multihead_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        # Implementation of Feedforward model\n        self.linear1 = nn.Linear(d_model, dim_feedforward)\n        self.dropout = nn.Dropout(dropout)\n        self.linear2 = nn.Linear(dim_feedforward, d_model)\n\n        self.norm1 = nn.LayerNorm(d_model)\n        self.norm2 = nn.LayerNorm(d_model)\n        self.norm3 = nn.LayerNorm(d_model)\n        self.dropout1 = nn.Dropout(dropout)\n        self.dropout2 = nn.Dropout(dropout)\n        self.dropout3 = nn.Dropout(dropout)\n\n        self.activation = _get_activation_fn(activation)\n        self.normalize_before = normalize_before\n\n    def with_pos_embed(self, tensor, pos: Optional[Tensor]):\n        return tensor if pos is None else tensor + pos\n\n    def forward_post(self, tgt, memory,\n                     tgt_mask: Optional[Tensor] = None,\n                     memory_mask: Optional[Tensor] = None,\n                     tgt_key_padding_mask: Optional[Tensor] = None,\n                     memory_key_padding_mask: Optional[Tensor] = None,\n                     pos: Optional[Tensor] = None,\n                     query_pos: Optional[Tensor] = None):\n        q = k = self.with_pos_embed(tgt, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt = self.norm1(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]\n        tgt = tgt + self.dropout2(tgt2)\n        tgt = self.norm2(tgt)\n        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt))))\n        tgt = tgt + self.dropout3(tgt2)\n        tgt = self.norm3(tgt)\n        return tgt\n\n    def forward_pre(self, tgt, memory,\n                    tgt_mask: Optional[Tensor] = None,\n                    memory_mask: Optional[Tensor] = None,\n                    tgt_key_padding_mask: Optional[Tensor] = None,\n                    memory_key_padding_mask: Optional[Tensor] = None,\n                    pos: Optional[Tensor] = None,\n                    query_pos: Optional[Tensor] = None):\n        tgt2 = self.norm1(tgt)\n        q = k = self.with_pos_embed(tgt2, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt2, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt2 = self.norm2(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt2, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]\n        tgt = tgt + self.dropout2(tgt2)\n        tgt2 = self.norm3(tgt)\n        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt2))))\n        tgt = tgt + self.dropout3(tgt2)\n        return tgt\n\n    def forward(self, tgt, memory,\n                tgt_mask: Optional[Tensor] = None,\n                memory_mask: Optional[Tensor] = None,\n                tgt_key_padding_mask: Optional[Tensor] = None,\n                memory_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None,\n                query_pos: Optional[Tensor] = None):\n        if self.normalize_before:\n            return self.forward_pre(tgt, memory, tgt_mask, memory_mask,\n                                    tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n        return self.forward_post(tgt, memory, tgt_mask, memory_mask,\n                                 tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n\n\ndef _get_clones(module, N):\n    return nn.ModuleList([copy.deepcopy(module) for i in range(N)])\n\n\ndef build_transformer(args):\n    return Transformer(\n        d_model=args.hidden_dim,\n        dropout=args.dropout,\n        nhead=args.nheads,\n        dim_feedforward=args.dim_feedforward,\n        num_encoder_layers=args.enc_layers,\n        num_decoder_layers=args.dec_layers,\n        normalize_before=args.pre_norm,\n        return_intermediate_dec=True,\n    )\n\n\ndef _get_activation_fn(activation):\n    \"\"\"Return an activation function given a string\"\"\"\n    if activation == \"relu\":\n        return F.relu\n    if activation == \"gelu\":\n        return F.gelu\n    if activation == \"glu\":\n        return F.glu\n    raise RuntimeError(F\"activation should be relu/gelu, not {activation}.\")\n\n\n","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.Transformer","uri":"program://LLaMA-Adapter/class/gorilla.finetune.transformer.Transformer#L17-L58","kind":"class","name":"Transformer","path":"gorilla/finetune/transformer.py","language":"python","start_line":17,"end_line":58,"context_start_line":1,"context_end_line":78,"code":"# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved\n\"\"\"\nDETR Transformer class.\nCopy-paste from torch.nn.Transformer with modifications:\n    * positional encodings are passed in MHattention\n    * extra LN at the end of encoder is removed\n    * decoder returns a stack of activations from all decoding layers\n\"\"\"\nimport copy\nfrom typing import Optional, List\n\nimport torch\nimport torch.nn.functional as F\nfrom torch import nn, Tensor\nimport pdb\n\nclass Transformer(nn.Module):\n\n    def __init__(self, d_model=512, nhead=8, num_encoder_layers=6,\n                 num_decoder_layers=6, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False,\n                 return_intermediate_dec=False):\n        super().__init__()\n\n        encoder_layer = TransformerEncoderLayer(d_model, nhead, dim_feedforward,\n                                                dropout, activation, normalize_before)\n        encoder_norm = nn.LayerNorm(d_model) if normalize_before else None\n        self.encoder = TransformerEncoder(encoder_layer, num_encoder_layers, encoder_norm)\n\n        decoder_layer = TransformerDecoderLayer(d_model, nhead, dim_feedforward,\n                                                dropout, activation, normalize_before)\n        decoder_norm = nn.LayerNorm(d_model)\n        self.decoder = TransformerDecoder(decoder_layer, num_decoder_layers, decoder_norm,\n                                          return_intermediate=return_intermediate_dec)\n\n        self._reset_parameters()\n\n        self.d_model = d_model\n        self.nhead = nhead\n\n    def _reset_parameters(self):\n        for p in self.parameters():\n            if p.dim() > 1:\n                nn.init.xavier_uniform_(p)\n\n    def forward(self, src, mask, query_embed, pos_embed):\n        # flatten NxCxHxW to HWxNxC\n        bs, l, c = src.shape\n        src = src.permute(1, 0, 2)\n        pos_embed = pos_embed.permute(1, 0, 2)\n        query_embed = query_embed.unsqueeze(1).repeat(1, bs, 1)\n        pos_embed = pos_embed.repeat(1, bs, 1)\n        tgt = torch.zeros_like(query_embed)\n        memory = src\n#        memory = self.encoder(src, src_key_padding_mask=mask, pos=pos_embed)\n        hs = self.decoder(tgt, memory, memory_key_padding_mask=mask,\n                          pos=pos_embed, query_pos=query_embed)\n        return hs[-1].permute(1, 0, 2), memory\n\n\nclass TransformerEncoder(nn.Module):\n\n    def __init__(self, encoder_layer, num_layers, norm=None):\n        super().__init__()\n        self.layers = _get_clones(encoder_layer, num_layers)\n        self.num_layers = num_layers\n        self.norm = norm\n\n    def forward(self, src,\n                mask: Optional[Tensor] = None,\n                src_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None):\n        output = src\n\n        for layer in self.layers:\n            output = layer(output, src_mask=mask,\n                           src_key_padding_mask=src_key_padding_mask, pos=pos)\n","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.TransformerEncoder","uri":"program://LLaMA-Adapter/class/gorilla.finetune.transformer.TransformerEncoder#L61-L82","kind":"class","name":"TransformerEncoder","path":"gorilla/finetune/transformer.py","language":"python","start_line":61,"end_line":82,"context_start_line":41,"context_end_line":102,"code":"    def _reset_parameters(self):\n        for p in self.parameters():\n            if p.dim() > 1:\n                nn.init.xavier_uniform_(p)\n\n    def forward(self, src, mask, query_embed, pos_embed):\n        # flatten NxCxHxW to HWxNxC\n        bs, l, c = src.shape\n        src = src.permute(1, 0, 2)\n        pos_embed = pos_embed.permute(1, 0, 2)\n        query_embed = query_embed.unsqueeze(1).repeat(1, bs, 1)\n        pos_embed = pos_embed.repeat(1, bs, 1)\n        tgt = torch.zeros_like(query_embed)\n        memory = src\n#        memory = self.encoder(src, src_key_padding_mask=mask, pos=pos_embed)\n        hs = self.decoder(tgt, memory, memory_key_padding_mask=mask,\n                          pos=pos_embed, query_pos=query_embed)\n        return hs[-1].permute(1, 0, 2), memory\n\n\nclass TransformerEncoder(nn.Module):\n\n    def __init__(self, encoder_layer, num_layers, norm=None):\n        super().__init__()\n        self.layers = _get_clones(encoder_layer, num_layers)\n        self.num_layers = num_layers\n        self.norm = norm\n\n    def forward(self, src,\n                mask: Optional[Tensor] = None,\n                src_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None):\n        output = src\n\n        for layer in self.layers:\n            output = layer(output, src_mask=mask,\n                           src_key_padding_mask=src_key_padding_mask, pos=pos)\n\n        if self.norm is not None:\n            output = self.norm(output)\n\n        return output\n\n\nclass TransformerDecoder(nn.Module):\n\n    def __init__(self, decoder_layer, num_layers, norm=None, return_intermediate=False):\n        super().__init__()\n        self.layers = _get_clones(decoder_layer, num_layers)\n        self.num_layers = num_layers\n        self.norm = norm\n        self.return_intermediate = return_intermediate\n\n    def forward(self, tgt, memory,\n                tgt_mask: Optional[Tensor] = None,\n                memory_mask: Optional[Tensor] = None,\n                tgt_key_padding_mask: Optional[Tensor] = None,\n                memory_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None,\n                query_pos: Optional[Tensor] = None):\n        output = tgt\n","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.TransformerDecoder","uri":"program://LLaMA-Adapter/class/gorilla.finetune.transformer.TransformerDecoder#L85-L123","kind":"class","name":"TransformerDecoder","path":"gorilla/finetune/transformer.py","language":"python","start_line":85,"end_line":123,"context_start_line":65,"context_end_line":143,"code":"        self.layers = _get_clones(encoder_layer, num_layers)\n        self.num_layers = num_layers\n        self.norm = norm\n\n    def forward(self, src,\n                mask: Optional[Tensor] = None,\n                src_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None):\n        output = src\n\n        for layer in self.layers:\n            output = layer(output, src_mask=mask,\n                           src_key_padding_mask=src_key_padding_mask, pos=pos)\n\n        if self.norm is not None:\n            output = self.norm(output)\n\n        return output\n\n\nclass TransformerDecoder(nn.Module):\n\n    def __init__(self, decoder_layer, num_layers, norm=None, return_intermediate=False):\n        super().__init__()\n        self.layers = _get_clones(decoder_layer, num_layers)\n        self.num_layers = num_layers\n        self.norm = norm\n        self.return_intermediate = return_intermediate\n\n    def forward(self, tgt, memory,\n                tgt_mask: Optional[Tensor] = None,\n                memory_mask: Optional[Tensor] = None,\n                tgt_key_padding_mask: Optional[Tensor] = None,\n                memory_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None,\n                query_pos: Optional[Tensor] = None):\n        output = tgt\n\n        intermediate = []\n\n        for layer in self.layers:\n            output = layer(output, memory, tgt_mask=tgt_mask,\n                           memory_mask=memory_mask,\n                           tgt_key_padding_mask=tgt_key_padding_mask,\n                           memory_key_padding_mask=memory_key_padding_mask,\n                           pos=pos, query_pos=query_pos)\n            if self.return_intermediate:\n                intermediate.append(self.norm(output))\n\n        if self.norm is not None:\n            output = self.norm(output)\n            if self.return_intermediate:\n                intermediate.pop()\n                intermediate.append(output)\n\n        if self.return_intermediate:\n            return torch.stack(intermediate)\n\n        return output.unsqueeze(0)\n\n\nclass TransformerEncoderLayer(nn.Module):\n\n    def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False):\n        super().__init__()\n        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        # Implementation of Feedforward model\n        self.linear1 = nn.Linear(d_model, dim_feedforward)\n        self.dropout = nn.Dropout(dropout)\n        self.linear2 = nn.Linear(dim_feedforward, d_model)\n\n        self.norm1 = nn.LayerNorm(d_model)\n        self.norm2 = nn.LayerNorm(d_model)\n        self.dropout1 = nn.Dropout(dropout)\n        self.dropout2 = nn.Dropout(dropout)\n\n        self.activation = _get_activation_fn(activation)\n        self.normalize_before = normalize_before","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.TransformerEncoderLayer","uri":"program://LLaMA-Adapter/class/gorilla.finetune.transformer.TransformerEncoderLayer#L126-L183","kind":"class","name":"TransformerEncoderLayer","path":"gorilla/finetune/transformer.py","language":"python","start_line":126,"end_line":183,"context_start_line":106,"context_end_line":203,"code":"            output = layer(output, memory, tgt_mask=tgt_mask,\n                           memory_mask=memory_mask,\n                           tgt_key_padding_mask=tgt_key_padding_mask,\n                           memory_key_padding_mask=memory_key_padding_mask,\n                           pos=pos, query_pos=query_pos)\n            if self.return_intermediate:\n                intermediate.append(self.norm(output))\n\n        if self.norm is not None:\n            output = self.norm(output)\n            if self.return_intermediate:\n                intermediate.pop()\n                intermediate.append(output)\n\n        if self.return_intermediate:\n            return torch.stack(intermediate)\n\n        return output.unsqueeze(0)\n\n\nclass TransformerEncoderLayer(nn.Module):\n\n    def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False):\n        super().__init__()\n        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        # Implementation of Feedforward model\n        self.linear1 = nn.Linear(d_model, dim_feedforward)\n        self.dropout = nn.Dropout(dropout)\n        self.linear2 = nn.Linear(dim_feedforward, d_model)\n\n        self.norm1 = nn.LayerNorm(d_model)\n        self.norm2 = nn.LayerNorm(d_model)\n        self.dropout1 = nn.Dropout(dropout)\n        self.dropout2 = nn.Dropout(dropout)\n\n        self.activation = _get_activation_fn(activation)\n        self.normalize_before = normalize_before\n\n    def with_pos_embed(self, tensor, pos: Optional[Tensor]):\n        return tensor if pos is None else tensor + pos\n\n    def forward_post(self,\n                     src,\n                     src_mask: Optional[Tensor] = None,\n                     src_key_padding_mask: Optional[Tensor] = None,\n                     pos: Optional[Tensor] = None):\n        q = k = self.with_pos_embed(src, pos)\n        src2 = self.self_attn(q, k, value=src, attn_mask=src_mask,\n                              key_padding_mask=src_key_padding_mask)[0]\n        src = src + self.dropout1(src2)\n        src = self.norm1(src)\n        src2 = self.linear2(self.dropout(self.activation(self.linear1(src))))\n        src = src + self.dropout2(src2)\n        src = self.norm2(src)\n        return src\n\n    def forward_pre(self, src,\n                    src_mask: Optional[Tensor] = None,\n                    src_key_padding_mask: Optional[Tensor] = None,\n                    pos: Optional[Tensor] = None):\n        src2 = self.norm1(src)\n        q = k = self.with_pos_embed(src2, pos)\n        src2 = self.self_attn(q, k, value=src2, attn_mask=src_mask,\n                              key_padding_mask=src_key_padding_mask)[0]\n        src = src + self.dropout1(src2)\n        src2 = self.norm2(src)\n        src2 = self.linear2(self.dropout(self.activation(self.linear1(src2))))\n        src = src + self.dropout2(src2)\n        return src\n\n    def forward(self, src,\n                src_mask: Optional[Tensor] = None,\n                src_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None):\n        if self.normalize_before:\n            return self.forward_pre(src, src_mask, src_key_padding_mask, pos)\n        return self.forward_post(src, src_mask, src_key_padding_mask, pos)\n\n\nclass TransformerDecoderLayer(nn.Module):\n\n    def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False):\n        super().__init__()\n        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        self.multihead_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        # Implementation of Feedforward model\n        self.linear1 = nn.Linear(d_model, dim_feedforward)\n        self.dropout = nn.Dropout(dropout)\n        self.linear2 = nn.Linear(dim_feedforward, d_model)\n\n        self.norm1 = nn.LayerNorm(d_model)\n        self.norm2 = nn.LayerNorm(d_model)\n        self.norm3 = nn.LayerNorm(d_model)\n        self.dropout1 = nn.Dropout(dropout)\n        self.dropout2 = nn.Dropout(dropout)\n        self.dropout3 = nn.Dropout(dropout)","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.TransformerDecoderLayer","uri":"program://LLaMA-Adapter/class/gorilla.finetune.transformer.TransformerDecoderLayer#L186-L268","kind":"class","name":"TransformerDecoderLayer","path":"gorilla/finetune/transformer.py","language":"python","start_line":186,"end_line":268,"context_start_line":166,"context_end_line":288,"code":"                    pos: Optional[Tensor] = None):\n        src2 = self.norm1(src)\n        q = k = self.with_pos_embed(src2, pos)\n        src2 = self.self_attn(q, k, value=src2, attn_mask=src_mask,\n                              key_padding_mask=src_key_padding_mask)[0]\n        src = src + self.dropout1(src2)\n        src2 = self.norm2(src)\n        src2 = self.linear2(self.dropout(self.activation(self.linear1(src2))))\n        src = src + self.dropout2(src2)\n        return src\n\n    def forward(self, src,\n                src_mask: Optional[Tensor] = None,\n                src_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None):\n        if self.normalize_before:\n            return self.forward_pre(src, src_mask, src_key_padding_mask, pos)\n        return self.forward_post(src, src_mask, src_key_padding_mask, pos)\n\n\nclass TransformerDecoderLayer(nn.Module):\n\n    def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False):\n        super().__init__()\n        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        self.multihead_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        # Implementation of Feedforward model\n        self.linear1 = nn.Linear(d_model, dim_feedforward)\n        self.dropout = nn.Dropout(dropout)\n        self.linear2 = nn.Linear(dim_feedforward, d_model)\n\n        self.norm1 = nn.LayerNorm(d_model)\n        self.norm2 = nn.LayerNorm(d_model)\n        self.norm3 = nn.LayerNorm(d_model)\n        self.dropout1 = nn.Dropout(dropout)\n        self.dropout2 = nn.Dropout(dropout)\n        self.dropout3 = nn.Dropout(dropout)\n\n        self.activation = _get_activation_fn(activation)\n        self.normalize_before = normalize_before\n\n    def with_pos_embed(self, tensor, pos: Optional[Tensor]):\n        return tensor if pos is None else tensor + pos\n\n    def forward_post(self, tgt, memory,\n                     tgt_mask: Optional[Tensor] = None,\n                     memory_mask: Optional[Tensor] = None,\n                     tgt_key_padding_mask: Optional[Tensor] = None,\n                     memory_key_padding_mask: Optional[Tensor] = None,\n                     pos: Optional[Tensor] = None,\n                     query_pos: Optional[Tensor] = None):\n        q = k = self.with_pos_embed(tgt, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt = self.norm1(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]\n        tgt = tgt + self.dropout2(tgt2)\n        tgt = self.norm2(tgt)\n        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt))))\n        tgt = tgt + self.dropout3(tgt2)\n        tgt = self.norm3(tgt)\n        return tgt\n\n    def forward_pre(self, tgt, memory,\n                    tgt_mask: Optional[Tensor] = None,\n                    memory_mask: Optional[Tensor] = None,\n                    tgt_key_padding_mask: Optional[Tensor] = None,\n                    memory_key_padding_mask: Optional[Tensor] = None,\n                    pos: Optional[Tensor] = None,\n                    query_pos: Optional[Tensor] = None):\n        tgt2 = self.norm1(tgt)\n        q = k = self.with_pos_embed(tgt2, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt2, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt2 = self.norm2(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt2, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]\n        tgt = tgt + self.dropout2(tgt2)\n        tgt2 = self.norm3(tgt)\n        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt2))))\n        tgt = tgt + self.dropout3(tgt2)\n        return tgt\n\n    def forward(self, tgt, memory,\n                tgt_mask: Optional[Tensor] = None,\n                memory_mask: Optional[Tensor] = None,\n                tgt_key_padding_mask: Optional[Tensor] = None,\n                memory_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None,\n                query_pos: Optional[Tensor] = None):\n        if self.normalize_before:\n            return self.forward_pre(tgt, memory, tgt_mask, memory_mask,\n                                    tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n        return self.forward_post(tgt, memory, tgt_mask, memory_mask,\n                                 tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n\n\ndef _get_clones(module, N):\n    return nn.ModuleList([copy.deepcopy(module) for i in range(N)])\n\n\ndef build_transformer(args):\n    return Transformer(\n        d_model=args.hidden_dim,\n        dropout=args.dropout,\n        nhead=args.nheads,\n        dim_feedforward=args.dim_feedforward,\n        num_encoder_layers=args.enc_layers,\n        num_decoder_layers=args.dec_layers,\n        normalize_before=args.pre_norm,\n        return_intermediate_dec=True,\n    )\n\n\ndef _get_activation_fn(activation):","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer._get_clones","uri":"program://LLaMA-Adapter/function/gorilla.finetune.transformer._get_clones#L271-L272","kind":"function","name":"_get_clones","path":"gorilla/finetune/transformer.py","language":"python","start_line":271,"end_line":272,"context_start_line":251,"context_end_line":292,"code":"        tgt = tgt + self.dropout2(tgt2)\n        tgt2 = self.norm3(tgt)\n        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt2))))\n        tgt = tgt + self.dropout3(tgt2)\n        return tgt\n\n    def forward(self, tgt, memory,\n                tgt_mask: Optional[Tensor] = None,\n                memory_mask: Optional[Tensor] = None,\n                tgt_key_padding_mask: Optional[Tensor] = None,\n                memory_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None,\n                query_pos: Optional[Tensor] = None):\n        if self.normalize_before:\n            return self.forward_pre(tgt, memory, tgt_mask, memory_mask,\n                                    tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n        return self.forward_post(tgt, memory, tgt_mask, memory_mask,\n                                 tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n\n\ndef _get_clones(module, N):\n    return nn.ModuleList([copy.deepcopy(module) for i in range(N)])\n\n\ndef build_transformer(args):\n    return Transformer(\n        d_model=args.hidden_dim,\n        dropout=args.dropout,\n        nhead=args.nheads,\n        dim_feedforward=args.dim_feedforward,\n        num_encoder_layers=args.enc_layers,\n        num_decoder_layers=args.dec_layers,\n        normalize_before=args.pre_norm,\n        return_intermediate_dec=True,\n    )\n\n\ndef _get_activation_fn(activation):\n    \"\"\"Return an activation function given a string\"\"\"\n    if activation == \"relu\":\n        return F.relu\n    if activation == \"gelu\":","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.build_transformer","uri":"program://LLaMA-Adapter/function/gorilla.finetune.transformer.build_transformer#L275-L285","kind":"function","name":"build_transformer","path":"gorilla/finetune/transformer.py","language":"python","start_line":275,"end_line":285,"context_start_line":255,"context_end_line":299,"code":"        return tgt\n\n    def forward(self, tgt, memory,\n                tgt_mask: Optional[Tensor] = None,\n                memory_mask: Optional[Tensor] = None,\n                tgt_key_padding_mask: Optional[Tensor] = None,\n                memory_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None,\n                query_pos: Optional[Tensor] = None):\n        if self.normalize_before:\n            return self.forward_pre(tgt, memory, tgt_mask, memory_mask,\n                                    tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n        return self.forward_post(tgt, memory, tgt_mask, memory_mask,\n                                 tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n\n\ndef _get_clones(module, N):\n    return nn.ModuleList([copy.deepcopy(module) for i in range(N)])\n\n\ndef build_transformer(args):\n    return Transformer(\n        d_model=args.hidden_dim,\n        dropout=args.dropout,\n        nhead=args.nheads,\n        dim_feedforward=args.dim_feedforward,\n        num_encoder_layers=args.enc_layers,\n        num_decoder_layers=args.dec_layers,\n        normalize_before=args.pre_norm,\n        return_intermediate_dec=True,\n    )\n\n\ndef _get_activation_fn(activation):\n    \"\"\"Return an activation function given a string\"\"\"\n    if activation == \"relu\":\n        return F.relu\n    if activation == \"gelu\":\n        return F.gelu\n    if activation == \"glu\":\n        return F.glu\n    raise RuntimeError(F\"activation should be relu/gelu, not {activation}.\")\n\n\n","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer._get_activation_fn","uri":"program://LLaMA-Adapter/function/gorilla.finetune.transformer._get_activation_fn#L288-L296","kind":"function","name":"_get_activation_fn","path":"gorilla/finetune/transformer.py","language":"python","start_line":288,"end_line":296,"context_start_line":268,"context_end_line":299,"code":"                                 tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n\n\ndef _get_clones(module, N):\n    return nn.ModuleList([copy.deepcopy(module) for i in range(N)])\n\n\ndef build_transformer(args):\n    return Transformer(\n        d_model=args.hidden_dim,\n        dropout=args.dropout,\n        nhead=args.nheads,\n        dim_feedforward=args.dim_feedforward,\n        num_encoder_layers=args.enc_layers,\n        num_decoder_layers=args.dec_layers,\n        normalize_before=args.pre_norm,\n        return_intermediate_dec=True,\n    )\n\n\ndef _get_activation_fn(activation):\n    \"\"\"Return an activation function given a string\"\"\"\n    if activation == \"relu\":\n        return F.relu\n    if activation == \"gelu\":\n        return F.gelu\n    if activation == \"glu\":\n        return F.glu\n    raise RuntimeError(F\"activation should be relu/gelu, not {activation}.\")\n\n\n","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.__init__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.transformer.__init__#L188-L206","kind":"function","name":"__init__","path":"gorilla/finetune/transformer.py","language":"python","start_line":188,"end_line":206,"context_start_line":168,"context_end_line":226,"code":"        q = k = self.with_pos_embed(src2, pos)\n        src2 = self.self_attn(q, k, value=src2, attn_mask=src_mask,\n                              key_padding_mask=src_key_padding_mask)[0]\n        src = src + self.dropout1(src2)\n        src2 = self.norm2(src)\n        src2 = self.linear2(self.dropout(self.activation(self.linear1(src2))))\n        src = src + self.dropout2(src2)\n        return src\n\n    def forward(self, src,\n                src_mask: Optional[Tensor] = None,\n                src_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None):\n        if self.normalize_before:\n            return self.forward_pre(src, src_mask, src_key_padding_mask, pos)\n        return self.forward_post(src, src_mask, src_key_padding_mask, pos)\n\n\nclass TransformerDecoderLayer(nn.Module):\n\n    def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False):\n        super().__init__()\n        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        self.multihead_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        # Implementation of Feedforward model\n        self.linear1 = nn.Linear(d_model, dim_feedforward)\n        self.dropout = nn.Dropout(dropout)\n        self.linear2 = nn.Linear(dim_feedforward, d_model)\n\n        self.norm1 = nn.LayerNorm(d_model)\n        self.norm2 = nn.LayerNorm(d_model)\n        self.norm3 = nn.LayerNorm(d_model)\n        self.dropout1 = nn.Dropout(dropout)\n        self.dropout2 = nn.Dropout(dropout)\n        self.dropout3 = nn.Dropout(dropout)\n\n        self.activation = _get_activation_fn(activation)\n        self.normalize_before = normalize_before\n\n    def with_pos_embed(self, tensor, pos: Optional[Tensor]):\n        return tensor if pos is None else tensor + pos\n\n    def forward_post(self, tgt, memory,\n                     tgt_mask: Optional[Tensor] = None,\n                     memory_mask: Optional[Tensor] = None,\n                     tgt_key_padding_mask: Optional[Tensor] = None,\n                     memory_key_padding_mask: Optional[Tensor] = None,\n                     pos: Optional[Tensor] = None,\n                     query_pos: Optional[Tensor] = None):\n        q = k = self.with_pos_embed(tgt, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt = self.norm1(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer._reset_parameters","uri":"program://LLaMA-Adapter/function/gorilla.finetune.transformer._reset_parameters#L41-L44","kind":"function","name":"_reset_parameters","path":"gorilla/finetune/transformer.py","language":"python","start_line":41,"end_line":44,"context_start_line":21,"context_end_line":64,"code":"                 activation=\"relu\", normalize_before=False,\n                 return_intermediate_dec=False):\n        super().__init__()\n\n        encoder_layer = TransformerEncoderLayer(d_model, nhead, dim_feedforward,\n                                                dropout, activation, normalize_before)\n        encoder_norm = nn.LayerNorm(d_model) if normalize_before else None\n        self.encoder = TransformerEncoder(encoder_layer, num_encoder_layers, encoder_norm)\n\n        decoder_layer = TransformerDecoderLayer(d_model, nhead, dim_feedforward,\n                                                dropout, activation, normalize_before)\n        decoder_norm = nn.LayerNorm(d_model)\n        self.decoder = TransformerDecoder(decoder_layer, num_decoder_layers, decoder_norm,\n                                          return_intermediate=return_intermediate_dec)\n\n        self._reset_parameters()\n\n        self.d_model = d_model\n        self.nhead = nhead\n\n    def _reset_parameters(self):\n        for p in self.parameters():\n            if p.dim() > 1:\n                nn.init.xavier_uniform_(p)\n\n    def forward(self, src, mask, query_embed, pos_embed):\n        # flatten NxCxHxW to HWxNxC\n        bs, l, c = src.shape\n        src = src.permute(1, 0, 2)\n        pos_embed = pos_embed.permute(1, 0, 2)\n        query_embed = query_embed.unsqueeze(1).repeat(1, bs, 1)\n        pos_embed = pos_embed.repeat(1, bs, 1)\n        tgt = torch.zeros_like(query_embed)\n        memory = src\n#        memory = self.encoder(src, src_key_padding_mask=mask, pos=pos_embed)\n        hs = self.decoder(tgt, memory, memory_key_padding_mask=mask,\n                          pos=pos_embed, query_pos=query_embed)\n        return hs[-1].permute(1, 0, 2), memory\n\n\nclass TransformerEncoder(nn.Module):\n\n    def __init__(self, encoder_layer, num_layers, norm=None):\n        super().__init__()","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.forward","uri":"program://LLaMA-Adapter/function/gorilla.finetune.transformer.forward#L257-L268","kind":"function","name":"forward","path":"gorilla/finetune/transformer.py","language":"python","start_line":257,"end_line":268,"context_start_line":237,"context_end_line":288,"code":"                    tgt_key_padding_mask: Optional[Tensor] = None,\n                    memory_key_padding_mask: Optional[Tensor] = None,\n                    pos: Optional[Tensor] = None,\n                    query_pos: Optional[Tensor] = None):\n        tgt2 = self.norm1(tgt)\n        q = k = self.with_pos_embed(tgt2, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt2, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt2 = self.norm2(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt2, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]\n        tgt = tgt + self.dropout2(tgt2)\n        tgt2 = self.norm3(tgt)\n        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt2))))\n        tgt = tgt + self.dropout3(tgt2)\n        return tgt\n\n    def forward(self, tgt, memory,\n                tgt_mask: Optional[Tensor] = None,\n                memory_mask: Optional[Tensor] = None,\n                tgt_key_padding_mask: Optional[Tensor] = None,\n                memory_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None,\n                query_pos: Optional[Tensor] = None):\n        if self.normalize_before:\n            return self.forward_pre(tgt, memory, tgt_mask, memory_mask,\n                                    tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n        return self.forward_post(tgt, memory, tgt_mask, memory_mask,\n                                 tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n\n\ndef _get_clones(module, N):\n    return nn.ModuleList([copy.deepcopy(module) for i in range(N)])\n\n\ndef build_transformer(args):\n    return Transformer(\n        d_model=args.hidden_dim,\n        dropout=args.dropout,\n        nhead=args.nheads,\n        dim_feedforward=args.dim_feedforward,\n        num_encoder_layers=args.enc_layers,\n        num_decoder_layers=args.dec_layers,\n        normalize_before=args.pre_norm,\n        return_intermediate_dec=True,\n    )\n\n\ndef _get_activation_fn(activation):","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.with_pos_embed","uri":"program://LLaMA-Adapter/function/gorilla.finetune.transformer.with_pos_embed#L208-L209","kind":"function","name":"with_pos_embed","path":"gorilla/finetune/transformer.py","language":"python","start_line":208,"end_line":209,"context_start_line":188,"context_end_line":229,"code":"    def __init__(self, d_model, nhead, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False):\n        super().__init__()\n        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        self.multihead_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        # Implementation of Feedforward model\n        self.linear1 = nn.Linear(d_model, dim_feedforward)\n        self.dropout = nn.Dropout(dropout)\n        self.linear2 = nn.Linear(dim_feedforward, d_model)\n\n        self.norm1 = nn.LayerNorm(d_model)\n        self.norm2 = nn.LayerNorm(d_model)\n        self.norm3 = nn.LayerNorm(d_model)\n        self.dropout1 = nn.Dropout(dropout)\n        self.dropout2 = nn.Dropout(dropout)\n        self.dropout3 = nn.Dropout(dropout)\n\n        self.activation = _get_activation_fn(activation)\n        self.normalize_before = normalize_before\n\n    def with_pos_embed(self, tensor, pos: Optional[Tensor]):\n        return tensor if pos is None else tensor + pos\n\n    def forward_post(self, tgt, memory,\n                     tgt_mask: Optional[Tensor] = None,\n                     memory_mask: Optional[Tensor] = None,\n                     tgt_key_padding_mask: Optional[Tensor] = None,\n                     memory_key_padding_mask: Optional[Tensor] = None,\n                     pos: Optional[Tensor] = None,\n                     query_pos: Optional[Tensor] = None):\n        q = k = self.with_pos_embed(tgt, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt = self.norm1(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]\n        tgt = tgt + self.dropout2(tgt2)\n        tgt = self.norm2(tgt)\n        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt))))","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.forward_post","uri":"program://LLaMA-Adapter/function/gorilla.finetune.transformer.forward_post#L211-L232","kind":"function","name":"forward_post","path":"gorilla/finetune/transformer.py","language":"python","start_line":211,"end_line":232,"context_start_line":191,"context_end_line":252,"code":"        self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        self.multihead_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)\n        # Implementation of Feedforward model\n        self.linear1 = nn.Linear(d_model, dim_feedforward)\n        self.dropout = nn.Dropout(dropout)\n        self.linear2 = nn.Linear(dim_feedforward, d_model)\n\n        self.norm1 = nn.LayerNorm(d_model)\n        self.norm2 = nn.LayerNorm(d_model)\n        self.norm3 = nn.LayerNorm(d_model)\n        self.dropout1 = nn.Dropout(dropout)\n        self.dropout2 = nn.Dropout(dropout)\n        self.dropout3 = nn.Dropout(dropout)\n\n        self.activation = _get_activation_fn(activation)\n        self.normalize_before = normalize_before\n\n    def with_pos_embed(self, tensor, pos: Optional[Tensor]):\n        return tensor if pos is None else tensor + pos\n\n    def forward_post(self, tgt, memory,\n                     tgt_mask: Optional[Tensor] = None,\n                     memory_mask: Optional[Tensor] = None,\n                     tgt_key_padding_mask: Optional[Tensor] = None,\n                     memory_key_padding_mask: Optional[Tensor] = None,\n                     pos: Optional[Tensor] = None,\n                     query_pos: Optional[Tensor] = None):\n        q = k = self.with_pos_embed(tgt, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt = self.norm1(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]\n        tgt = tgt + self.dropout2(tgt2)\n        tgt = self.norm2(tgt)\n        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt))))\n        tgt = tgt + self.dropout3(tgt2)\n        tgt = self.norm3(tgt)\n        return tgt\n\n    def forward_pre(self, tgt, memory,\n                    tgt_mask: Optional[Tensor] = None,\n                    memory_mask: Optional[Tensor] = None,\n                    tgt_key_padding_mask: Optional[Tensor] = None,\n                    memory_key_padding_mask: Optional[Tensor] = None,\n                    pos: Optional[Tensor] = None,\n                    query_pos: Optional[Tensor] = None):\n        tgt2 = self.norm1(tgt)\n        q = k = self.with_pos_embed(tgt2, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt2, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt2 = self.norm2(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt2, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]\n        tgt = tgt + self.dropout2(tgt2)\n        tgt2 = self.norm3(tgt)","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.transformer.forward_pre","uri":"program://LLaMA-Adapter/function/gorilla.finetune.transformer.forward_pre#L234-L255","kind":"function","name":"forward_pre","path":"gorilla/finetune/transformer.py","language":"python","start_line":234,"end_line":255,"context_start_line":214,"context_end_line":275,"code":"                     tgt_key_padding_mask: Optional[Tensor] = None,\n                     memory_key_padding_mask: Optional[Tensor] = None,\n                     pos: Optional[Tensor] = None,\n                     query_pos: Optional[Tensor] = None):\n        q = k = self.with_pos_embed(tgt, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt = self.norm1(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]\n        tgt = tgt + self.dropout2(tgt2)\n        tgt = self.norm2(tgt)\n        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt))))\n        tgt = tgt + self.dropout3(tgt2)\n        tgt = self.norm3(tgt)\n        return tgt\n\n    def forward_pre(self, tgt, memory,\n                    tgt_mask: Optional[Tensor] = None,\n                    memory_mask: Optional[Tensor] = None,\n                    tgt_key_padding_mask: Optional[Tensor] = None,\n                    memory_key_padding_mask: Optional[Tensor] = None,\n                    pos: Optional[Tensor] = None,\n                    query_pos: Optional[Tensor] = None):\n        tgt2 = self.norm1(tgt)\n        q = k = self.with_pos_embed(tgt2, query_pos)\n        tgt2 = self.self_attn(q, k, value=tgt2, attn_mask=tgt_mask,\n                              key_padding_mask=tgt_key_padding_mask)[0]\n        tgt = tgt + self.dropout1(tgt2)\n        tgt2 = self.norm2(tgt)\n        tgt2 = self.multihead_attn(query=self.with_pos_embed(tgt2, query_pos),\n                                   key=self.with_pos_embed(memory, pos),\n                                   value=memory, attn_mask=memory_mask,\n                                   key_padding_mask=memory_key_padding_mask)[0]\n        tgt = tgt + self.dropout2(tgt2)\n        tgt2 = self.norm3(tgt)\n        tgt2 = self.linear2(self.dropout(self.activation(self.linear1(tgt2))))\n        tgt = tgt + self.dropout3(tgt2)\n        return tgt\n\n    def forward(self, tgt, memory,\n                tgt_mask: Optional[Tensor] = None,\n                memory_mask: Optional[Tensor] = None,\n                tgt_key_padding_mask: Optional[Tensor] = None,\n                memory_key_padding_mask: Optional[Tensor] = None,\n                pos: Optional[Tensor] = None,\n                query_pos: Optional[Tensor] = None):\n        if self.normalize_before:\n            return self.forward_pre(tgt, memory, tgt_mask, memory_mask,\n                                    tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n        return self.forward_post(tgt, memory, tgt_mask, memory_mask,\n                                 tgt_key_padding_mask, memory_key_padding_mask, pos, query_pos)\n\n\ndef _get_clones(module, N):\n    return nn.ModuleList([copy.deepcopy(module) for i in range(N)])\n\n\ndef build_transformer(args):","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.conversation","uri":"program://LLaMA-Adapter/module/gorilla.finetune.conversation#L1-L159","kind":"module","name":"gorilla.finetune.conversation","path":"gorilla/finetune/conversation.py","language":"python","start_line":1,"end_line":159,"context_start_line":1,"context_end_line":159,"code":"import dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Tuple\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Different separator style.\"\"\"\n    SINGLE = auto()\n    TWO = auto()\n\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n    system: str\n    roles: List[str]\n    messages: List[List[str]]\n    offset: int\n    sep_style: SeparatorStyle = SeparatorStyle.SINGLE\n    sep: str = \"###\"\n    sep2: str = None\n\n    skip_next: bool = False\n\n    def get_prompt(self):\n        if self.sep_style == SeparatorStyle.SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n\n    def dict(self):\n        return {\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n            \"sep\": self.sep,\n            \"sep2\": self.sep2,\n        }\n\n\nconv_v1 = Conversation(\n    system=\"A chat between a curious human and an artificial intelligence assistant. \"\n           \"The assistant gives helpful, detailed, and polite answers to the human's questions.\",\n    roles=(\"Human\", \"Assistant\"),\n    messages=(\n        (\"Human\", \"Give three tips for staying healthy.\"),\n        (\"Assistant\",\n            \"Sure, here are three tips for staying healthy:\\n\"\n            \"1. Exercise regularly: Regular physical activity can help improve your overall health and wellbeing. \"\n            \"It can also help reduce your risk of chronic conditions such as obesity, diabetes, heart disease, \"\n            \"and certain cancers. Aim for at least 150 minutes of moderate-intensity aerobic exercise or \"\n            \"75 minutes of vigorous-intensity aerobic exercise per week, along with muscle-strengthening \"\n            \"activities at least two days per week.\\n\"\n            \"2. Eat a balanced diet: Eating a balanced diet that is rich in fruits, \"\n            \"vegetables, whole grains, lean proteins, and healthy fats can help support \"\n            \"your overall health. Try to limit your intake of processed and high-sugar foods, \"\n            \"and aim to drink plenty of water throughout the day.\\n\"\n            \"3. Get enough sleep: Getting enough quality sleep is essential for your physical \"\n            \"and mental health. Adults should aim for seven to nine hours of sleep per night. \"\n            \"Establish a regular sleep schedule and try to create a relaxing bedtime routine to \"\n            \"help improve the quality of your sleep.\")\n    ),\n    offset=2,\n    sep_style=SeparatorStyle.SINGLE,\n    sep=\"###\",\n)\n\nconv_v1_2 = Conversation(\n    system=\"A chat between a curious human and an artificial intelligence assistant. \"\n           \"The assistant gives helpful, detailed, and polite answers to the human's questions.\",\n    roles=(\"Human\", \"Assistant\"),\n    messages=(\n        (\"Human\", \"What are the key differences between renewable and non-renewable energy sources?\"),\n        (\"Assistant\",\n            \"Renewable energy sources are those that can be replenished naturally in a relatively \"\n            \"short amount of time, such as solar, wind, hydro, geothermal, and biomass. \"\n            \"Non-renewable energy sources, on the other hand, are finite and will eventually be \"\n            \"depleted, such as coal, oil, and natural gas. Here are some key differences between \"\n            \"renewable and non-renewable energy sources:\\n\"\n            \"1. Availability: Renewable energy sources are virtually inexhaustible, while non-renewable \"\n            \"energy sources are finite and will eventually run out.\\n\"\n            \"2. Environmental impact: Renewable energy sources have a much lower environmental impact \"\n            \"than non-renewable sources, which can lead to air and water pollution, greenhouse gas emissions, \"\n            \"and other negative effects.\\n\"\n            \"3. Cost: Renewable energy sources can be more expensive to initially set up, but they typically \"\n            \"have lower operational costs than non-renewable sources.\\n\"\n            \"4. Reliability: Renewable energy sources are often more reliable and can be used in more remote \"\n            \"locations than non-renewable sources.\\n\"\n            \"5. Flexibility: Renewable energy sources are often more flexible and can be adapted to different \"\n            \"situations and needs, while non-renewable sources are more rigid and inflexible.\\n\"\n            \"6. Sustainability: Renewable energy sources are more sustainable over the long term, while \"\n            \"non-renewable sources are not, and their depletion can lead to economic and social instability.\\n\")\n    ),\n    offset=2,\n    sep_style=SeparatorStyle.SINGLE,\n    sep=\"###\",\n)\n\nconv_bair_v1 = Conversation(\n    system=\"BEGINNING OF CONVERSATION:\",\n    roles=(\"USER\", \"GPT\"),\n    messages=(),\n    offset=0,\n    sep_style=SeparatorStyle.TWO,\n    sep=\" \",\n    sep2=\"</s>\",\n)\n\n\ndefault_conversation = conv_v1_2\nconv_templates = {\n    \"v1\": conv_v1_2,\n    \"bair_v1\": conv_bair_v1,\n}\n\n\nif __name__ == \"__main__\":\n    import pdb\n    pdb.set_trace()\n    default_conversation.get_prompt()\n    print(default_conversation.get_prompt())","source_hash":"26bc22f298f041b21f9c0877b66425f314dd0c1e0222e349779be6e1de949cf6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.conversation.SeparatorStyle","uri":"program://LLaMA-Adapter/class/gorilla.finetune.conversation.SeparatorStyle#L6-L9","kind":"class","name":"SeparatorStyle","path":"gorilla/finetune/conversation.py","language":"python","start_line":6,"end_line":9,"context_start_line":1,"context_end_line":29,"code":"import dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Tuple\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Different separator style.\"\"\"\n    SINGLE = auto()\n    TWO = auto()\n\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n    system: str\n    roles: List[str]\n    messages: List[List[str]]\n    offset: int\n    sep_style: SeparatorStyle = SeparatorStyle.SINGLE\n    sep: str = \"###\"\n    sep2: str = None\n\n    skip_next: bool = False\n\n    def get_prompt(self):\n        if self.sep_style == SeparatorStyle.SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:","source_hash":"26bc22f298f041b21f9c0877b66425f314dd0c1e0222e349779be6e1de949cf6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.conversation.Conversation","uri":"program://LLaMA-Adapter/class/gorilla.finetune.conversation.Conversation#L13-L76","kind":"class","name":"Conversation","path":"gorilla/finetune/conversation.py","language":"python","start_line":13,"end_line":76,"context_start_line":1,"context_end_line":96,"code":"import dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Tuple\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Different separator style.\"\"\"\n    SINGLE = auto()\n    TWO = auto()\n\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n    system: str\n    roles: List[str]\n    messages: List[List[str]]\n    offset: int\n    sep_style: SeparatorStyle = SeparatorStyle.SINGLE\n    sep: str = \"###\"\n    sep2: str = None\n\n    skip_next: bool = False\n\n    def get_prompt(self):\n        if self.sep_style == SeparatorStyle.SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n\n    def dict(self):\n        return {\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n            \"sep\": self.sep,\n            \"sep2\": self.sep2,\n        }\n\n\nconv_v1 = Conversation(\n    system=\"A chat between a curious human and an artificial intelligence assistant. \"\n           \"The assistant gives helpful, detailed, and polite answers to the human's questions.\",\n    roles=(\"Human\", \"Assistant\"),\n    messages=(\n        (\"Human\", \"Give three tips for staying healthy.\"),\n        (\"Assistant\",\n            \"Sure, here are three tips for staying healthy:\\n\"\n            \"1. Exercise regularly: Regular physical activity can help improve your overall health and wellbeing. \"\n            \"It can also help reduce your risk of chronic conditions such as obesity, diabetes, heart disease, \"\n            \"and certain cancers. Aim for at least 150 minutes of moderate-intensity aerobic exercise or \"\n            \"75 minutes of vigorous-intensity aerobic exercise per week, along with muscle-strengthening \"\n            \"activities at least two days per week.\\n\"\n            \"2. Eat a balanced diet: Eating a balanced diet that is rich in fruits, \"\n            \"vegetables, whole grains, lean proteins, and healthy fats can help support \"\n            \"your overall health. Try to limit your intake of processed and high-sugar foods, \"\n            \"and aim to drink plenty of water throughout the day.\\n\"\n            \"3. Get enough sleep: Getting enough quality sleep is essential for your physical \"","source_hash":"26bc22f298f041b21f9c0877b66425f314dd0c1e0222e349779be6e1de949cf6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.conversation.get_prompt","uri":"program://LLaMA-Adapter/function/gorilla.finetune.conversation.get_prompt#L25-L44","kind":"function","name":"get_prompt","path":"gorilla/finetune/conversation.py","language":"python","start_line":25,"end_line":44,"context_start_line":5,"context_end_line":64,"code":"\nclass SeparatorStyle(Enum):\n    \"\"\"Different separator style.\"\"\"\n    SINGLE = auto()\n    TWO = auto()\n\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n    system: str\n    roles: List[str]\n    messages: List[List[str]]\n    offset: int\n    sep_style: SeparatorStyle = SeparatorStyle.SINGLE\n    sep: str = \"###\"\n    sep2: str = None\n\n    skip_next: bool = False\n\n    def get_prompt(self):\n        if self.sep_style == SeparatorStyle.SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,","source_hash":"26bc22f298f041b21f9c0877b66425f314dd0c1e0222e349779be6e1de949cf6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.conversation.append_message","uri":"program://LLaMA-Adapter/function/gorilla.finetune.conversation.append_message#L46-L47","kind":"function","name":"append_message","path":"gorilla/finetune/conversation.py","language":"python","start_line":46,"end_line":47,"context_start_line":26,"context_end_line":67,"code":"        if self.sep_style == SeparatorStyle.SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n","source_hash":"26bc22f298f041b21f9c0877b66425f314dd0c1e0222e349779be6e1de949cf6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.conversation.to_gradio_chatbot","uri":"program://LLaMA-Adapter/function/gorilla.finetune.conversation.to_gradio_chatbot#L49-L56","kind":"function","name":"to_gradio_chatbot","path":"gorilla/finetune/conversation.py","language":"python","start_line":49,"end_line":56,"context_start_line":29,"context_end_line":76,"code":"                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n\n    def dict(self):\n        return {\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n            \"sep\": self.sep,\n            \"sep2\": self.sep2,\n        }","source_hash":"26bc22f298f041b21f9c0877b66425f314dd0c1e0222e349779be6e1de949cf6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.conversation.copy","uri":"program://LLaMA-Adapter/function/gorilla.finetune.conversation.copy#L58-L66","kind":"function","name":"copy","path":"gorilla/finetune/conversation.py","language":"python","start_line":58,"end_line":66,"context_start_line":38,"context_end_line":86,"code":"                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n\n    def dict(self):\n        return {\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n            \"sep\": self.sep,\n            \"sep2\": self.sep2,\n        }\n\n\nconv_v1 = Conversation(\n    system=\"A chat between a curious human and an artificial intelligence assistant. \"\n           \"The assistant gives helpful, detailed, and polite answers to the human's questions.\",\n    roles=(\"Human\", \"Assistant\"),\n    messages=(\n        (\"Human\", \"Give three tips for staying healthy.\"),\n        (\"Assistant\",\n            \"Sure, here are three tips for staying healthy:\\n\"","source_hash":"26bc22f298f041b21f9c0877b66425f314dd0c1e0222e349779be6e1de949cf6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.conversation.dict","uri":"program://LLaMA-Adapter/function/gorilla.finetune.conversation.dict#L68-L76","kind":"function","name":"dict","path":"gorilla/finetune/conversation.py","language":"python","start_line":68,"end_line":76,"context_start_line":48,"context_end_line":96,"code":"\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n\n    def dict(self):\n        return {\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n            \"sep\": self.sep,\n            \"sep2\": self.sep2,\n        }\n\n\nconv_v1 = Conversation(\n    system=\"A chat between a curious human and an artificial intelligence assistant. \"\n           \"The assistant gives helpful, detailed, and polite answers to the human's questions.\",\n    roles=(\"Human\", \"Assistant\"),\n    messages=(\n        (\"Human\", \"Give three tips for staying healthy.\"),\n        (\"Assistant\",\n            \"Sure, here are three tips for staying healthy:\\n\"\n            \"1. Exercise regularly: Regular physical activity can help improve your overall health and wellbeing. \"\n            \"It can also help reduce your risk of chronic conditions such as obesity, diabetes, heart disease, \"\n            \"and certain cancers. Aim for at least 150 minutes of moderate-intensity aerobic exercise or \"\n            \"75 minutes of vigorous-intensity aerobic exercise per week, along with muscle-strengthening \"\n            \"activities at least two days per week.\\n\"\n            \"2. Eat a balanced diet: Eating a balanced diet that is rich in fruits, \"\n            \"vegetables, whole grains, lean proteins, and healthy fats can help support \"\n            \"your overall health. Try to limit your intake of processed and high-sugar foods, \"\n            \"and aim to drink plenty of water throughout the day.\\n\"\n            \"3. Get enough sleep: Getting enough quality sleep is essential for your physical \"","source_hash":"26bc22f298f041b21f9c0877b66425f314dd0c1e0222e349779be6e1de949cf6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.global_configs","uri":"program://LLaMA-Adapter/module/gorilla.finetune.global_configs#L1-L4","kind":"module","name":"gorilla.finetune.global_configs","path":"gorilla/finetune/global_configs.py","language":"python","start_line":1,"end_line":4,"context_start_line":1,"context_end_line":4,"code":"tokenizer_path = '/data1/llma/tokenizer.model'\npetrel_conf = \"/mnt/petrelfs/share_data/gaopeng/ldy/petreloss_all.conf\"\npetrel_prefix = \"cluster_p_ssd:s3://falcon-refinedweb/data\"\ndata_meta_path = \"/mnt/petrelfs/share_data/gaopeng/ldy/falcon_list.json\"","source_hash":"ef7e1236d31fb389e2d529063d6ad7f7ba242fde9261f8943c12bd4b200d9e88","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.data_preprocess","uri":"program://LLaMA-Adapter/module/gorilla.finetune.data_preprocess#L1-L30","kind":"module","name":"gorilla.finetune.data_preprocess","path":"gorilla/finetune/data_preprocess.py","language":"python","start_line":1,"end_line":30,"context_start_line":1,"context_end_line":30,"code":"import json\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\nimport pdb\npdb.set_trace()\n\ndatas = json.load(open('/home/pgao/stanford_alpaca/stanford_alpaca/alpaca_data.json'))\nprompt_input, prompt_no_input = PROMPT_DICT[\"prompt_input\"], PROMPT_DICT[\"prompt_no_input\"]\nsources = [\n        prompt_input.format_map(example) if example.get(\"input\", \"\") != \"\" else prompt_no_input.format_map(example)\n        for example in datas\n        ]\n\n\ntargets = [f\"{example['output']}\" for example in datas]\nexamples = [s + t for s, t in zip(sources, targets)]\nfor strings in (examples, sources):\n    print(strings)\n","source_hash":"4456043a285aeef35fe923702414b326d7cce06887487ce4942834dfe370e279","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.main_pretrain","uri":"program://LLaMA-Adapter/module/gorilla.finetune.main_pretrain#L1-L279","kind":"module","name":"gorilla.finetune.main_pretrain","path":"gorilla/finetune/main_pretrain.py","language":"python","start_line":1,"end_line":279,"context_start_line":1,"context_end_line":279,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\nimport functools\nfrom functools import partial\n\nimport torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.distributed.fsdp import (\n    FullyShardedDataParallel as FSDP,\n    MixedPrecision,\n    ShardingStrategy,\n)\nfrom torch.distributed.algorithms._checkpoint.checkpoint_wrapper import (\n   checkpoint_wrapper,\n   CheckpointImpl,\n   apply_activation_checkpointing,\n)\nfrom torch.distributed.fsdp.wrap import (\n    transformer_auto_wrap_policy,\n)\n\nfrom fairscale.nn.model_parallel import initialize as fs_init\n\nfrom apex.optimizers import FusedAdam\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom model.meta import MetaModel\nfrom model.LLM.llama import Attention, FeedForward\nfrom engine_pretrain import train_one_epoch, val_one_epoch\nfrom torch.utils.data import Dataset\nfrom data.data import Falcon, FalconVal\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('MAE pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=64, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--accum_iter', default=4, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='llama', type=str, metavar='MODEL', choices=['llama', 'revllama'],\n                        help='Name of model to train')\n\n    parser.add_argument('--llama_config', default='params.json', type=str,\n                        help='Path to llama model config')\n\n    parser.add_argument('--reversible_grad', action='store_true', default=False,\n                        help='Whether to use reversible grad')\n\n    parser.add_argument('--input_size', default=224, type=int,\n                        help='images input size')\n\n    parser.add_argument('--mask_ratio', default=0.75, type=float,\n                        help='Masking ratio (percentage of removed patches).')\n\n    parser.add_argument('--norm_pix_loss', action='store_true',\n                        help='Use (per-patch) normalized pixels as targets for computing loss')\n    parser.set_defaults(norm_pix_loss=False)\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.02,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=0.001, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--min_lr', type=float, default=0.0001, metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_iters', type=int, default=20000, metavar='N',\n                        help='iterations to warmup LR')\n    parser.add_argument('--lr_decay_iters', type=int, default=1800000, metavar='N',\n                        help='iters before keeping minimal learning rate')\n\n    parser.add_argument('--clip_grad', type=int, default=-1,\n                        help='grad clipping norm')\n\n    # Dataset parameters\n    parser.add_argument('--data_path', default='/datasets01/imagenet_full_size/061417/', type=str,\n                        help='dataset path')\n\n    parser.add_argument('--output_dir', default='./output_dir',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output_dir',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n    parser.add_argument('--resume', default='',\n                        help='resume from checkpoint')\n\n\n    parser.add_argument('--num_workers', default=5, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--model_parallel_size', type=int, default=1)\n    parser.add_argument('--data_parallel', type=str, choices=['ddp', 'sdp', 'fsdp'], default='sdp')\n    parser.add_argument('--precision', type=str, choices=['fp16', 'bf16', 'tf32'], default='bf16')\n    parser.add_argument('--save_freq', type=int, default=5000)\n    parser.add_argument('--save_consolidated', action=\"store_true\",\n                        help=\"save consolidated model weights along with regular checkpoints \"\n                             \"used to resume training. useful for convenient deployment but \"\n                             \"will occupy some additional disk space.\")\n    parser.add_argument('--checkpointing', action=\"store_true\", default=False,\n                        help=\"enable gradient checkopointing\")\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n    fs_init.initialize_model_parallel(args.model_parallel_size)\n    if args.precision == \"tf32\":\n        torch.backends.cuda.matmul.allow_tf32 = True\n        torch.backends.cudnn.allow_tf32 = True\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n\n    cudnn.benchmark = True\n\n    global_rank = misc.get_rank()\n    mp_rank = fs_init.get_model_parallel_rank()\n    dp_rank = fs_init.get_data_parallel_rank()\n    dp_world_size = fs_init.get_data_parallel_world_size()\n    dp_group = fs_init.get_data_parallel_group()\n\n    dataset_train = Falcon(\n        max_words=2048, num_processes=dp_world_size, process_rank=dp_rank,\n    )\n    dataset_val = FalconVal(max_words=2048)\n    print(dataset_train)\n\n\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    sampler_val = torch.utils.data.DistributedSampler(\n        dataset_val, num_replicas=dp_world_size, rank=dp_rank, shuffle=False\n    )\n    data_loader_val = torch.utils.data.DataLoader(\n        dataset_val,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        sampler=sampler_val,\n        drop_last=True,\n    )\n\n    \n    # define the model\n    model = MetaModel(args.llama_type, args.reversible_grad, args.llama_config)\n    model.to(device)\n    print(\"Unwrapped Model = %s\" % str(model))\n\n    mixed_precision_dtype = {\n        \"fp16\": torch.float16,\n        \"bf16\": torch.bfloat16,\n        \"tf32\": torch.float32,\n    }[args.precision]\n    model = FSDP(\n        model,\n        process_group=fs_init.get_data_parallel_group(),\n        auto_wrap_policy=functools.partial(\n            transformer_auto_wrap_policy,\n            transformer_layer_cls=[Attention, FeedForward],\n        ),\n        limit_all_gathers=True,\n        use_orig_params=True,\n        sync_module_states=True,\n        mixed_precision=MixedPrecision(\n            param_dtype=mixed_precision_dtype,\n            reduce_dtype=mixed_precision_dtype,\n            buffer_dtype=mixed_precision_dtype,\n        ),\n        sharding_strategy={\n            \"sdp\": ShardingStrategy.SHARD_GRAD_OP,\n            \"ddp\": ShardingStrategy.NO_SHARD,\n            \"fsdp\": ShardingStrategy.FULL_SHARD,\n        }[args.data_parallel],\n    )\n\n    # gradient checkpointing\n    if args.checkpointing:\n        print(\"apply gradient checkpointing\")\n        non_reentrant_wrapper = partial(\n            checkpoint_wrapper,\n            offload_to_cpu=False,\n            checkpoint_impl=CheckpointImpl.NO_REENTRANT,\n        )\n        check_fn = lambda submodule: isinstance(submodule, (Attention, FeedForward))\n        apply_activation_checkpointing(model, checkpoint_wrapper_fn=non_reentrant_wrapper, check_fn=check_fn)\n\n    print(\"Model = %s\" % str(model))\n\n    eff_batch_size = args.batch_size * args.accum_iter * fs_init.get_data_parallel_world_size()\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model, args.weight_decay)\n    optimizer = FusedAdam(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler(args)\n\n    start_iter = 0\n    if args.resume:\n        _, start_iter = misc.load_model(args=args, model=model, optimizer=optimizer, loss_scaler=loss_scaler, dataset_train=dataset_train)\n\n    print(f\"Start training\")\n    start_time = time.time()\n\n    train_stats = train_one_epoch(\n        model, data_loader_train, data_loader_val,\n        optimizer, device, 0, start_iter, loss_scaler,\n        log_writer=log_writer,\n        args=args\n    )\n\n    if args.output_dir and misc.is_main_process():\n        if log_writer is not None:\n            log_writer.flush()\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"efb38fa3fda169a2a9cbde46b53e5cb3ad6bbf5b2f4d2c4aa8dc86da164f5ee6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.main_pretrain.get_args_parser","uri":"program://LLaMA-Adapter/function/gorilla.finetune.main_pretrain.get_args_parser#L51-L135","kind":"function","name":"get_args_parser","path":"gorilla/finetune/main_pretrain.py","language":"python","start_line":51,"end_line":135,"context_start_line":31,"context_end_line":155,"code":"   CheckpointImpl,\n   apply_activation_checkpointing,\n)\nfrom torch.distributed.fsdp.wrap import (\n    transformer_auto_wrap_policy,\n)\n\nfrom fairscale.nn.model_parallel import initialize as fs_init\n\nfrom apex.optimizers import FusedAdam\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom model.meta import MetaModel\nfrom model.LLM.llama import Attention, FeedForward\nfrom engine_pretrain import train_one_epoch, val_one_epoch\nfrom torch.utils.data import Dataset\nfrom data.data import Falcon, FalconVal\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('MAE pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=64, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--accum_iter', default=4, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='llama', type=str, metavar='MODEL', choices=['llama', 'revllama'],\n                        help='Name of model to train')\n\n    parser.add_argument('--llama_config', default='params.json', type=str,\n                        help='Path to llama model config')\n\n    parser.add_argument('--reversible_grad', action='store_true', default=False,\n                        help='Whether to use reversible grad')\n\n    parser.add_argument('--input_size', default=224, type=int,\n                        help='images input size')\n\n    parser.add_argument('--mask_ratio', default=0.75, type=float,\n                        help='Masking ratio (percentage of removed patches).')\n\n    parser.add_argument('--norm_pix_loss', action='store_true',\n                        help='Use (per-patch) normalized pixels as targets for computing loss')\n    parser.set_defaults(norm_pix_loss=False)\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.02,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=0.001, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--min_lr', type=float, default=0.0001, metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_iters', type=int, default=20000, metavar='N',\n                        help='iterations to warmup LR')\n    parser.add_argument('--lr_decay_iters', type=int, default=1800000, metavar='N',\n                        help='iters before keeping minimal learning rate')\n\n    parser.add_argument('--clip_grad', type=int, default=-1,\n                        help='grad clipping norm')\n\n    # Dataset parameters\n    parser.add_argument('--data_path', default='/datasets01/imagenet_full_size/061417/', type=str,\n                        help='dataset path')\n\n    parser.add_argument('--output_dir', default='./output_dir',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output_dir',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n    parser.add_argument('--resume', default='',\n                        help='resume from checkpoint')\n\n\n    parser.add_argument('--num_workers', default=5, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--model_parallel_size', type=int, default=1)\n    parser.add_argument('--data_parallel', type=str, choices=['ddp', 'sdp', 'fsdp'], default='sdp')\n    parser.add_argument('--precision', type=str, choices=['fp16', 'bf16', 'tf32'], default='bf16')\n    parser.add_argument('--save_freq', type=int, default=5000)\n    parser.add_argument('--save_consolidated', action=\"store_true\",\n                        help=\"save consolidated model weights along with regular checkpoints \"\n                             \"used to resume training. useful for convenient deployment but \"\n                             \"will occupy some additional disk space.\")\n    parser.add_argument('--checkpointing', action=\"store_true\", default=False,\n                        help=\"enable gradient checkopointing\")\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n    fs_init.initialize_model_parallel(args.model_parallel_size)\n    if args.precision == \"tf32\":\n        torch.backends.cuda.matmul.allow_tf32 = True\n        torch.backends.cudnn.allow_tf32 = True\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n\n    cudnn.benchmark = True","source_hash":"efb38fa3fda169a2a9cbde46b53e5cb3ad6bbf5b2f4d2c4aa8dc86da164f5ee6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.main_pretrain.main","uri":"program://LLaMA-Adapter/function/gorilla.finetune.main_pretrain.main#L138-L271","kind":"function","name":"main","path":"gorilla/finetune/main_pretrain.py","language":"python","start_line":138,"end_line":271,"context_start_line":118,"context_end_line":279,"code":"                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--model_parallel_size', type=int, default=1)\n    parser.add_argument('--data_parallel', type=str, choices=['ddp', 'sdp', 'fsdp'], default='sdp')\n    parser.add_argument('--precision', type=str, choices=['fp16', 'bf16', 'tf32'], default='bf16')\n    parser.add_argument('--save_freq', type=int, default=5000)\n    parser.add_argument('--save_consolidated', action=\"store_true\",\n                        help=\"save consolidated model weights along with regular checkpoints \"\n                             \"used to resume training. useful for convenient deployment but \"\n                             \"will occupy some additional disk space.\")\n    parser.add_argument('--checkpointing', action=\"store_true\", default=False,\n                        help=\"enable gradient checkopointing\")\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n    fs_init.initialize_model_parallel(args.model_parallel_size)\n    if args.precision == \"tf32\":\n        torch.backends.cuda.matmul.allow_tf32 = True\n        torch.backends.cudnn.allow_tf32 = True\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n\n    cudnn.benchmark = True\n\n    global_rank = misc.get_rank()\n    mp_rank = fs_init.get_model_parallel_rank()\n    dp_rank = fs_init.get_data_parallel_rank()\n    dp_world_size = fs_init.get_data_parallel_world_size()\n    dp_group = fs_init.get_data_parallel_group()\n\n    dataset_train = Falcon(\n        max_words=2048, num_processes=dp_world_size, process_rank=dp_rank,\n    )\n    dataset_val = FalconVal(max_words=2048)\n    print(dataset_train)\n\n\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    sampler_val = torch.utils.data.DistributedSampler(\n        dataset_val, num_replicas=dp_world_size, rank=dp_rank, shuffle=False\n    )\n    data_loader_val = torch.utils.data.DataLoader(\n        dataset_val,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        sampler=sampler_val,\n        drop_last=True,\n    )\n\n    \n    # define the model\n    model = MetaModel(args.llama_type, args.reversible_grad, args.llama_config)\n    model.to(device)\n    print(\"Unwrapped Model = %s\" % str(model))\n\n    mixed_precision_dtype = {\n        \"fp16\": torch.float16,\n        \"bf16\": torch.bfloat16,\n        \"tf32\": torch.float32,\n    }[args.precision]\n    model = FSDP(\n        model,\n        process_group=fs_init.get_data_parallel_group(),\n        auto_wrap_policy=functools.partial(\n            transformer_auto_wrap_policy,\n            transformer_layer_cls=[Attention, FeedForward],\n        ),\n        limit_all_gathers=True,\n        use_orig_params=True,\n        sync_module_states=True,\n        mixed_precision=MixedPrecision(\n            param_dtype=mixed_precision_dtype,\n            reduce_dtype=mixed_precision_dtype,\n            buffer_dtype=mixed_precision_dtype,\n        ),\n        sharding_strategy={\n            \"sdp\": ShardingStrategy.SHARD_GRAD_OP,\n            \"ddp\": ShardingStrategy.NO_SHARD,\n            \"fsdp\": ShardingStrategy.FULL_SHARD,\n        }[args.data_parallel],\n    )\n\n    # gradient checkpointing\n    if args.checkpointing:\n        print(\"apply gradient checkpointing\")\n        non_reentrant_wrapper = partial(\n            checkpoint_wrapper,\n            offload_to_cpu=False,\n            checkpoint_impl=CheckpointImpl.NO_REENTRANT,\n        )\n        check_fn = lambda submodule: isinstance(submodule, (Attention, FeedForward))\n        apply_activation_checkpointing(model, checkpoint_wrapper_fn=non_reentrant_wrapper, check_fn=check_fn)\n\n    print(\"Model = %s\" % str(model))\n\n    eff_batch_size = args.batch_size * args.accum_iter * fs_init.get_data_parallel_world_size()\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model, args.weight_decay)\n    optimizer = FusedAdam(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler(args)\n\n    start_iter = 0\n    if args.resume:\n        _, start_iter = misc.load_model(args=args, model=model, optimizer=optimizer, loss_scaler=loss_scaler, dataset_train=dataset_train)\n\n    print(f\"Start training\")\n    start_time = time.time()\n\n    train_stats = train_one_epoch(\n        model, data_loader_train, data_loader_val,\n        optimizer, device, 0, start_iter, loss_scaler,\n        log_writer=log_writer,\n        args=args\n    )\n\n    if args.output_dir and misc.is_main_process():\n        if log_writer is not None:\n            log_writer.flush()\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"efb38fa3fda169a2a9cbde46b53e5cb3ad6bbf5b2f4d2c4aa8dc86da164f5ee6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.engine_finetune","uri":"program://LLaMA-Adapter/module/gorilla.finetune.engine_finetune#L1-L151","kind":"module","name":"gorilla.finetune.engine_finetune","path":"gorilla/finetune/engine_finetune.py","language":"python","start_line":1,"end_line":151,"context_start_line":1,"context_end_line":151,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\nimport json\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport math\nimport sys\nimport os\nfrom typing import Iterable\nimport contextlib\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\nimport pdb\n\ndef train_one_epoch(model: torch.nn.Module,\n                    data_loader, optimizer: torch.optim.Optimizer,\n                    epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, imgs) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header)):\n\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate_epoch(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        autocast_ctx = {\n            \"bf16\": torch.cuda.amp.autocast(dtype=torch.bfloat16),\n            \"fp16\": torch.cuda.amp.autocast(dtype=torch.float16),\n            \"tf32\": contextlib.nullcontext(),\n        }[args.precision]\n        with autocast_ctx:\n             c_loss, m_loss, _, _ = model(examples, labels)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n        update_grad = (data_iter_step + 1) % accum_iter == 0\n        grad_norm = loss_scaler(\n            loss, optimizer, model,\n            parameters=model.parameters(),\n            update_grad=update_grad,\n            clip_grad=None if args.clip_grad <= 0 else args.clip_grad,\n        )\n        if update_grad:\n            assert grad_norm is not None\n            metric_logger.update(grad_norm=grad_norm)\n\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if update_grad:\n            grad_norm_reduce = misc.all_reduce_mean(grad_norm)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            if update_grad:\n                log_writer.add_scalar('grad_norm', grad_norm_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n@ torch.no_grad()\ndef val_one_epoch(model: torch.nn.Module,\n                  data_loader: Iterable, epoch: int,\n                  log_writer=None,\n                  args=None):\n    print(\"!!!start validation!!!\")\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n\n        with torch.cuda.amp.autocast(dtype=torch.bfloat16):\n             c_loss, m_loss, _, _ = model(examples, labels)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss.item()\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n\n        # loss_value_reduce = misc.all_reduce_mean(loss_value)\n        # c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        # m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        # if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n        #     \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n        #     This calibrates different curves when batch size changes.\n        #     \"\"\"\n        #     epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n        #     log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n        #     log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n        #     log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n","source_hash":"04485c8d845a90473030686c62d742ae5b1837a84fdc72b8d89c5a51516f94e7","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.engine_finetune.train_one_epoch","uri":"program://LLaMA-Adapter/function/gorilla.finetune.engine_finetune.train_one_epoch#L23-L103","kind":"function","name":"train_one_epoch","path":"gorilla/finetune/engine_finetune.py","language":"python","start_line":23,"end_line":103,"context_start_line":3,"context_end_line":123,"code":"import json\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport math\nimport sys\nimport os\nfrom typing import Iterable\nimport contextlib\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\nimport pdb\n\ndef train_one_epoch(model: torch.nn.Module,\n                    data_loader, optimizer: torch.optim.Optimizer,\n                    epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, imgs) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header)):\n\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate_epoch(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        autocast_ctx = {\n            \"bf16\": torch.cuda.amp.autocast(dtype=torch.bfloat16),\n            \"fp16\": torch.cuda.amp.autocast(dtype=torch.float16),\n            \"tf32\": contextlib.nullcontext(),\n        }[args.precision]\n        with autocast_ctx:\n             c_loss, m_loss, _, _ = model(examples, labels)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n        update_grad = (data_iter_step + 1) % accum_iter == 0\n        grad_norm = loss_scaler(\n            loss, optimizer, model,\n            parameters=model.parameters(),\n            update_grad=update_grad,\n            clip_grad=None if args.clip_grad <= 0 else args.clip_grad,\n        )\n        if update_grad:\n            assert grad_norm is not None\n            metric_logger.update(grad_norm=grad_norm)\n\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if update_grad:\n            grad_norm_reduce = misc.all_reduce_mean(grad_norm)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            if update_grad:\n                log_writer.add_scalar('grad_norm', grad_norm_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n@ torch.no_grad()\ndef val_one_epoch(model: torch.nn.Module,\n                  data_loader: Iterable, epoch: int,\n                  log_writer=None,\n                  args=None):\n    print(\"!!!start validation!!!\")\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n\n        with torch.cuda.amp.autocast(dtype=torch.bfloat16):\n             c_loss, m_loss, _, _ = model(examples, labels)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()","source_hash":"04485c8d845a90473030686c62d742ae5b1837a84fdc72b8d89c5a51516f94e7","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.engine_finetune.val_one_epoch","uri":"program://LLaMA-Adapter/function/gorilla.finetune.engine_finetune.val_one_epoch#L106-L150","kind":"function","name":"val_one_epoch","path":"gorilla/finetune/engine_finetune.py","language":"python","start_line":106,"end_line":150,"context_start_line":86,"context_end_line":151,"code":"        if update_grad:\n            grad_norm_reduce = misc.all_reduce_mean(grad_norm)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            if update_grad:\n                log_writer.add_scalar('grad_norm', grad_norm_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n@ torch.no_grad()\ndef val_one_epoch(model: torch.nn.Module,\n                  data_loader: Iterable, epoch: int,\n                  log_writer=None,\n                  args=None):\n    print(\"!!!start validation!!!\")\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n\n        with torch.cuda.amp.autocast(dtype=torch.bfloat16):\n             c_loss, m_loss, _, _ = model(examples, labels)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss.item()\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n\n        # loss_value_reduce = misc.all_reduce_mean(loss_value)\n        # c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        # m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        # if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n        #     \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n        #     This calibrates different curves when batch size changes.\n        #     \"\"\"\n        #     epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n        #     log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n        #     log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n        #     log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n","source_hash":"04485c8d845a90473030686c62d742ae5b1837a84fdc72b8d89c5a51516f94e7","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.engine_pretrain","uri":"program://LLaMA-Adapter/module/gorilla.finetune.engine_pretrain#L1-L172","kind":"module","name":"gorilla.finetune.engine_pretrain","path":"gorilla/finetune/engine_pretrain.py","language":"python","start_line":1,"end_line":172,"context_start_line":1,"context_end_line":172,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\nimport json\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport math\nimport sys\nimport os\nfrom typing import Iterable\nimport contextlib\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\nimport pdb\n\ndef train_one_epoch(model: torch.nn.Module,\n                    data_loader, val_loader, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, start_iter, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    dataset_state = {}\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, item_states) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header, start_iter), start=start_iter\n    ):\n\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step, args)\n\n        autocast_ctx = {\n            \"bf16\": torch.cuda.amp.autocast(dtype=torch.bfloat16),\n            \"fp16\": torch.cuda.amp.autocast(dtype=torch.float16),\n            \"tf32\": contextlib.nullcontext(),\n        }[args.precision]\n        with autocast_ctx:\n             c_loss, m_loss, _, _ = model(examples, labels)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n#        loss.backward()\n\n        update_grad = (data_iter_step + 1) % accum_iter == 0\n        grad_norm = loss_scaler(\n            loss, optimizer, model,\n            parameters=model.parameters(),\n            update_grad=update_grad,\n            clip_grad=None if args.clip_grad <= 0 else args.clip_grad,\n        )\n        if update_grad:\n            assert grad_norm is not None\n            metric_logger.update(grad_norm=grad_norm)\n\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        # process item states for resume\n        for i in range(len(item_states['worker_id'])):\n            worker_id, _curr_idx, _file_idx = item_states['worker_id'][i], item_states['_curr_idx'][i], item_states['_file_idx'][i]\n            worker_id, _curr_idx, _file_idx = worker_id.item(), _curr_idx.item(), _file_idx.item()\n            if worker_id not in dataset_state or \\\n            dataset_state[worker_id]['_file_idx'] < _file_idx or \\\n            (dataset_state[worker_id]['_file_idx'] == _file_idx and dataset_state[worker_id]['_curr_idx'] < _curr_idx):\n                dataset_state[worker_id] = {\"_curr_idx\": _curr_idx, \"_file_idx\":  _file_idx}\n\n        # save checkpoint\n        if (data_iter_step + 1) % args.save_freq == 0:\n            misc.save_model(\n                output_dir=args.output_dir,\n                args=args, epoch=epoch, iteration=data_iter_step, model=model, optimizer=optimizer,\n                loss_scaler=loss_scaler, dataset_state=dataset_state)\n\n        # validation\n        if (data_iter_step + 1) % 10000 == 0:\n            val_one_epoch(model, val_loader, epoch, log_writer=log_writer, args=args)\n            model.train(True)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if update_grad:\n            grad_norm_reduce = misc.all_reduce_mean(grad_norm)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, data_iter_step)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, data_iter_step)\n            if update_grad:\n                log_writer.add_scalar('grad_norm', grad_norm_reduce, data_iter_step)\n            log_writer.add_scalar('lr', lr, data_iter_step)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n@ torch.no_grad()\ndef val_one_epoch(model: torch.nn.Module,\n                  data_loader: Iterable, epoch: int,\n                  log_writer=None,\n                  args=None):\n    print(\"!!!start validation!!!\")\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n\n        with torch.cuda.amp.autocast(dtype=torch.bfloat16):\n             c_loss, m_loss, _, _ = model(examples, labels)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss.item()\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n\n        # loss_value_reduce = misc.all_reduce_mean(loss_value)\n        # c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        # m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        # if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n        #     \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n        #     This calibrates different curves when batch size changes.\n        #     \"\"\"\n        #     epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n        #     log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n        #     log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n        #     log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n","source_hash":"04b42f9156ec322fcfc9789802a9bb6c60279c84ac1fa0ae1f941a8349a4260b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.engine_pretrain.train_one_epoch","uri":"program://LLaMA-Adapter/function/gorilla.finetune.engine_pretrain.train_one_epoch#L23-L124","kind":"function","name":"train_one_epoch","path":"gorilla/finetune/engine_pretrain.py","language":"python","start_line":23,"end_line":124,"context_start_line":3,"context_end_line":144,"code":"import json\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport math\nimport sys\nimport os\nfrom typing import Iterable\nimport contextlib\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\nimport pdb\n\ndef train_one_epoch(model: torch.nn.Module,\n                    data_loader, val_loader, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, start_iter, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    dataset_state = {}\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, item_states) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header, start_iter), start=start_iter\n    ):\n\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step, args)\n\n        autocast_ctx = {\n            \"bf16\": torch.cuda.amp.autocast(dtype=torch.bfloat16),\n            \"fp16\": torch.cuda.amp.autocast(dtype=torch.float16),\n            \"tf32\": contextlib.nullcontext(),\n        }[args.precision]\n        with autocast_ctx:\n             c_loss, m_loss, _, _ = model(examples, labels)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n#        loss.backward()\n\n        update_grad = (data_iter_step + 1) % accum_iter == 0\n        grad_norm = loss_scaler(\n            loss, optimizer, model,\n            parameters=model.parameters(),\n            update_grad=update_grad,\n            clip_grad=None if args.clip_grad <= 0 else args.clip_grad,\n        )\n        if update_grad:\n            assert grad_norm is not None\n            metric_logger.update(grad_norm=grad_norm)\n\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        # process item states for resume\n        for i in range(len(item_states['worker_id'])):\n            worker_id, _curr_idx, _file_idx = item_states['worker_id'][i], item_states['_curr_idx'][i], item_states['_file_idx'][i]\n            worker_id, _curr_idx, _file_idx = worker_id.item(), _curr_idx.item(), _file_idx.item()\n            if worker_id not in dataset_state or \\\n            dataset_state[worker_id]['_file_idx'] < _file_idx or \\\n            (dataset_state[worker_id]['_file_idx'] == _file_idx and dataset_state[worker_id]['_curr_idx'] < _curr_idx):\n                dataset_state[worker_id] = {\"_curr_idx\": _curr_idx, \"_file_idx\":  _file_idx}\n\n        # save checkpoint\n        if (data_iter_step + 1) % args.save_freq == 0:\n            misc.save_model(\n                output_dir=args.output_dir,\n                args=args, epoch=epoch, iteration=data_iter_step, model=model, optimizer=optimizer,\n                loss_scaler=loss_scaler, dataset_state=dataset_state)\n\n        # validation\n        if (data_iter_step + 1) % 10000 == 0:\n            val_one_epoch(model, val_loader, epoch, log_writer=log_writer, args=args)\n            model.train(True)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if update_grad:\n            grad_norm_reduce = misc.all_reduce_mean(grad_norm)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, data_iter_step)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, data_iter_step)\n            if update_grad:\n                log_writer.add_scalar('grad_norm', grad_norm_reduce, data_iter_step)\n            log_writer.add_scalar('lr', lr, data_iter_step)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n@ torch.no_grad()\ndef val_one_epoch(model: torch.nn.Module,\n                  data_loader: Iterable, epoch: int,\n                  log_writer=None,\n                  args=None):\n    print(\"!!!start validation!!!\")\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n\n        with torch.cuda.amp.autocast(dtype=torch.bfloat16):\n             c_loss, m_loss, _, _ = model(examples, labels)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()","source_hash":"04b42f9156ec322fcfc9789802a9bb6c60279c84ac1fa0ae1f941a8349a4260b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.engine_pretrain.val_one_epoch","uri":"program://LLaMA-Adapter/function/gorilla.finetune.engine_pretrain.val_one_epoch#L127-L171","kind":"function","name":"val_one_epoch","path":"gorilla/finetune/engine_pretrain.py","language":"python","start_line":127,"end_line":171,"context_start_line":107,"context_end_line":172,"code":"            model.train(True)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if update_grad:\n            grad_norm_reduce = misc.all_reduce_mean(grad_norm)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, data_iter_step)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, data_iter_step)\n            if update_grad:\n                log_writer.add_scalar('grad_norm', grad_norm_reduce, data_iter_step)\n            log_writer.add_scalar('lr', lr, data_iter_step)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n@ torch.no_grad()\ndef val_one_epoch(model: torch.nn.Module,\n                  data_loader: Iterable, epoch: int,\n                  log_writer=None,\n                  args=None):\n    print(\"!!!start validation!!!\")\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n\n        with torch.cuda.amp.autocast(dtype=torch.bfloat16):\n             c_loss, m_loss, _, _ = model(examples, labels)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss.item()\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n\n        # loss_value_reduce = misc.all_reduce_mean(loss_value)\n        # c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        # m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        # if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n        #     \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n        #     This calibrates different curves when batch size changes.\n        #     \"\"\"\n        #     epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n        #     log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n        #     log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n        #     log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n","source_hash":"04b42f9156ec322fcfc9789802a9bb6c60279c84ac1fa0ae1f941a8349a4260b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.submitit_pretrain","uri":"program://LLaMA-Adapter/module/gorilla.finetune.submitit_pretrain#L1-L131","kind":"module","name":"gorilla.finetune.submitit_pretrain","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":1,"end_line":131,"context_start_line":1,"context_end_line":131,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# A script to run multinode training with submitit.\n# --------------------------------------------------------\n\nimport argparse\nimport os\nimport uuid\nfrom pathlib import Path\n\nimport main_pretrain as trainer\nimport submitit\n\n\ndef parse_args():\n    trainer_parser = trainer.get_args_parser()\n    parser = argparse.ArgumentParser(\"Submitit for MAE pretrain\", parents=[trainer_parser])\n    parser.add_argument(\"--ngpus\", default=8, type=int, help=\"Number of gpus to request on each node\")\n    parser.add_argument(\"--nodes\", default=2, type=int, help=\"Number of nodes to request\")\n    parser.add_argument(\"--timeout\", default=4320, type=int, help=\"Duration of the job\")\n    parser.add_argument(\"--job_dir\", default=\"\", type=str, help=\"Job dir. Leave empty for automatic.\")\n\n    parser.add_argument(\"--partition\", default=\"learnfair\", type=str, help=\"Partition where to submit\")\n    parser.add_argument(\"--use_volta32\", action='store_true', help=\"Request 32G V100 GPUs\")\n    parser.add_argument('--comment', default=\"\", type=str, help=\"Comment to pass to scheduler\")\n    return parser.parse_args()\n\n\ndef get_shared_folder() -> Path:\n    user = os.getenv(\"USER\")\n    if Path(\"/checkpoint/\").is_dir():\n        p = Path(f\"/checkpoint/{user}/experiments\")\n        p.mkdir(exist_ok=True)\n        return p\n    raise RuntimeError(\"No shared folder available\")\n\n\ndef get_init_file():\n    # Init file must not exist, but it's parent dir must exist.\n    os.makedirs(str(get_shared_folder()), exist_ok=True)\n    init_file = get_shared_folder() / f\"{uuid.uuid4().hex}_init\"\n    if init_file.exists():\n        os.remove(str(init_file))\n    return init_file\n\n\nclass Trainer(object):\n    def __init__(self, args):\n        self.args = args\n\n    def __call__(self):\n        import main_pretrain as trainer\n\n        self._setup_gpu_args()\n        trainer.main(self.args)\n\n    def checkpoint(self):\n        import os\n        import submitit\n\n        self.args.dist_url = get_init_file().as_uri()\n        checkpoint_file = os.path.join(self.args.output_dir, \"checkpoint.pth\")\n        if os.path.exists(checkpoint_file):\n            self.args.resume = checkpoint_file\n        print(\"Requeuing \", self.args)\n        empty_trainer = type(self)(self.args)\n        return submitit.helpers.DelayedSubmission(empty_trainer)\n\n    def _setup_gpu_args(self):\n        import submitit\n        from pathlib import Path\n\n        job_env = submitit.JobEnvironment()\n        self.args.output_dir = Path(str(self.args.output_dir).replace(\"%j\", str(job_env.job_id)))\n        self.args.log_dir = self.args.output_dir\n        self.args.gpu = job_env.local_rank\n        self.args.rank = job_env.global_rank\n        self.args.world_size = job_env.num_tasks\n        print(f\"Process group: {job_env.num_tasks} tasks, rank: {job_env.global_rank}\")\n\n\ndef main():\n    args = parse_args()\n    if args.job_dir == \"\":\n        args.job_dir = get_shared_folder() / \"%j\"\n\n    # Note that the folder will depend on the job_id, to easily track experiments\n    executor = submitit.AutoExecutor(folder=args.job_dir, slurm_max_num_timeout=30)\n\n    num_gpus_per_node = args.ngpus\n    nodes = args.nodes\n    timeout_min = args.timeout\n\n    partition = args.partition\n    kwargs = {}\n    if args.use_volta32:\n        kwargs['slurm_constraint'] = 'volta32gb'\n    if args.comment:\n        kwargs['slurm_comment'] = args.comment\n\n    executor.update_parameters(\n        mem_gb=40 * num_gpus_per_node,\n        gpus_per_node=num_gpus_per_node,\n        tasks_per_node=num_gpus_per_node,  # one task per GPU\n        cpus_per_task=10,\n        nodes=nodes,\n        timeout_min=timeout_min,  # max is 60 * 72\n        # Below are cluster dependent parameters\n        slurm_partition=partition,\n        slurm_signal_delay_s=120,\n        **kwargs\n    )\n\n    executor.update_parameters(name=\"mae\")\n\n    args.dist_url = get_init_file().as_uri()\n    args.output_dir = args.job_dir\n\n    trainer = Trainer(args)\n    job = executor.submit(trainer)\n\n    # print(\"Submitted job_id:\", job.job_id)\n    print(job.job_id)\n\n\nif __name__ == \"__main__\":\n    main()","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.submitit_pretrain.parse_args","uri":"program://LLaMA-Adapter/function/gorilla.finetune.submitit_pretrain.parse_args#L19-L30","kind":"function","name":"parse_args","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":19,"end_line":30,"context_start_line":1,"context_end_line":50,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# A script to run multinode training with submitit.\n# --------------------------------------------------------\n\nimport argparse\nimport os\nimport uuid\nfrom pathlib import Path\n\nimport main_pretrain as trainer\nimport submitit\n\n\ndef parse_args():\n    trainer_parser = trainer.get_args_parser()\n    parser = argparse.ArgumentParser(\"Submitit for MAE pretrain\", parents=[trainer_parser])\n    parser.add_argument(\"--ngpus\", default=8, type=int, help=\"Number of gpus to request on each node\")\n    parser.add_argument(\"--nodes\", default=2, type=int, help=\"Number of nodes to request\")\n    parser.add_argument(\"--timeout\", default=4320, type=int, help=\"Duration of the job\")\n    parser.add_argument(\"--job_dir\", default=\"\", type=str, help=\"Job dir. Leave empty for automatic.\")\n\n    parser.add_argument(\"--partition\", default=\"learnfair\", type=str, help=\"Partition where to submit\")\n    parser.add_argument(\"--use_volta32\", action='store_true', help=\"Request 32G V100 GPUs\")\n    parser.add_argument('--comment', default=\"\", type=str, help=\"Comment to pass to scheduler\")\n    return parser.parse_args()\n\n\ndef get_shared_folder() -> Path:\n    user = os.getenv(\"USER\")\n    if Path(\"/checkpoint/\").is_dir():\n        p = Path(f\"/checkpoint/{user}/experiments\")\n        p.mkdir(exist_ok=True)\n        return p\n    raise RuntimeError(\"No shared folder available\")\n\n\ndef get_init_file():\n    # Init file must not exist, but it's parent dir must exist.\n    os.makedirs(str(get_shared_folder()), exist_ok=True)\n    init_file = get_shared_folder() / f\"{uuid.uuid4().hex}_init\"\n    if init_file.exists():\n        os.remove(str(init_file))\n    return init_file\n\n","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.submitit_pretrain.get_shared_folder","uri":"program://LLaMA-Adapter/function/gorilla.finetune.submitit_pretrain.get_shared_folder#L33-L39","kind":"function","name":"get_shared_folder","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":33,"end_line":39,"context_start_line":13,"context_end_line":59,"code":"from pathlib import Path\n\nimport main_pretrain as trainer\nimport submitit\n\n\ndef parse_args():\n    trainer_parser = trainer.get_args_parser()\n    parser = argparse.ArgumentParser(\"Submitit for MAE pretrain\", parents=[trainer_parser])\n    parser.add_argument(\"--ngpus\", default=8, type=int, help=\"Number of gpus to request on each node\")\n    parser.add_argument(\"--nodes\", default=2, type=int, help=\"Number of nodes to request\")\n    parser.add_argument(\"--timeout\", default=4320, type=int, help=\"Duration of the job\")\n    parser.add_argument(\"--job_dir\", default=\"\", type=str, help=\"Job dir. Leave empty for automatic.\")\n\n    parser.add_argument(\"--partition\", default=\"learnfair\", type=str, help=\"Partition where to submit\")\n    parser.add_argument(\"--use_volta32\", action='store_true', help=\"Request 32G V100 GPUs\")\n    parser.add_argument('--comment', default=\"\", type=str, help=\"Comment to pass to scheduler\")\n    return parser.parse_args()\n\n\ndef get_shared_folder() -> Path:\n    user = os.getenv(\"USER\")\n    if Path(\"/checkpoint/\").is_dir():\n        p = Path(f\"/checkpoint/{user}/experiments\")\n        p.mkdir(exist_ok=True)\n        return p\n    raise RuntimeError(\"No shared folder available\")\n\n\ndef get_init_file():\n    # Init file must not exist, but it's parent dir must exist.\n    os.makedirs(str(get_shared_folder()), exist_ok=True)\n    init_file = get_shared_folder() / f\"{uuid.uuid4().hex}_init\"\n    if init_file.exists():\n        os.remove(str(init_file))\n    return init_file\n\n\nclass Trainer(object):\n    def __init__(self, args):\n        self.args = args\n\n    def __call__(self):\n        import main_pretrain as trainer\n\n        self._setup_gpu_args()\n        trainer.main(self.args)","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.submitit_pretrain.get_init_file","uri":"program://LLaMA-Adapter/function/gorilla.finetune.submitit_pretrain.get_init_file#L42-L48","kind":"function","name":"get_init_file","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":42,"end_line":48,"context_start_line":22,"context_end_line":68,"code":"    parser.add_argument(\"--ngpus\", default=8, type=int, help=\"Number of gpus to request on each node\")\n    parser.add_argument(\"--nodes\", default=2, type=int, help=\"Number of nodes to request\")\n    parser.add_argument(\"--timeout\", default=4320, type=int, help=\"Duration of the job\")\n    parser.add_argument(\"--job_dir\", default=\"\", type=str, help=\"Job dir. Leave empty for automatic.\")\n\n    parser.add_argument(\"--partition\", default=\"learnfair\", type=str, help=\"Partition where to submit\")\n    parser.add_argument(\"--use_volta32\", action='store_true', help=\"Request 32G V100 GPUs\")\n    parser.add_argument('--comment', default=\"\", type=str, help=\"Comment to pass to scheduler\")\n    return parser.parse_args()\n\n\ndef get_shared_folder() -> Path:\n    user = os.getenv(\"USER\")\n    if Path(\"/checkpoint/\").is_dir():\n        p = Path(f\"/checkpoint/{user}/experiments\")\n        p.mkdir(exist_ok=True)\n        return p\n    raise RuntimeError(\"No shared folder available\")\n\n\ndef get_init_file():\n    # Init file must not exist, but it's parent dir must exist.\n    os.makedirs(str(get_shared_folder()), exist_ok=True)\n    init_file = get_shared_folder() / f\"{uuid.uuid4().hex}_init\"\n    if init_file.exists():\n        os.remove(str(init_file))\n    return init_file\n\n\nclass Trainer(object):\n    def __init__(self, args):\n        self.args = args\n\n    def __call__(self):\n        import main_pretrain as trainer\n\n        self._setup_gpu_args()\n        trainer.main(self.args)\n\n    def checkpoint(self):\n        import os\n        import submitit\n\n        self.args.dist_url = get_init_file().as_uri()\n        checkpoint_file = os.path.join(self.args.output_dir, \"checkpoint.pth\")\n        if os.path.exists(checkpoint_file):\n            self.args.resume = checkpoint_file","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.submitit_pretrain.Trainer","uri":"program://LLaMA-Adapter/class/gorilla.finetune.submitit_pretrain.Trainer#L51-L83","kind":"class","name":"Trainer","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":51,"end_line":83,"context_start_line":31,"context_end_line":103,"code":"\n\ndef get_shared_folder() -> Path:\n    user = os.getenv(\"USER\")\n    if Path(\"/checkpoint/\").is_dir():\n        p = Path(f\"/checkpoint/{user}/experiments\")\n        p.mkdir(exist_ok=True)\n        return p\n    raise RuntimeError(\"No shared folder available\")\n\n\ndef get_init_file():\n    # Init file must not exist, but it's parent dir must exist.\n    os.makedirs(str(get_shared_folder()), exist_ok=True)\n    init_file = get_shared_folder() / f\"{uuid.uuid4().hex}_init\"\n    if init_file.exists():\n        os.remove(str(init_file))\n    return init_file\n\n\nclass Trainer(object):\n    def __init__(self, args):\n        self.args = args\n\n    def __call__(self):\n        import main_pretrain as trainer\n\n        self._setup_gpu_args()\n        trainer.main(self.args)\n\n    def checkpoint(self):\n        import os\n        import submitit\n\n        self.args.dist_url = get_init_file().as_uri()\n        checkpoint_file = os.path.join(self.args.output_dir, \"checkpoint.pth\")\n        if os.path.exists(checkpoint_file):\n            self.args.resume = checkpoint_file\n        print(\"Requeuing \", self.args)\n        empty_trainer = type(self)(self.args)\n        return submitit.helpers.DelayedSubmission(empty_trainer)\n\n    def _setup_gpu_args(self):\n        import submitit\n        from pathlib import Path\n\n        job_env = submitit.JobEnvironment()\n        self.args.output_dir = Path(str(self.args.output_dir).replace(\"%j\", str(job_env.job_id)))\n        self.args.log_dir = self.args.output_dir\n        self.args.gpu = job_env.local_rank\n        self.args.rank = job_env.global_rank\n        self.args.world_size = job_env.num_tasks\n        print(f\"Process group: {job_env.num_tasks} tasks, rank: {job_env.global_rank}\")\n\n\ndef main():\n    args = parse_args()\n    if args.job_dir == \"\":\n        args.job_dir = get_shared_folder() / \"%j\"\n\n    # Note that the folder will depend on the job_id, to easily track experiments\n    executor = submitit.AutoExecutor(folder=args.job_dir, slurm_max_num_timeout=30)\n\n    num_gpus_per_node = args.ngpus\n    nodes = args.nodes\n    timeout_min = args.timeout\n\n    partition = args.partition\n    kwargs = {}\n    if args.use_volta32:\n        kwargs['slurm_constraint'] = 'volta32gb'\n    if args.comment:\n        kwargs['slurm_comment'] = args.comment","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.submitit_pretrain.main","uri":"program://LLaMA-Adapter/function/gorilla.finetune.submitit_pretrain.main#L86-L127","kind":"function","name":"main","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":86,"end_line":127,"context_start_line":66,"context_end_line":131,"code":"        checkpoint_file = os.path.join(self.args.output_dir, \"checkpoint.pth\")\n        if os.path.exists(checkpoint_file):\n            self.args.resume = checkpoint_file\n        print(\"Requeuing \", self.args)\n        empty_trainer = type(self)(self.args)\n        return submitit.helpers.DelayedSubmission(empty_trainer)\n\n    def _setup_gpu_args(self):\n        import submitit\n        from pathlib import Path\n\n        job_env = submitit.JobEnvironment()\n        self.args.output_dir = Path(str(self.args.output_dir).replace(\"%j\", str(job_env.job_id)))\n        self.args.log_dir = self.args.output_dir\n        self.args.gpu = job_env.local_rank\n        self.args.rank = job_env.global_rank\n        self.args.world_size = job_env.num_tasks\n        print(f\"Process group: {job_env.num_tasks} tasks, rank: {job_env.global_rank}\")\n\n\ndef main():\n    args = parse_args()\n    if args.job_dir == \"\":\n        args.job_dir = get_shared_folder() / \"%j\"\n\n    # Note that the folder will depend on the job_id, to easily track experiments\n    executor = submitit.AutoExecutor(folder=args.job_dir, slurm_max_num_timeout=30)\n\n    num_gpus_per_node = args.ngpus\n    nodes = args.nodes\n    timeout_min = args.timeout\n\n    partition = args.partition\n    kwargs = {}\n    if args.use_volta32:\n        kwargs['slurm_constraint'] = 'volta32gb'\n    if args.comment:\n        kwargs['slurm_comment'] = args.comment\n\n    executor.update_parameters(\n        mem_gb=40 * num_gpus_per_node,\n        gpus_per_node=num_gpus_per_node,\n        tasks_per_node=num_gpus_per_node,  # one task per GPU\n        cpus_per_task=10,\n        nodes=nodes,\n        timeout_min=timeout_min,  # max is 60 * 72\n        # Below are cluster dependent parameters\n        slurm_partition=partition,\n        slurm_signal_delay_s=120,\n        **kwargs\n    )\n\n    executor.update_parameters(name=\"mae\")\n\n    args.dist_url = get_init_file().as_uri()\n    args.output_dir = args.job_dir\n\n    trainer = Trainer(args)\n    job = executor.submit(trainer)\n\n    # print(\"Submitted job_id:\", job.job_id)\n    print(job.job_id)\n\n\nif __name__ == \"__main__\":\n    main()","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.submitit_pretrain.__init__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.submitit_pretrain.__init__#L52-L53","kind":"function","name":"__init__","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":52,"end_line":53,"context_start_line":32,"context_end_line":73,"code":"\ndef get_shared_folder() -> Path:\n    user = os.getenv(\"USER\")\n    if Path(\"/checkpoint/\").is_dir():\n        p = Path(f\"/checkpoint/{user}/experiments\")\n        p.mkdir(exist_ok=True)\n        return p\n    raise RuntimeError(\"No shared folder available\")\n\n\ndef get_init_file():\n    # Init file must not exist, but it's parent dir must exist.\n    os.makedirs(str(get_shared_folder()), exist_ok=True)\n    init_file = get_shared_folder() / f\"{uuid.uuid4().hex}_init\"\n    if init_file.exists():\n        os.remove(str(init_file))\n    return init_file\n\n\nclass Trainer(object):\n    def __init__(self, args):\n        self.args = args\n\n    def __call__(self):\n        import main_pretrain as trainer\n\n        self._setup_gpu_args()\n        trainer.main(self.args)\n\n    def checkpoint(self):\n        import os\n        import submitit\n\n        self.args.dist_url = get_init_file().as_uri()\n        checkpoint_file = os.path.join(self.args.output_dir, \"checkpoint.pth\")\n        if os.path.exists(checkpoint_file):\n            self.args.resume = checkpoint_file\n        print(\"Requeuing \", self.args)\n        empty_trainer = type(self)(self.args)\n        return submitit.helpers.DelayedSubmission(empty_trainer)\n\n    def _setup_gpu_args(self):","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.submitit_pretrain.__call__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.submitit_pretrain.__call__#L55-L59","kind":"function","name":"__call__","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":55,"end_line":59,"context_start_line":35,"context_end_line":79,"code":"    if Path(\"/checkpoint/\").is_dir():\n        p = Path(f\"/checkpoint/{user}/experiments\")\n        p.mkdir(exist_ok=True)\n        return p\n    raise RuntimeError(\"No shared folder available\")\n\n\ndef get_init_file():\n    # Init file must not exist, but it's parent dir must exist.\n    os.makedirs(str(get_shared_folder()), exist_ok=True)\n    init_file = get_shared_folder() / f\"{uuid.uuid4().hex}_init\"\n    if init_file.exists():\n        os.remove(str(init_file))\n    return init_file\n\n\nclass Trainer(object):\n    def __init__(self, args):\n        self.args = args\n\n    def __call__(self):\n        import main_pretrain as trainer\n\n        self._setup_gpu_args()\n        trainer.main(self.args)\n\n    def checkpoint(self):\n        import os\n        import submitit\n\n        self.args.dist_url = get_init_file().as_uri()\n        checkpoint_file = os.path.join(self.args.output_dir, \"checkpoint.pth\")\n        if os.path.exists(checkpoint_file):\n            self.args.resume = checkpoint_file\n        print(\"Requeuing \", self.args)\n        empty_trainer = type(self)(self.args)\n        return submitit.helpers.DelayedSubmission(empty_trainer)\n\n    def _setup_gpu_args(self):\n        import submitit\n        from pathlib import Path\n\n        job_env = submitit.JobEnvironment()\n        self.args.output_dir = Path(str(self.args.output_dir).replace(\"%j\", str(job_env.job_id)))\n        self.args.log_dir = self.args.output_dir","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.submitit_pretrain.checkpoint","uri":"program://LLaMA-Adapter/function/gorilla.finetune.submitit_pretrain.checkpoint#L61-L71","kind":"function","name":"checkpoint","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":61,"end_line":71,"context_start_line":41,"context_end_line":91,"code":"\ndef get_init_file():\n    # Init file must not exist, but it's parent dir must exist.\n    os.makedirs(str(get_shared_folder()), exist_ok=True)\n    init_file = get_shared_folder() / f\"{uuid.uuid4().hex}_init\"\n    if init_file.exists():\n        os.remove(str(init_file))\n    return init_file\n\n\nclass Trainer(object):\n    def __init__(self, args):\n        self.args = args\n\n    def __call__(self):\n        import main_pretrain as trainer\n\n        self._setup_gpu_args()\n        trainer.main(self.args)\n\n    def checkpoint(self):\n        import os\n        import submitit\n\n        self.args.dist_url = get_init_file().as_uri()\n        checkpoint_file = os.path.join(self.args.output_dir, \"checkpoint.pth\")\n        if os.path.exists(checkpoint_file):\n            self.args.resume = checkpoint_file\n        print(\"Requeuing \", self.args)\n        empty_trainer = type(self)(self.args)\n        return submitit.helpers.DelayedSubmission(empty_trainer)\n\n    def _setup_gpu_args(self):\n        import submitit\n        from pathlib import Path\n\n        job_env = submitit.JobEnvironment()\n        self.args.output_dir = Path(str(self.args.output_dir).replace(\"%j\", str(job_env.job_id)))\n        self.args.log_dir = self.args.output_dir\n        self.args.gpu = job_env.local_rank\n        self.args.rank = job_env.global_rank\n        self.args.world_size = job_env.num_tasks\n        print(f\"Process group: {job_env.num_tasks} tasks, rank: {job_env.global_rank}\")\n\n\ndef main():\n    args = parse_args()\n    if args.job_dir == \"\":\n        args.job_dir = get_shared_folder() / \"%j\"\n\n    # Note that the folder will depend on the job_id, to easily track experiments","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.submitit_pretrain._setup_gpu_args","uri":"program://LLaMA-Adapter/function/gorilla.finetune.submitit_pretrain._setup_gpu_args#L73-L83","kind":"function","name":"_setup_gpu_args","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":73,"end_line":83,"context_start_line":53,"context_end_line":103,"code":"        self.args = args\n\n    def __call__(self):\n        import main_pretrain as trainer\n\n        self._setup_gpu_args()\n        trainer.main(self.args)\n\n    def checkpoint(self):\n        import os\n        import submitit\n\n        self.args.dist_url = get_init_file().as_uri()\n        checkpoint_file = os.path.join(self.args.output_dir, \"checkpoint.pth\")\n        if os.path.exists(checkpoint_file):\n            self.args.resume = checkpoint_file\n        print(\"Requeuing \", self.args)\n        empty_trainer = type(self)(self.args)\n        return submitit.helpers.DelayedSubmission(empty_trainer)\n\n    def _setup_gpu_args(self):\n        import submitit\n        from pathlib import Path\n\n        job_env = submitit.JobEnvironment()\n        self.args.output_dir = Path(str(self.args.output_dir).replace(\"%j\", str(job_env.job_id)))\n        self.args.log_dir = self.args.output_dir\n        self.args.gpu = job_env.local_rank\n        self.args.rank = job_env.global_rank\n        self.args.world_size = job_env.num_tasks\n        print(f\"Process group: {job_env.num_tasks} tasks, rank: {job_env.global_rank}\")\n\n\ndef main():\n    args = parse_args()\n    if args.job_dir == \"\":\n        args.job_dir = get_shared_folder() / \"%j\"\n\n    # Note that the folder will depend on the job_id, to easily track experiments\n    executor = submitit.AutoExecutor(folder=args.job_dir, slurm_max_num_timeout=30)\n\n    num_gpus_per_node = args.ngpus\n    nodes = args.nodes\n    timeout_min = args.timeout\n\n    partition = args.partition\n    kwargs = {}\n    if args.use_volta32:\n        kwargs['slurm_constraint'] = 'volta32gb'\n    if args.comment:\n        kwargs['slurm_comment'] = args.comment","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc","uri":"program://LLaMA-Adapter/module/gorilla.finetune.util.misc#L1-L462","kind":"module","name":"gorilla.finetune.util.misc","path":"gorilla/finetune/util/misc.py","language":"python","start_line":1,"end_line":462,"context_start_line":1,"context_end_line":462,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\nimport subprocess\n\nimport torch\nimport torch.distributed as dist\nfrom torch import inf\nfrom torch.distributed.fsdp.sharded_grad_scaler import ShardedGradScaler\nfrom torch.distributed.fsdp import (\n    FullyShardedDataParallel as FSDP,\n    StateDictType,\n    FullStateDictConfig,\n)\n\nfrom fairscale.nn.model_parallel import initialize as fs_init\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None, start_iter=0):\n        i = start_iter\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        log_msg = [\n            header,\n            '[{0' + '}/{1}]',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0:\n                try:\n                    total_len = len(iterable)\n                except:\n                    total_len = \"unknown\"\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, total_len,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, total_len,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n#        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        os.environ['MASTER_PORT'] = '8964'\n        while 'MASTER_ADDR' not in os.environ or len(os.environ['MASTER_ADDR'].strip()) == 0:\n            os.environ['MASTER_ADDR'] = subprocess.check_output('sinfo -Nh -n %s | head -n 1 | awk \\'{print $1}\\'' % os.environ['SLURM_NODELIST'], shell=True, ).decode().strip()\n            time.sleep(1)\n        print(os.environ['MASTER_ADDR'])\n        args.world_size = int(os.environ['SLURM_NPROCS'])\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n        args.local_rank = args.gpu\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        os.environ['RANK'] = str(args.rank)\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\ndef init_distributed_mode1(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self, args):\n        self._scaler = ShardedGradScaler(enabled=args.precision in [\"fp16\"])\n\n    def __call__(self, loss, optimizer, model, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        if update_grad:\n            self._scaler.scale(loss).backward(create_graph=create_graph)\n            if clip_grad is not None:\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = model.clip_grad_norm_(clip_grad)\n            else:\n                raise NotImplementedError(\"please set clip_grad to a very large value if you do not want to clip.\")\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            with model.no_sync():\n                self._scaler.scale(loss).backward(create_graph=create_graph)\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(output_dir, args, model, optimizer, loss_scaler, dataset_state, epoch=None, iteration=None):\n    save_name = f\"epoch{epoch}\"\n    if iteration is not None:\n        save_name += f\"-iter{iteration}\"\n    save_dir = os.path.join(output_dir, save_name)\n\n    os.makedirs(save_dir, exist_ok=True)\n    with FSDP.state_dict_type(model, StateDictType.SHARDED_STATE_DICT):\n        to_save = {\n            \"model\": model.state_dict(),\n            \"optimizer\": optimizer.state_dict(),\n            \"epoch\": epoch,\n            \"iter\": iteration,\n            \"scaler\": loss_scaler.state_dict(),\n            \"args\": args,\n            \"dataset_state\": dataset_state,\n        }\n        save_path = os.path.join(\n            save_dir,\n            f\"checkpoint.{dist.get_rank():05d}-of-{dist.get_world_size():05d}.pth\",\n        )\n        torch.save(to_save, save_path)\n\n    if args.save_consolidated:\n        mp_rank = fs_init.get_model_parallel_rank()\n        mp_world_size = fs_init.get_model_parallel_world_size()\n        consolidated_model_save_path = os.path.join(\n            save_dir,\n            f\"consolidated.{mp_rank:02d}-of-{mp_world_size:02d}.pth\",\n        )\n        with FSDP.state_dict_type(\n            model,\n            StateDictType.FULL_STATE_DICT,\n            FullStateDictConfig(rank0_only=True, offload_to_cpu=True),\n        ):\n            save_dtype = {\n                \"fp16\": torch.float16,\n                \"bf16\": torch.bfloat16,\n                \"tf32\": torch.float32,\n            }[args.precision]\n            consolidated_model_state_dict = {\n                k: v.to(save_dtype) for k, v in model.state_dict().items()\n            }\n        if fs_init.get_data_parallel_rank() == 0:\n            torch.save(consolidated_model_state_dict, consolidated_model_save_path)\n\ndef load_model(args, model, optimizer, loss_scaler, dataset_train):\n    if args.resume:\n        print(\"Resume checkpoint %s\" % args.resume)\n        local_checkpoint_path = os.path.join(\n            args.resume,\n            f\"checkpoint.{dist.get_rank():05d}-of-{dist.get_world_size():05d}.pth\",\n        )\n        checkpoint = torch.load(local_checkpoint_path, map_location='cpu')\n        with FSDP.state_dict_type(model, StateDictType.SHARDED_STATE_DICT):\n            model.load_state_dict(checkpoint['model'])\n        optimizer.load_state_dict(checkpoint['optimizer'])\n        loss_scaler.load_state_dict(checkpoint['scaler'])\n        dataset_train.load_state_dict(checkpoint[\"dataset_state\"])\n\n        to_return = [\n            int(checkpoint['epoch']) + 1 if hasattr(checkpoint, 'epoch') else None,\n            int(checkpoint['iter']) + 1 if hasattr(checkpoint, 'iter') else None\n        ]\n        return to_return\n\n\ndef load_pretrained(load_dir, args, model):\n    mp_rank = fs_init.get_model_parallel_rank()\n    mp_world_size = fs_init.get_model_parallel_world_size()\n\n    dp_rank = fs_init.get_data_parallel_rank()\n    if dp_rank == 0: # later broadcast to all ranks through FSDP init\n        if args.pretrained_type == \"consolidated\":\n            state_dict_path = os.path.join(load_dir, f\"consolidated.{mp_rank:02d}-of-{mp_world_size:02d}.pth\")\n            state_dict = torch.load(state_dict_path, map_location='cpu')\n            model.load_state_dict(state_dict['model'])\n        elif args.pretrained_type == \"meta_ori\":\n            state_dict_path = os.path.join(load_dir, f\"consolidated.{mp_rank:02d}.pth\")\n            state_dict = torch.load(state_dict_path, map_location='cpu')\n            model_state = {f\"llma.{key}\": val for key, val in state_dict.items()}\n            model.load_state_dict(model_state, strict=False)\n\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        if isinstance(x, torch.Tensor):\n            x_reduce = x.clone().cuda()\n        else:\n            x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        #if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n        if name.endswith(\".bias\") or name.endswith(\"norm.weight\"):\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.SmoothedValue","uri":"program://LLaMA-Adapter/class/gorilla.finetune.util.misc.SmoothedValue#L33-L92","kind":"class","name":"SmoothedValue","path":"gorilla/finetune/util/misc.py","language":"python","start_line":33,"end_line":92,"context_start_line":13,"context_end_line":112,"code":"import datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\nimport subprocess\n\nimport torch\nimport torch.distributed as dist\nfrom torch import inf\nfrom torch.distributed.fsdp.sharded_grad_scaler import ShardedGradScaler\nfrom torch.distributed.fsdp import (\n    FullyShardedDataParallel as FSDP,\n    StateDictType,\n    FullStateDictConfig,\n)\n\nfrom fairscale.nn.model_parallel import initialize as fs_init\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.MetricLogger","uri":"program://LLaMA-Adapter/class/gorilla.finetune.util.misc.MetricLogger#L95-L176","kind":"class","name":"MetricLogger","path":"gorilla/finetune/util/misc.py","language":"python","start_line":95,"end_line":176,"context_start_line":75,"context_end_line":196,"code":"    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None, start_iter=0):\n        i = start_iter\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        log_msg = [\n            header,\n            '[{0' + '}/{1}]',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0:\n                try:\n                    total_len = len(iterable)\n                except:\n                    total_len = \"unknown\"\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, total_len,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, total_len,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n#        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.setup_for_distributed","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.setup_for_distributed#L179-L193","kind":"function","name":"setup_for_distributed","path":"gorilla/finetune/util/misc.py","language":"python","start_line":179,"end_line":193,"context_start_line":159,"context_end_line":213,"code":"                    total_len = \"unknown\"\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, total_len,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, total_len,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n#        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.is_dist_avail_and_initialized","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.is_dist_avail_and_initialized#L196-L201","kind":"function","name":"is_dist_avail_and_initialized","path":"gorilla/finetune/util/misc.py","language":"python","start_line":196,"end_line":201,"context_start_line":176,"context_end_line":221,"code":"            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n#        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.get_world_size","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.get_world_size#L204-L207","kind":"function","name":"get_world_size","path":"gorilla/finetune/util/misc.py","language":"python","start_line":204,"end_line":207,"context_start_line":184,"context_end_line":227,"code":"\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n#        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.get_rank","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.get_rank#L210-L213","kind":"function","name":"get_rank","path":"gorilla/finetune/util/misc.py","language":"python","start_line":210,"end_line":213,"context_start_line":190,"context_end_line":233,"code":"            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.is_main_process","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.is_main_process#L216-L217","kind":"function","name":"is_main_process","path":"gorilla/finetune/util/misc.py","language":"python","start_line":216,"end_line":217,"context_start_line":196,"context_end_line":237,"code":"def is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.save_on_master","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.save_on_master#L220-L222","kind":"function","name":"save_on_master","path":"gorilla/finetune/util/misc.py","language":"python","start_line":220,"end_line":222,"context_start_line":200,"context_end_line":242,"code":"        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        os.environ['MASTER_PORT'] = '8964'\n        while 'MASTER_ADDR' not in os.environ or len(os.environ['MASTER_ADDR'].strip()) == 0:\n            os.environ['MASTER_ADDR'] = subprocess.check_output('sinfo -Nh -n %s | head -n 1 | awk \\'{print $1}\\'' % os.environ['SLURM_NODELIST'], shell=True, ).decode().strip()\n            time.sleep(1)","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.init_distributed_mode","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.init_distributed_mode#L224-L266","kind":"function","name":"init_distributed_mode","path":"gorilla/finetune/util/misc.py","language":"python","start_line":224,"end_line":266,"context_start_line":204,"context_end_line":286,"code":"def get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        os.environ['MASTER_PORT'] = '8964'\n        while 'MASTER_ADDR' not in os.environ or len(os.environ['MASTER_ADDR'].strip()) == 0:\n            os.environ['MASTER_ADDR'] = subprocess.check_output('sinfo -Nh -n %s | head -n 1 | awk \\'{print $1}\\'' % os.environ['SLURM_NODELIST'], shell=True, ).decode().strip()\n            time.sleep(1)\n        print(os.environ['MASTER_ADDR'])\n        args.world_size = int(os.environ['SLURM_NPROCS'])\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n        args.local_rank = args.gpu\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        os.environ['RANK'] = str(args.rank)\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\ndef init_distributed_mode1(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.init_distributed_mode1","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.init_distributed_mode1#L269-L301","kind":"function","name":"init_distributed_mode1","path":"gorilla/finetune/util/misc.py","language":"python","start_line":269,"end_line":301,"context_start_line":249,"context_end_line":321,"code":"        os.environ['WORLD_SIZE'] = str(args.world_size)\n        os.environ['RANK'] = str(args.rank)\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\ndef init_distributed_mode1(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self, args):\n        self._scaler = ShardedGradScaler(enabled=args.precision in [\"fp16\"])\n\n    def __call__(self, loss, optimizer, model, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        if update_grad:\n            self._scaler.scale(loss).backward(create_graph=create_graph)\n            if clip_grad is not None:\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = model.clip_grad_norm_(clip_grad)\n            else:\n                raise NotImplementedError(\"please set clip_grad to a very large value if you do not want to clip.\")\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.NativeScalerWithGradNormCount","uri":"program://LLaMA-Adapter/class/gorilla.finetune.util.misc.NativeScalerWithGradNormCount#L304-L332","kind":"class","name":"NativeScalerWithGradNormCount","path":"gorilla/finetune/util/misc.py","language":"python","start_line":304,"end_line":332,"context_start_line":284,"context_end_line":352,"code":"        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self, args):\n        self._scaler = ShardedGradScaler(enabled=args.precision in [\"fp16\"])\n\n    def __call__(self, loss, optimizer, model, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        if update_grad:\n            self._scaler.scale(loss).backward(create_graph=create_graph)\n            if clip_grad is not None:\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = model.clip_grad_norm_(clip_grad)\n            else:\n                raise NotImplementedError(\"please set clip_grad to a very large value if you do not want to clip.\")\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            with model.no_sync():\n                self._scaler.scale(loss).backward(create_graph=create_graph)\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(output_dir, args, model, optimizer, loss_scaler, dataset_state, epoch=None, iteration=None):\n    save_name = f\"epoch{epoch}\"\n    if iteration is not None:","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.get_grad_norm_","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.get_grad_norm_#L335-L347","kind":"function","name":"get_grad_norm_","path":"gorilla/finetune/util/misc.py","language":"python","start_line":335,"end_line":347,"context_start_line":315,"context_end_line":367,"code":"                norm = model.clip_grad_norm_(clip_grad)\n            else:\n                raise NotImplementedError(\"please set clip_grad to a very large value if you do not want to clip.\")\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            with model.no_sync():\n                self._scaler.scale(loss).backward(create_graph=create_graph)\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(output_dir, args, model, optimizer, loss_scaler, dataset_state, epoch=None, iteration=None):\n    save_name = f\"epoch{epoch}\"\n    if iteration is not None:\n        save_name += f\"-iter{iteration}\"\n    save_dir = os.path.join(output_dir, save_name)\n\n    os.makedirs(save_dir, exist_ok=True)\n    with FSDP.state_dict_type(model, StateDictType.SHARDED_STATE_DICT):\n        to_save = {\n            \"model\": model.state_dict(),\n            \"optimizer\": optimizer.state_dict(),\n            \"epoch\": epoch,\n            \"iter\": iteration,\n            \"scaler\": loss_scaler.state_dict(),\n            \"args\": args,\n            \"dataset_state\": dataset_state,\n        }\n        save_path = os.path.join(","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.save_model","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.save_model#L350-L394","kind":"function","name":"save_model","path":"gorilla/finetune/util/misc.py","language":"python","start_line":350,"end_line":394,"context_start_line":330,"context_end_line":414,"code":"\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(output_dir, args, model, optimizer, loss_scaler, dataset_state, epoch=None, iteration=None):\n    save_name = f\"epoch{epoch}\"\n    if iteration is not None:\n        save_name += f\"-iter{iteration}\"\n    save_dir = os.path.join(output_dir, save_name)\n\n    os.makedirs(save_dir, exist_ok=True)\n    with FSDP.state_dict_type(model, StateDictType.SHARDED_STATE_DICT):\n        to_save = {\n            \"model\": model.state_dict(),\n            \"optimizer\": optimizer.state_dict(),\n            \"epoch\": epoch,\n            \"iter\": iteration,\n            \"scaler\": loss_scaler.state_dict(),\n            \"args\": args,\n            \"dataset_state\": dataset_state,\n        }\n        save_path = os.path.join(\n            save_dir,\n            f\"checkpoint.{dist.get_rank():05d}-of-{dist.get_world_size():05d}.pth\",\n        )\n        torch.save(to_save, save_path)\n\n    if args.save_consolidated:\n        mp_rank = fs_init.get_model_parallel_rank()\n        mp_world_size = fs_init.get_model_parallel_world_size()\n        consolidated_model_save_path = os.path.join(\n            save_dir,\n            f\"consolidated.{mp_rank:02d}-of-{mp_world_size:02d}.pth\",\n        )\n        with FSDP.state_dict_type(\n            model,\n            StateDictType.FULL_STATE_DICT,\n            FullStateDictConfig(rank0_only=True, offload_to_cpu=True),\n        ):\n            save_dtype = {\n                \"fp16\": torch.float16,\n                \"bf16\": torch.bfloat16,\n                \"tf32\": torch.float32,\n            }[args.precision]\n            consolidated_model_state_dict = {\n                k: v.to(save_dtype) for k, v in model.state_dict().items()\n            }\n        if fs_init.get_data_parallel_rank() == 0:\n            torch.save(consolidated_model_state_dict, consolidated_model_save_path)\n\ndef load_model(args, model, optimizer, loss_scaler, dataset_train):\n    if args.resume:\n        print(\"Resume checkpoint %s\" % args.resume)\n        local_checkpoint_path = os.path.join(\n            args.resume,\n            f\"checkpoint.{dist.get_rank():05d}-of-{dist.get_world_size():05d}.pth\",\n        )\n        checkpoint = torch.load(local_checkpoint_path, map_location='cpu')\n        with FSDP.state_dict_type(model, StateDictType.SHARDED_STATE_DICT):\n            model.load_state_dict(checkpoint['model'])\n        optimizer.load_state_dict(checkpoint['optimizer'])\n        loss_scaler.load_state_dict(checkpoint['scaler'])\n        dataset_train.load_state_dict(checkpoint[\"dataset_state\"])\n\n        to_return = [\n            int(checkpoint['epoch']) + 1 if hasattr(checkpoint, 'epoch') else None,\n            int(checkpoint['iter']) + 1 if hasattr(checkpoint, 'iter') else None\n        ]\n        return to_return","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.load_model","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.load_model#L396-L414","kind":"function","name":"load_model","path":"gorilla/finetune/util/misc.py","language":"python","start_line":396,"end_line":414,"context_start_line":376,"context_end_line":434,"code":"        consolidated_model_save_path = os.path.join(\n            save_dir,\n            f\"consolidated.{mp_rank:02d}-of-{mp_world_size:02d}.pth\",\n        )\n        with FSDP.state_dict_type(\n            model,\n            StateDictType.FULL_STATE_DICT,\n            FullStateDictConfig(rank0_only=True, offload_to_cpu=True),\n        ):\n            save_dtype = {\n                \"fp16\": torch.float16,\n                \"bf16\": torch.bfloat16,\n                \"tf32\": torch.float32,\n            }[args.precision]\n            consolidated_model_state_dict = {\n                k: v.to(save_dtype) for k, v in model.state_dict().items()\n            }\n        if fs_init.get_data_parallel_rank() == 0:\n            torch.save(consolidated_model_state_dict, consolidated_model_save_path)\n\ndef load_model(args, model, optimizer, loss_scaler, dataset_train):\n    if args.resume:\n        print(\"Resume checkpoint %s\" % args.resume)\n        local_checkpoint_path = os.path.join(\n            args.resume,\n            f\"checkpoint.{dist.get_rank():05d}-of-{dist.get_world_size():05d}.pth\",\n        )\n        checkpoint = torch.load(local_checkpoint_path, map_location='cpu')\n        with FSDP.state_dict_type(model, StateDictType.SHARDED_STATE_DICT):\n            model.load_state_dict(checkpoint['model'])\n        optimizer.load_state_dict(checkpoint['optimizer'])\n        loss_scaler.load_state_dict(checkpoint['scaler'])\n        dataset_train.load_state_dict(checkpoint[\"dataset_state\"])\n\n        to_return = [\n            int(checkpoint['epoch']) + 1 if hasattr(checkpoint, 'epoch') else None,\n            int(checkpoint['iter']) + 1 if hasattr(checkpoint, 'iter') else None\n        ]\n        return to_return\n\n\ndef load_pretrained(load_dir, args, model):\n    mp_rank = fs_init.get_model_parallel_rank()\n    mp_world_size = fs_init.get_model_parallel_world_size()\n\n    dp_rank = fs_init.get_data_parallel_rank()\n    if dp_rank == 0: # later broadcast to all ranks through FSDP init\n        if args.pretrained_type == \"consolidated\":\n            state_dict_path = os.path.join(load_dir, f\"consolidated.{mp_rank:02d}-of-{mp_world_size:02d}.pth\")\n            state_dict = torch.load(state_dict_path, map_location='cpu')\n            model.load_state_dict(state_dict['model'])\n        elif args.pretrained_type == \"meta_ori\":\n            state_dict_path = os.path.join(load_dir, f\"consolidated.{mp_rank:02d}.pth\")\n            state_dict = torch.load(state_dict_path, map_location='cpu')\n            model_state = {f\"llma.{key}\": val for key, val in state_dict.items()}\n            model.load_state_dict(model_state, strict=False)\n\n\n","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.load_pretrained","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.load_pretrained#L417-L431","kind":"function","name":"load_pretrained","path":"gorilla/finetune/util/misc.py","language":"python","start_line":417,"end_line":431,"context_start_line":397,"context_end_line":451,"code":"    if args.resume:\n        print(\"Resume checkpoint %s\" % args.resume)\n        local_checkpoint_path = os.path.join(\n            args.resume,\n            f\"checkpoint.{dist.get_rank():05d}-of-{dist.get_world_size():05d}.pth\",\n        )\n        checkpoint = torch.load(local_checkpoint_path, map_location='cpu')\n        with FSDP.state_dict_type(model, StateDictType.SHARDED_STATE_DICT):\n            model.load_state_dict(checkpoint['model'])\n        optimizer.load_state_dict(checkpoint['optimizer'])\n        loss_scaler.load_state_dict(checkpoint['scaler'])\n        dataset_train.load_state_dict(checkpoint[\"dataset_state\"])\n\n        to_return = [\n            int(checkpoint['epoch']) + 1 if hasattr(checkpoint, 'epoch') else None,\n            int(checkpoint['iter']) + 1 if hasattr(checkpoint, 'iter') else None\n        ]\n        return to_return\n\n\ndef load_pretrained(load_dir, args, model):\n    mp_rank = fs_init.get_model_parallel_rank()\n    mp_world_size = fs_init.get_model_parallel_world_size()\n\n    dp_rank = fs_init.get_data_parallel_rank()\n    if dp_rank == 0: # later broadcast to all ranks through FSDP init\n        if args.pretrained_type == \"consolidated\":\n            state_dict_path = os.path.join(load_dir, f\"consolidated.{mp_rank:02d}-of-{mp_world_size:02d}.pth\")\n            state_dict = torch.load(state_dict_path, map_location='cpu')\n            model.load_state_dict(state_dict['model'])\n        elif args.pretrained_type == \"meta_ori\":\n            state_dict_path = os.path.join(load_dir, f\"consolidated.{mp_rank:02d}.pth\")\n            state_dict = torch.load(state_dict_path, map_location='cpu')\n            model_state = {f\"llma.{key}\": val for key, val in state_dict.items()}\n            model.load_state_dict(model_state, strict=False)\n\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        if isinstance(x, torch.Tensor):\n            x_reduce = x.clone().cuda()\n        else:\n            x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.all_reduce_mean","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.all_reduce_mean#L435-L446","kind":"function","name":"all_reduce_mean","path":"gorilla/finetune/util/misc.py","language":"python","start_line":435,"end_line":446,"context_start_line":415,"context_end_line":462,"code":"\n\ndef load_pretrained(load_dir, args, model):\n    mp_rank = fs_init.get_model_parallel_rank()\n    mp_world_size = fs_init.get_model_parallel_world_size()\n\n    dp_rank = fs_init.get_data_parallel_rank()\n    if dp_rank == 0: # later broadcast to all ranks through FSDP init\n        if args.pretrained_type == \"consolidated\":\n            state_dict_path = os.path.join(load_dir, f\"consolidated.{mp_rank:02d}-of-{mp_world_size:02d}.pth\")\n            state_dict = torch.load(state_dict_path, map_location='cpu')\n            model.load_state_dict(state_dict['model'])\n        elif args.pretrained_type == \"meta_ori\":\n            state_dict_path = os.path.join(load_dir, f\"consolidated.{mp_rank:02d}.pth\")\n            state_dict = torch.load(state_dict_path, map_location='cpu')\n            model_state = {f\"llma.{key}\": val for key, val in state_dict.items()}\n            model.load_state_dict(model_state, strict=False)\n\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        if isinstance(x, torch.Tensor):\n            x_reduce = x.clone().cuda()\n        else:\n            x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        #if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n        if name.endswith(\".bias\") or name.endswith(\"norm.weight\"):\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.add_weight_decay","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.add_weight_decay#L449-L462","kind":"function","name":"add_weight_decay","path":"gorilla/finetune/util/misc.py","language":"python","start_line":449,"end_line":462,"context_start_line":429,"context_end_line":462,"code":"            state_dict = torch.load(state_dict_path, map_location='cpu')\n            model_state = {f\"llma.{key}\": val for key, val in state_dict.items()}\n            model.load_state_dict(model_state, strict=False)\n\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        if isinstance(x, torch.Tensor):\n            x_reduce = x.clone().cuda()\n        else:\n            x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        #if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n        if name.endswith(\".bias\") or name.endswith(\"norm.weight\"):\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.__init__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.__init__#L307-L308","kind":"function","name":"__init__","path":"gorilla/finetune/util/misc.py","language":"python","start_line":307,"end_line":308,"context_start_line":287,"context_end_line":328,"code":"        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self, args):\n        self._scaler = ShardedGradScaler(enabled=args.precision in [\"fp16\"])\n\n    def __call__(self, loss, optimizer, model, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        if update_grad:\n            self._scaler.scale(loss).backward(create_graph=create_graph)\n            if clip_grad is not None:\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = model.clip_grad_norm_(clip_grad)\n            else:\n                raise NotImplementedError(\"please set clip_grad to a very large value if you do not want to clip.\")\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            with model.no_sync():\n                self._scaler.scale(loss).backward(create_graph=create_graph)\n            norm = None\n        return norm\n\n    def state_dict(self):","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.update","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.update#L100-L107","kind":"function","name":"update","path":"gorilla/finetune/util/misc.py","language":"python","start_line":100,"end_line":107,"context_start_line":80,"context_end_line":127,"code":"        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.synchronize_between_processes","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.synchronize_between_processes#L125-L127","kind":"function","name":"synchronize_between_processes","path":"gorilla/finetune/util/misc.py","language":"python","start_line":125,"end_line":127,"context_start_line":105,"context_end_line":147,"code":"                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None, start_iter=0):\n        i = start_iter\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        log_msg = [\n            header,\n            '[{0' + '}/{1}]',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.median","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.median#L65-L67","kind":"function","name":"median","path":"gorilla/finetune/util/misc.py","language":"python","start_line":65,"end_line":67,"context_start_line":45,"context_end_line":87,"code":"\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.avg","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.avg#L70-L72","kind":"function","name":"avg","path":"gorilla/finetune/util/misc.py","language":"python","start_line":70,"end_line":72,"context_start_line":50,"context_end_line":92,"code":"\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.global_avg","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.global_avg#L75-L76","kind":"function","name":"global_avg","path":"gorilla/finetune/util/misc.py","language":"python","start_line":75,"end_line":76,"context_start_line":55,"context_end_line":96,"code":"        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.max","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.max#L79-L80","kind":"function","name":"max","path":"gorilla/finetune/util/misc.py","language":"python","start_line":79,"end_line":80,"context_start_line":59,"context_end_line":100,"code":"        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.value","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.value#L83-L84","kind":"function","name":"value","path":"gorilla/finetune/util/misc.py","language":"python","start_line":83,"end_line":84,"context_start_line":63,"context_end_line":104,"code":"\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.__str__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.__str__#L117-L123","kind":"function","name":"__str__","path":"gorilla/finetune/util/misc.py","language":"python","start_line":117,"end_line":123,"context_start_line":97,"context_end_line":143,"code":"        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None, start_iter=0):\n        i = start_iter\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        log_msg = [\n            header,\n            '[{0' + '}/{1}]',\n            '{meters}',","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.__getattr__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.__getattr__#L109-L115","kind":"function","name":"__getattr__","path":"gorilla/finetune/util/misc.py","language":"python","start_line":109,"end_line":115,"context_start_line":89,"context_end_line":135,"code":"            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None, start_iter=0):\n        i = start_iter\n        if not header:\n            header = ''","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.add_meter","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.add_meter#L129-L130","kind":"function","name":"add_meter","path":"gorilla/finetune/util/misc.py","language":"python","start_line":129,"end_line":130,"context_start_line":109,"context_end_line":150,"code":"    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None, start_iter=0):\n        i = start_iter\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        log_msg = [\n            header,\n            '[{0' + '}/{1}]',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.log_every","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.log_every#L132-L176","kind":"function","name":"log_every","path":"gorilla/finetune/util/misc.py","language":"python","start_line":132,"end_line":176,"context_start_line":112,"context_end_line":196,"code":"        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None, start_iter=0):\n        i = start_iter\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        log_msg = [\n            header,\n            '[{0' + '}/{1}]',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0:\n                try:\n                    total_len = len(iterable)\n                except:\n                    total_len = \"unknown\"\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, total_len,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, total_len,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n#        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.print","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.print#L185-L191","kind":"function","name":"print","path":"gorilla/finetune/util/misc.py","language":"python","start_line":185,"end_line":191,"context_start_line":165,"context_end_line":211,"code":"                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, total_len,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n#        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.__call__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.__call__#L310-L326","kind":"function","name":"__call__","path":"gorilla/finetune/util/misc.py","language":"python","start_line":310,"end_line":326,"context_start_line":290,"context_end_line":346,"code":"        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self, args):\n        self._scaler = ShardedGradScaler(enabled=args.precision in [\"fp16\"])\n\n    def __call__(self, loss, optimizer, model, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        if update_grad:\n            self._scaler.scale(loss).backward(create_graph=create_graph)\n            if clip_grad is not None:\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = model.clip_grad_norm_(clip_grad)\n            else:\n                raise NotImplementedError(\"please set clip_grad to a very large value if you do not want to clip.\")\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            with model.no_sync():\n                self._scaler.scale(loss).backward(create_graph=create_graph)\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.state_dict","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.state_dict#L328-L329","kind":"function","name":"state_dict","path":"gorilla/finetune/util/misc.py","language":"python","start_line":328,"end_line":329,"context_start_line":308,"context_end_line":349,"code":"        self._scaler = ShardedGradScaler(enabled=args.precision in [\"fp16\"])\n\n    def __call__(self, loss, optimizer, model, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        if update_grad:\n            self._scaler.scale(loss).backward(create_graph=create_graph)\n            if clip_grad is not None:\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = model.clip_grad_norm_(clip_grad)\n            else:\n                raise NotImplementedError(\"please set clip_grad to a very large value if you do not want to clip.\")\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            with model.no_sync():\n                self._scaler.scale(loss).backward(create_graph=create_graph)\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.misc.load_state_dict","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.misc.load_state_dict#L331-L332","kind":"function","name":"load_state_dict","path":"gorilla/finetune/util/misc.py","language":"python","start_line":331,"end_line":332,"context_start_line":311,"context_end_line":352,"code":"        if update_grad:\n            self._scaler.scale(loss).backward(create_graph=create_graph)\n            if clip_grad is not None:\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = model.clip_grad_norm_(clip_grad)\n            else:\n                raise NotImplementedError(\"please set clip_grad to a very large value if you do not want to clip.\")\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            with model.no_sync():\n                self._scaler.scale(loss).backward(create_graph=create_graph)\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(output_dir, args, model, optimizer, loss_scaler, dataset_state, epoch=None, iteration=None):\n    save_name = f\"epoch{epoch}\"\n    if iteration is not None:","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.lr_decay","uri":"program://LLaMA-Adapter/module/gorilla.finetune.util.lr_decay#L1-L76","kind":"module","name":"gorilla.finetune.util.lr_decay","path":"gorilla/finetune/util/lr_decay.py","language":"python","start_line":1,"end_line":76,"context_start_line":1,"context_end_line":76,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# ELECTRA https://github.com/google-research/electra\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport json\n\n\ndef param_groups_lrd(model, weight_decay=0.05, no_weight_decay_list=[], layer_decay=.75):\n    \"\"\"\n    Parameter groups for layer-wise lr decay\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L58\n    \"\"\"\n    param_group_names = {}\n    param_groups = {}\n\n    num_layers = len(model.blocks) + 1\n\n    layer_scales = list(layer_decay ** (num_layers - i) for i in range(num_layers + 1))\n\n    for n, p in model.named_parameters():\n        if not p.requires_grad:\n            continue\n\n        # no decay: all 1D parameters and model specific ones\n        if p.ndim == 1 or n in no_weight_decay_list:\n            g_decay = \"no_decay\"\n            this_decay = 0.\n        else:\n            g_decay = \"decay\"\n            this_decay = weight_decay\n            \n        layer_id = get_layer_id_for_vit(n, num_layers)\n        group_name = \"layer_%d_%s\" % (layer_id, g_decay)\n\n        if group_name not in param_group_names:\n            this_scale = layer_scales[layer_id]\n\n            param_group_names[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n            param_groups[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n\n        param_group_names[group_name][\"params\"].append(n)\n        param_groups[group_name][\"params\"].append(p)\n\n    # print(\"parameter groups: \\n%s\" % json.dumps(param_group_names, indent=2))\n\n    return list(param_groups.values())\n\n\ndef get_layer_id_for_vit(name, num_layers):\n    \"\"\"\n    Assign a parameter with its layer id\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L33\n    \"\"\"\n    if name in ['cls_token', 'pos_embed']:\n        return 0\n    elif name.startswith('patch_embed'):\n        return 0\n    elif name.startswith('blocks'):\n        return int(name.split('.')[1]) + 1\n    else:\n        return num_layers","source_hash":"ff0ea0cf26230819d56c2c839e6c8e1b8c4f9834c46a8da226d601042bd2cdd5","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.lr_decay.param_groups_lrd","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.lr_decay.param_groups_lrd#L15-L61","kind":"function","name":"param_groups_lrd","path":"gorilla/finetune/util/lr_decay.py","language":"python","start_line":15,"end_line":61,"context_start_line":1,"context_end_line":76,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# ELECTRA https://github.com/google-research/electra\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport json\n\n\ndef param_groups_lrd(model, weight_decay=0.05, no_weight_decay_list=[], layer_decay=.75):\n    \"\"\"\n    Parameter groups for layer-wise lr decay\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L58\n    \"\"\"\n    param_group_names = {}\n    param_groups = {}\n\n    num_layers = len(model.blocks) + 1\n\n    layer_scales = list(layer_decay ** (num_layers - i) for i in range(num_layers + 1))\n\n    for n, p in model.named_parameters():\n        if not p.requires_grad:\n            continue\n\n        # no decay: all 1D parameters and model specific ones\n        if p.ndim == 1 or n in no_weight_decay_list:\n            g_decay = \"no_decay\"\n            this_decay = 0.\n        else:\n            g_decay = \"decay\"\n            this_decay = weight_decay\n            \n        layer_id = get_layer_id_for_vit(n, num_layers)\n        group_name = \"layer_%d_%s\" % (layer_id, g_decay)\n\n        if group_name not in param_group_names:\n            this_scale = layer_scales[layer_id]\n\n            param_group_names[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n            param_groups[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n\n        param_group_names[group_name][\"params\"].append(n)\n        param_groups[group_name][\"params\"].append(p)\n\n    # print(\"parameter groups: \\n%s\" % json.dumps(param_group_names, indent=2))\n\n    return list(param_groups.values())\n\n\ndef get_layer_id_for_vit(name, num_layers):\n    \"\"\"\n    Assign a parameter with its layer id\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L33\n    \"\"\"\n    if name in ['cls_token', 'pos_embed']:\n        return 0\n    elif name.startswith('patch_embed'):\n        return 0\n    elif name.startswith('blocks'):\n        return int(name.split('.')[1]) + 1\n    else:\n        return num_layers","source_hash":"ff0ea0cf26230819d56c2c839e6c8e1b8c4f9834c46a8da226d601042bd2cdd5","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.lr_decay.get_layer_id_for_vit","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.lr_decay.get_layer_id_for_vit#L64-L76","kind":"function","name":"get_layer_id_for_vit","path":"gorilla/finetune/util/lr_decay.py","language":"python","start_line":64,"end_line":76,"context_start_line":44,"context_end_line":76,"code":"\n            param_group_names[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n            param_groups[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n\n        param_group_names[group_name][\"params\"].append(n)\n        param_groups[group_name][\"params\"].append(p)\n\n    # print(\"parameter groups: \\n%s\" % json.dumps(param_group_names, indent=2))\n\n    return list(param_groups.values())\n\n\ndef get_layer_id_for_vit(name, num_layers):\n    \"\"\"\n    Assign a parameter with its layer id\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L33\n    \"\"\"\n    if name in ['cls_token', 'pos_embed']:\n        return 0\n    elif name.startswith('patch_embed'):\n        return 0\n    elif name.startswith('blocks'):\n        return int(name.split('.')[1]) + 1\n    else:\n        return num_layers","source_hash":"ff0ea0cf26230819d56c2c839e6c8e1b8c4f9834c46a8da226d601042bd2cdd5","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.crop","uri":"program://LLaMA-Adapter/module/gorilla.finetune.util.crop#L1-L42","kind":"module","name":"gorilla.finetune.util.crop","path":"gorilla/finetune/util/crop.py","language":"python","start_line":1,"end_line":42,"context_start_line":1,"context_end_line":42,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\nimport torch\n\nfrom torchvision import transforms\nfrom torchvision.transforms import functional as F\n\n\nclass RandomResizedCrop(transforms.RandomResizedCrop):\n    \"\"\"\n    RandomResizedCrop for matching TF/TPU implementation: no for-loop is used.\n    This may lead to results different with torchvision's version.\n    Following BYOL's TF code:\n    https://github.com/deepmind/deepmind-research/blob/master/byol/utils/dataset.py#L206\n    \"\"\"\n    @staticmethod\n    def get_params(img, scale, ratio):\n        width, height = F._get_image_size(img)\n        area = height * width\n\n        target_area = area * torch.empty(1).uniform_(scale[0], scale[1]).item()\n        log_ratio = torch.log(torch.tensor(ratio))\n        aspect_ratio = torch.exp(\n            torch.empty(1).uniform_(log_ratio[0], log_ratio[1])\n        ).item()\n\n        w = int(round(math.sqrt(target_area * aspect_ratio)))\n        h = int(round(math.sqrt(target_area / aspect_ratio)))\n\n        w = min(w, width)\n        h = min(h, height)\n\n        i = torch.randint(0, height - h + 1, size=(1,)).item()\n        j = torch.randint(0, width - w + 1, size=(1,)).item()\n\n        return i, j, h, w","source_hash":"494b97fecddf698fa1089b9c16fae97a4fa22a809c1a4f908dcb09387d2b36ff","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.crop.RandomResizedCrop","uri":"program://LLaMA-Adapter/class/gorilla.finetune.util.crop.RandomResizedCrop#L15-L42","kind":"class","name":"RandomResizedCrop","path":"gorilla/finetune/util/crop.py","language":"python","start_line":15,"end_line":42,"context_start_line":1,"context_end_line":42,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\nimport torch\n\nfrom torchvision import transforms\nfrom torchvision.transforms import functional as F\n\n\nclass RandomResizedCrop(transforms.RandomResizedCrop):\n    \"\"\"\n    RandomResizedCrop for matching TF/TPU implementation: no for-loop is used.\n    This may lead to results different with torchvision's version.\n    Following BYOL's TF code:\n    https://github.com/deepmind/deepmind-research/blob/master/byol/utils/dataset.py#L206\n    \"\"\"\n    @staticmethod\n    def get_params(img, scale, ratio):\n        width, height = F._get_image_size(img)\n        area = height * width\n\n        target_area = area * torch.empty(1).uniform_(scale[0], scale[1]).item()\n        log_ratio = torch.log(torch.tensor(ratio))\n        aspect_ratio = torch.exp(\n            torch.empty(1).uniform_(log_ratio[0], log_ratio[1])\n        ).item()\n\n        w = int(round(math.sqrt(target_area * aspect_ratio)))\n        h = int(round(math.sqrt(target_area / aspect_ratio)))\n\n        w = min(w, width)\n        h = min(h, height)\n\n        i = torch.randint(0, height - h + 1, size=(1,)).item()\n        j = torch.randint(0, width - w + 1, size=(1,)).item()\n\n        return i, j, h, w","source_hash":"494b97fecddf698fa1089b9c16fae97a4fa22a809c1a4f908dcb09387d2b36ff","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.crop.get_params","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.crop.get_params#L23-L42","kind":"function","name":"get_params","path":"gorilla/finetune/util/crop.py","language":"python","start_line":23,"end_line":42,"context_start_line":3,"context_end_line":42,"code":"\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\nimport torch\n\nfrom torchvision import transforms\nfrom torchvision.transforms import functional as F\n\n\nclass RandomResizedCrop(transforms.RandomResizedCrop):\n    \"\"\"\n    RandomResizedCrop for matching TF/TPU implementation: no for-loop is used.\n    This may lead to results different with torchvision's version.\n    Following BYOL's TF code:\n    https://github.com/deepmind/deepmind-research/blob/master/byol/utils/dataset.py#L206\n    \"\"\"\n    @staticmethod\n    def get_params(img, scale, ratio):\n        width, height = F._get_image_size(img)\n        area = height * width\n\n        target_area = area * torch.empty(1).uniform_(scale[0], scale[1]).item()\n        log_ratio = torch.log(torch.tensor(ratio))\n        aspect_ratio = torch.exp(\n            torch.empty(1).uniform_(log_ratio[0], log_ratio[1])\n        ).item()\n\n        w = int(round(math.sqrt(target_area * aspect_ratio)))\n        h = int(round(math.sqrt(target_area / aspect_ratio)))\n\n        w = min(w, width)\n        h = min(h, height)\n\n        i = torch.randint(0, height - h + 1, size=(1,)).item()\n        j = torch.randint(0, width - w + 1, size=(1,)).item()\n\n        return i, j, h, w","source_hash":"494b97fecddf698fa1089b9c16fae97a4fa22a809c1a4f908dcb09387d2b36ff","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.lr_sched","uri":"program://LLaMA-Adapter/module/gorilla.finetune.util.lr_sched#L1-L41","kind":"module","name":"gorilla.finetune.util.lr_sched","path":"gorilla/finetune/util/lr_sched.py","language":"python","start_line":1,"end_line":41,"context_start_line":1,"context_end_line":41,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\ndef adjust_learning_rate(optimizer, it, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if it < args.warmup_iters: # 1) linear warmup for warmup_iters steps\n        lr = args.lr * it / args.warmup_iters\n    elif it > args.lr_decay_iters: # 2) if it > lr_decay_iters, return min learning rate\n        lr = args.min_lr\n    else: # 3) in between, use cosine decay down to min learning rate\n        decay_ratio = (it - args.warmup_iters) / (args.lr_decay_iters - args.warmup_iters)\n        assert 0 <= decay_ratio <= 1\n        coeff = 0.5 * (1.0 + math.cos(math.pi * decay_ratio))  # coeff ranges 0..1\n        lr = args.min_lr + (args.lr - args.min_lr) * coeff\n\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr\n\n\ndef adjust_learning_rate_epoch(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs\n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * \\\n            (1. + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs)))\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"57ea96156d0b62bb6af56bda534e9f6f906d1e1b7939799c9b1402a9c659ab43","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.lr_sched.adjust_learning_rate","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.lr_sched.adjust_learning_rate#L9-L26","kind":"function","name":"adjust_learning_rate","path":"gorilla/finetune/util/lr_sched.py","language":"python","start_line":9,"end_line":26,"context_start_line":1,"context_end_line":41,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\ndef adjust_learning_rate(optimizer, it, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if it < args.warmup_iters: # 1) linear warmup for warmup_iters steps\n        lr = args.lr * it / args.warmup_iters\n    elif it > args.lr_decay_iters: # 2) if it > lr_decay_iters, return min learning rate\n        lr = args.min_lr\n    else: # 3) in between, use cosine decay down to min learning rate\n        decay_ratio = (it - args.warmup_iters) / (args.lr_decay_iters - args.warmup_iters)\n        assert 0 <= decay_ratio <= 1\n        coeff = 0.5 * (1.0 + math.cos(math.pi * decay_ratio))  # coeff ranges 0..1\n        lr = args.min_lr + (args.lr - args.min_lr) * coeff\n\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr\n\n\ndef adjust_learning_rate_epoch(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs\n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * \\\n            (1. + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs)))\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"57ea96156d0b62bb6af56bda534e9f6f906d1e1b7939799c9b1402a9c659ab43","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.lr_sched.adjust_learning_rate_epoch","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.lr_sched.adjust_learning_rate_epoch#L29-L41","kind":"function","name":"adjust_learning_rate_epoch","path":"gorilla/finetune/util/lr_sched.py","language":"python","start_line":29,"end_line":41,"context_start_line":9,"context_end_line":41,"code":"def adjust_learning_rate(optimizer, it, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if it < args.warmup_iters: # 1) linear warmup for warmup_iters steps\n        lr = args.lr * it / args.warmup_iters\n    elif it > args.lr_decay_iters: # 2) if it > lr_decay_iters, return min learning rate\n        lr = args.min_lr\n    else: # 3) in between, use cosine decay down to min learning rate\n        decay_ratio = (it - args.warmup_iters) / (args.lr_decay_iters - args.warmup_iters)\n        assert 0 <= decay_ratio <= 1\n        coeff = 0.5 * (1.0 + math.cos(math.pi * decay_ratio))  # coeff ranges 0..1\n        lr = args.min_lr + (args.lr - args.min_lr) * coeff\n\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr\n\n\ndef adjust_learning_rate_epoch(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs\n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * \\\n            (1. + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs)))\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"57ea96156d0b62bb6af56bda534e9f6f906d1e1b7939799c9b1402a9c659ab43","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.pos_embed","uri":"program://LLaMA-Adapter/module/gorilla.finetune.util.pos_embed#L1-L96","kind":"module","name":"gorilla.finetune.util.pos_embed","path":"gorilla/finetune/util/pos_embed.py","language":"python","start_line":1,"end_line":96,"context_start_line":1,"context_end_line":96,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# Position embedding utils\n# --------------------------------------------------------\n\nimport numpy as np\n\nimport torch\n\n# --------------------------------------------------------\n# 2D sine-cosine position embedding\n# References:\n# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py\n# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"\n    grid_size: int of the grid height and width\n    return:\n    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)\n    \"\"\"\n    grid_h = np.arange(grid_size, dtype=np.float32)\n    grid_w = np.arange(grid_size, dtype=np.float32)\n    grid = np.meshgrid(grid_w, grid_h)  # here w goes first\n    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1) # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.\n    omega = 1. / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum('m,d->md', pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out) # (M, D/2)\n    emb_cos = np.cos(out) # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)\n    return emb\n\n\n# --------------------------------------------------------\n# Interpolate position embeddings for high-resolution\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\ndef interpolate_pos_embed(model, checkpoint_model):\n    if 'pos_embed' in checkpoint_model:\n        pos_embed_checkpoint = checkpoint_model['pos_embed']\n        embedding_size = pos_embed_checkpoint.shape[-1]\n        num_patches = model.patch_embed.num_patches\n        num_extra_tokens = model.pos_embed.shape[-2] - num_patches\n        # height (== width) for the checkpoint position embedding\n        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)\n        # height (== width) for the new position embedding\n        new_size = int(num_patches ** 0.5)\n        # class_token and dist_token are kept unchanged\n        if orig_size != new_size:\n            print(\"Position interpolate from %dx%d to %dx%d\" % (orig_size, orig_size, new_size, new_size))\n            extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]\n            # only the position tokens are interpolated\n            pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:]\n            pos_tokens = pos_tokens.reshape(-1, orig_size, orig_size, embedding_size).permute(0, 3, 1, 2)\n            pos_tokens = torch.nn.functional.interpolate(\n                pos_tokens, size=(new_size, new_size), mode='bicubic', align_corners=False)\n            pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)\n            new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)\n            checkpoint_model['pos_embed'] = new_pos_embed","source_hash":"b50f20cb689db0bc58b40aec97a3767f124e9ae1c747efbb4f2c51a0b744a67f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.pos_embed.get_2d_sincos_pos_embed","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.pos_embed.get_2d_sincos_pos_embed#L20-L35","kind":"function","name":"get_2d_sincos_pos_embed","path":"gorilla/finetune/util/pos_embed.py","language":"python","start_line":20,"end_line":35,"context_start_line":1,"context_end_line":55,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# Position embedding utils\n# --------------------------------------------------------\n\nimport numpy as np\n\nimport torch\n\n# --------------------------------------------------------\n# 2D sine-cosine position embedding\n# References:\n# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py\n# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"\n    grid_size: int of the grid height and width\n    return:\n    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)\n    \"\"\"\n    grid_h = np.arange(grid_size, dtype=np.float32)\n    grid_w = np.arange(grid_size, dtype=np.float32)\n    grid = np.meshgrid(grid_w, grid_h)  # here w goes first\n    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1) # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0","source_hash":"b50f20cb689db0bc58b40aec97a3767f124e9ae1c747efbb4f2c51a0b744a67f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.pos_embed.get_2d_sincos_pos_embed_from_grid","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.pos_embed.get_2d_sincos_pos_embed_from_grid#L38-L46","kind":"function","name":"get_2d_sincos_pos_embed_from_grid","path":"gorilla/finetune/util/pos_embed.py","language":"python","start_line":38,"end_line":46,"context_start_line":18,"context_end_line":66,"code":"# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"\n    grid_size: int of the grid height and width\n    return:\n    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)\n    \"\"\"\n    grid_h = np.arange(grid_size, dtype=np.float32)\n    grid_w = np.arange(grid_size, dtype=np.float32)\n    grid = np.meshgrid(grid_w, grid_h)  # here w goes first\n    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1) # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.\n    omega = 1. / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum('m,d->md', pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out) # (M, D/2)\n    emb_cos = np.cos(out) # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)","source_hash":"b50f20cb689db0bc58b40aec97a3767f124e9ae1c747efbb4f2c51a0b744a67f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.pos_embed.get_1d_sincos_pos_embed_from_grid","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.pos_embed.get_1d_sincos_pos_embed_from_grid#L49-L67","kind":"function","name":"get_1d_sincos_pos_embed_from_grid","path":"gorilla/finetune/util/pos_embed.py","language":"python","start_line":49,"end_line":67,"context_start_line":29,"context_end_line":87,"code":"    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1) # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.\n    omega = 1. / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum('m,d->md', pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out) # (M, D/2)\n    emb_cos = np.cos(out) # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)\n    return emb\n\n\n# --------------------------------------------------------\n# Interpolate position embeddings for high-resolution\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\ndef interpolate_pos_embed(model, checkpoint_model):\n    if 'pos_embed' in checkpoint_model:\n        pos_embed_checkpoint = checkpoint_model['pos_embed']\n        embedding_size = pos_embed_checkpoint.shape[-1]\n        num_patches = model.patch_embed.num_patches\n        num_extra_tokens = model.pos_embed.shape[-2] - num_patches\n        # height (== width) for the checkpoint position embedding\n        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)\n        # height (== width) for the new position embedding\n        new_size = int(num_patches ** 0.5)\n        # class_token and dist_token are kept unchanged\n        if orig_size != new_size:\n            print(\"Position interpolate from %dx%d to %dx%d\" % (orig_size, orig_size, new_size, new_size))","source_hash":"b50f20cb689db0bc58b40aec97a3767f124e9ae1c747efbb4f2c51a0b744a67f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.util.pos_embed.interpolate_pos_embed","uri":"program://LLaMA-Adapter/function/gorilla.finetune.util.pos_embed.interpolate_pos_embed#L75-L96","kind":"function","name":"interpolate_pos_embed","path":"gorilla/finetune/util/pos_embed.py","language":"python","start_line":75,"end_line":96,"context_start_line":55,"context_end_line":96,"code":"    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.\n    omega = 1. / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum('m,d->md', pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out) # (M, D/2)\n    emb_cos = np.cos(out) # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)\n    return emb\n\n\n# --------------------------------------------------------\n# Interpolate position embeddings for high-resolution\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\ndef interpolate_pos_embed(model, checkpoint_model):\n    if 'pos_embed' in checkpoint_model:\n        pos_embed_checkpoint = checkpoint_model['pos_embed']\n        embedding_size = pos_embed_checkpoint.shape[-1]\n        num_patches = model.patch_embed.num_patches\n        num_extra_tokens = model.pos_embed.shape[-2] - num_patches\n        # height (== width) for the checkpoint position embedding\n        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)\n        # height (== width) for the new position embedding\n        new_size = int(num_patches ** 0.5)\n        # class_token and dist_token are kept unchanged\n        if orig_size != new_size:\n            print(\"Position interpolate from %dx%d to %dx%d\" % (orig_size, orig_size, new_size, new_size))\n            extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]\n            # only the position tokens are interpolated\n            pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:]\n            pos_tokens = pos_tokens.reshape(-1, orig_size, orig_size, embedding_size).permute(0, 3, 1, 2)\n            pos_tokens = torch.nn.functional.interpolate(\n                pos_tokens, size=(new_size, new_size), mode='bicubic', align_corners=False)\n            pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)\n            new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)\n            checkpoint_model['pos_embed'] = new_pos_embed","source_hash":"b50f20cb689db0bc58b40aec97a3767f124e9ae1c747efbb4f2c51a0b744a67f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.data.alpaca","uri":"program://LLaMA-Adapter/module/gorilla.finetune.data.alpaca#L1-L113","kind":"module","name":"gorilla.finetune.data.alpaca","path":"gorilla/finetune/data/alpaca.py","language":"python","start_line":1,"end_line":113,"context_start_line":1,"context_end_line":113,"code":"import torch\nimport yaml\nfrom torch.utils.data import Dataset\nfrom PIL import Image\nimport json\nfrom model.tokenizer import Tokenizer\nimport copy\nimport torchvision.transforms as transforms\nimport pdb\n\ntry:\n    from torchvision.transforms import InterpolationMode\n    BICUBIC = InterpolationMode.BICUBIC\nexcept ImportError:\n    BICUBIC = Image.BICUBIC\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\n\n# create data\ntransform_train = transforms.Compose([\n    transforms.RandomResizedCrop(size=(224, 224), scale=(0.9, 1.0), ratio=(0.75, 1.3333), interpolation=BICUBIC,\n                                 antialias=None),  # 3 is bicubic\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\n\ntransform_val = transforms.Compose([\n    transforms.Resize(\n        224, interpolation=transforms.InterpolationMode.BICUBIC\n    ),\n    transforms.CenterCrop(224),\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\nclass FinetuneDataset(Dataset):\n    def __init__(self, config_path, transform=transform_train, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        ann = []\n\n        data_lists = list(open(self.config['META'][0]))\n        for data in data_lists:\n            data = json.loads(data)['code']\n            if '###Instruction.' in data:\n                data = data.replace('###Instruction.', '###Instruction:')\n            if 'Output:' in data:\n               ann.append({'instruction':data.split('Output:')[0].split('Instruction:')[1].split('###')[0],\n                           'input': '',\n                           'output': data.split('Output:')[1]})\n        self.ann = ann\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n        data_item = self.ann[index]\n        if 'image' in data_item.keys():\n            filename = data_item['image']\n            question = data_item['conversations'][0]['value']\n            answer = data_item['conversations'][1]['value']\n\n            image = Image.open(filename).convert('RGB')\n            image = self.transform(image)\n            format_instruction = question\n            format_input = None\n        else:\n            image = torch.zeros(3, 224, 224)\n            format_instruction = data_item['instruction'],\n            format_input = data_item['input']\n            answer = data_item['output']\n        input1 = format_prompt(format_instruction, format_input)\n        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image","source_hash":"f1c356152d9979a79d2368bf2a2d3d8b55031e165d0ac81b96535f11fb44ddd1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.data.alpaca.format_prompt","uri":"program://LLaMA-Adapter/function/gorilla.finetune.data.alpaca.format_prompt#L18-L34","kind":"function","name":"format_prompt","path":"gorilla/finetune/data/alpaca.py","language":"python","start_line":18,"end_line":34,"context_start_line":1,"context_end_line":54,"code":"import torch\nimport yaml\nfrom torch.utils.data import Dataset\nfrom PIL import Image\nimport json\nfrom model.tokenizer import Tokenizer\nimport copy\nimport torchvision.transforms as transforms\nimport pdb\n\ntry:\n    from torchvision.transforms import InterpolationMode\n    BICUBIC = InterpolationMode.BICUBIC\nexcept ImportError:\n    BICUBIC = Image.BICUBIC\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\n\n# create data\ntransform_train = transforms.Compose([\n    transforms.RandomResizedCrop(size=(224, 224), scale=(0.9, 1.0), ratio=(0.75, 1.3333), interpolation=BICUBIC,\n                                 antialias=None),  # 3 is bicubic\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\n\ntransform_val = transforms.Compose([\n    transforms.Resize(\n        224, interpolation=transforms.InterpolationMode.BICUBIC\n    ),\n    transforms.CenterCrop(224),\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\nclass FinetuneDataset(Dataset):\n    def __init__(self, config_path, transform=transform_train, max_words=30, tokenizer_path=None):","source_hash":"f1c356152d9979a79d2368bf2a2d3d8b55031e165d0ac81b96535f11fb44ddd1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.data.alpaca.FinetuneDataset","uri":"program://LLaMA-Adapter/class/gorilla.finetune.data.alpaca.FinetuneDataset#L53-L113","kind":"class","name":"FinetuneDataset","path":"gorilla/finetune/data/alpaca.py","language":"python","start_line":53,"end_line":113,"context_start_line":33,"context_end_line":113,"code":"    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\n\n# create data\ntransform_train = transforms.Compose([\n    transforms.RandomResizedCrop(size=(224, 224), scale=(0.9, 1.0), ratio=(0.75, 1.3333), interpolation=BICUBIC,\n                                 antialias=None),  # 3 is bicubic\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\n\ntransform_val = transforms.Compose([\n    transforms.Resize(\n        224, interpolation=transforms.InterpolationMode.BICUBIC\n    ),\n    transforms.CenterCrop(224),\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\nclass FinetuneDataset(Dataset):\n    def __init__(self, config_path, transform=transform_train, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        ann = []\n\n        data_lists = list(open(self.config['META'][0]))\n        for data in data_lists:\n            data = json.loads(data)['code']\n            if '###Instruction.' in data:\n                data = data.replace('###Instruction.', '###Instruction:')\n            if 'Output:' in data:\n               ann.append({'instruction':data.split('Output:')[0].split('Instruction:')[1].split('###')[0],\n                           'input': '',\n                           'output': data.split('Output:')[1]})\n        self.ann = ann\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n        data_item = self.ann[index]\n        if 'image' in data_item.keys():\n            filename = data_item['image']\n            question = data_item['conversations'][0]['value']\n            answer = data_item['conversations'][1]['value']\n\n            image = Image.open(filename).convert('RGB')\n            image = self.transform(image)\n            format_instruction = question\n            format_input = None\n        else:\n            image = torch.zeros(3, 224, 224)\n            format_instruction = data_item['instruction'],\n            format_input = data_item['input']\n            answer = data_item['output']\n        input1 = format_prompt(format_instruction, format_input)\n        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image","source_hash":"f1c356152d9979a79d2368bf2a2d3d8b55031e165d0ac81b96535f11fb44ddd1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.data.alpaca.__init__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.data.alpaca.__init__#L54-L75","kind":"function","name":"__init__","path":"gorilla/finetune/data/alpaca.py","language":"python","start_line":54,"end_line":75,"context_start_line":34,"context_end_line":95,"code":"        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\n\n# create data\ntransform_train = transforms.Compose([\n    transforms.RandomResizedCrop(size=(224, 224), scale=(0.9, 1.0), ratio=(0.75, 1.3333), interpolation=BICUBIC,\n                                 antialias=None),  # 3 is bicubic\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\n\ntransform_val = transforms.Compose([\n    transforms.Resize(\n        224, interpolation=transforms.InterpolationMode.BICUBIC\n    ),\n    transforms.CenterCrop(224),\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\nclass FinetuneDataset(Dataset):\n    def __init__(self, config_path, transform=transform_train, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        ann = []\n\n        data_lists = list(open(self.config['META'][0]))\n        for data in data_lists:\n            data = json.loads(data)['code']\n            if '###Instruction.' in data:\n                data = data.replace('###Instruction.', '###Instruction:')\n            if 'Output:' in data:\n               ann.append({'instruction':data.split('Output:')[0].split('Instruction:')[1].split('###')[0],\n                           'input': '',\n                           'output': data.split('Output:')[1]})\n        self.ann = ann\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n        data_item = self.ann[index]\n        if 'image' in data_item.keys():\n            filename = data_item['image']\n            question = data_item['conversations'][0]['value']\n            answer = data_item['conversations'][1]['value']\n\n            image = Image.open(filename).convert('RGB')\n            image = self.transform(image)\n            format_instruction = question\n            format_input = None\n        else:\n            image = torch.zeros(3, 224, 224)\n            format_instruction = data_item['instruction'],\n            format_input = data_item['input']\n            answer = data_item['output']","source_hash":"f1c356152d9979a79d2368bf2a2d3d8b55031e165d0ac81b96535f11fb44ddd1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.data.alpaca.__len__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.data.alpaca.__len__#L77-L78","kind":"function","name":"__len__","path":"gorilla/finetune/data/alpaca.py","language":"python","start_line":77,"end_line":78,"context_start_line":57,"context_end_line":98,"code":"            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        ann = []\n\n        data_lists = list(open(self.config['META'][0]))\n        for data in data_lists:\n            data = json.loads(data)['code']\n            if '###Instruction.' in data:\n                data = data.replace('###Instruction.', '###Instruction:')\n            if 'Output:' in data:\n               ann.append({'instruction':data.split('Output:')[0].split('Instruction:')[1].split('###')[0],\n                           'input': '',\n                           'output': data.split('Output:')[1]})\n        self.ann = ann\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n        data_item = self.ann[index]\n        if 'image' in data_item.keys():\n            filename = data_item['image']\n            question = data_item['conversations'][0]['value']\n            answer = data_item['conversations'][1]['value']\n\n            image = Image.open(filename).convert('RGB')\n            image = self.transform(image)\n            format_instruction = question\n            format_input = None\n        else:\n            image = torch.zeros(3, 224, 224)\n            format_instruction = data_item['instruction'],\n            format_input = data_item['input']\n            answer = data_item['output']\n        input1 = format_prompt(format_instruction, format_input)\n        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)","source_hash":"f1c356152d9979a79d2368bf2a2d3d8b55031e165d0ac81b96535f11fb44ddd1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.data.alpaca.__getitem__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.data.alpaca.__getitem__#L80-L113","kind":"function","name":"__getitem__","path":"gorilla/finetune/data/alpaca.py","language":"python","start_line":80,"end_line":113,"context_start_line":60,"context_end_line":113,"code":"        ann = []\n\n        data_lists = list(open(self.config['META'][0]))\n        for data in data_lists:\n            data = json.loads(data)['code']\n            if '###Instruction.' in data:\n                data = data.replace('###Instruction.', '###Instruction:')\n            if 'Output:' in data:\n               ann.append({'instruction':data.split('Output:')[0].split('Instruction:')[1].split('###')[0],\n                           'input': '',\n                           'output': data.split('Output:')[1]})\n        self.ann = ann\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n        data_item = self.ann[index]\n        if 'image' in data_item.keys():\n            filename = data_item['image']\n            question = data_item['conversations'][0]['value']\n            answer = data_item['conversations'][1]['value']\n\n            image = Image.open(filename).convert('RGB')\n            image = self.transform(image)\n            format_instruction = question\n            format_input = None\n        else:\n            image = torch.zeros(3, 224, 224)\n            format_instruction = data_item['instruction'],\n            format_input = data_item['input']\n            answer = data_item['output']\n        input1 = format_prompt(format_instruction, format_input)\n        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image","source_hash":"f1c356152d9979a79d2368bf2a2d3d8b55031e165d0ac81b96535f11fb44ddd1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.meta","uri":"program://LLaMA-Adapter/module/gorilla.finetune.model.meta#L1-L52","kind":"module","name":"gorilla.finetune.model.meta","path":"gorilla/finetune/model/meta.py","language":"python","start_line":1,"end_line":52,"context_start_line":1,"context_end_line":52,"code":"import torch\nimport torch.nn as nn\nimport json\nfrom .tokenizer import Tokenizer\nfrom . import LLM\nfrom global_configs import tokenizer_path\n\n\nclass MetaModel(nn.Module):\n    \"\"\" Masked Autoencoder with VisionTransformer backbone\n    \"\"\"\n    def __init__(self, llama_type, reversible_grad: bool, llama_config):\n        super().__init__()\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        ModelArgs = LLM.__dict__[llama_type].ModelArgs\n        Transformer = LLM.__dict__[llama_type].Transformer\n\n        with open(llama_config, \"r\") as f:\n            params = json.loads(f.read())\n        model_args: ModelArgs = ModelArgs(\n            max_seq_len=2048, max_batch_size=32, **params\n        )\n        tokenizer = Tokenizer(model_path=tokenizer_path)\n        model_args.vocab_size = tokenizer.n_words\n        if reversible_grad:\n            if hasattr(model_args, \"reversible_gradient\"):\n                model_args.reversible_gradient = True\n            else:\n                raise KeyError (f\"{ModelArgs} object has no attribute reversible_gradient\")\n\n        model = Transformer(model_args)\n        self.llma = model\n        for name, param in self.named_parameters():\n            if param.requires_grad:\n               print(f\"Trainable param: {name}, {param.shape}, {param.dtype}\")\n        count = sum(p.numel() for p in self.parameters() if p.requires_grad)\n        print(f\"Parameter count : {count}\")\n\n    def forward(self, examples, labels):\n        output = self.llma(examples)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n           c_loss = output.mean() * 0\n        else:\n           c_loss = self.criterion(output.reshape(-1, 32000), labels.flatten())\n        pred = 0\n        mask = 0\n        return c_loss, c_loss, pred, mask","source_hash":"5ae0a461056ec2a2ee50a5c7efde1e77ef8c57d3751d844da3bc8629b25863a8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.meta.MetaModel","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.meta.MetaModel#L9-L52","kind":"class","name":"MetaModel","path":"gorilla/finetune/model/meta.py","language":"python","start_line":9,"end_line":52,"context_start_line":1,"context_end_line":52,"code":"import torch\nimport torch.nn as nn\nimport json\nfrom .tokenizer import Tokenizer\nfrom . import LLM\nfrom global_configs import tokenizer_path\n\n\nclass MetaModel(nn.Module):\n    \"\"\" Masked Autoencoder with VisionTransformer backbone\n    \"\"\"\n    def __init__(self, llama_type, reversible_grad: bool, llama_config):\n        super().__init__()\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        ModelArgs = LLM.__dict__[llama_type].ModelArgs\n        Transformer = LLM.__dict__[llama_type].Transformer\n\n        with open(llama_config, \"r\") as f:\n            params = json.loads(f.read())\n        model_args: ModelArgs = ModelArgs(\n            max_seq_len=2048, max_batch_size=32, **params\n        )\n        tokenizer = Tokenizer(model_path=tokenizer_path)\n        model_args.vocab_size = tokenizer.n_words\n        if reversible_grad:\n            if hasattr(model_args, \"reversible_gradient\"):\n                model_args.reversible_gradient = True\n            else:\n                raise KeyError (f\"{ModelArgs} object has no attribute reversible_gradient\")\n\n        model = Transformer(model_args)\n        self.llma = model\n        for name, param in self.named_parameters():\n            if param.requires_grad:\n               print(f\"Trainable param: {name}, {param.shape}, {param.dtype}\")\n        count = sum(p.numel() for p in self.parameters() if p.requires_grad)\n        print(f\"Parameter count : {count}\")\n\n    def forward(self, examples, labels):\n        output = self.llma(examples)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n           c_loss = output.mean() * 0\n        else:\n           c_loss = self.criterion(output.reshape(-1, 32000), labels.flatten())\n        pred = 0\n        mask = 0\n        return c_loss, c_loss, pred, mask","source_hash":"5ae0a461056ec2a2ee50a5c7efde1e77ef8c57d3751d844da3bc8629b25863a8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.meta.__init__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.meta.__init__#L12-L39","kind":"function","name":"__init__","path":"gorilla/finetune/model/meta.py","language":"python","start_line":12,"end_line":39,"context_start_line":1,"context_end_line":52,"code":"import torch\nimport torch.nn as nn\nimport json\nfrom .tokenizer import Tokenizer\nfrom . import LLM\nfrom global_configs import tokenizer_path\n\n\nclass MetaModel(nn.Module):\n    \"\"\" Masked Autoencoder with VisionTransformer backbone\n    \"\"\"\n    def __init__(self, llama_type, reversible_grad: bool, llama_config):\n        super().__init__()\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        ModelArgs = LLM.__dict__[llama_type].ModelArgs\n        Transformer = LLM.__dict__[llama_type].Transformer\n\n        with open(llama_config, \"r\") as f:\n            params = json.loads(f.read())\n        model_args: ModelArgs = ModelArgs(\n            max_seq_len=2048, max_batch_size=32, **params\n        )\n        tokenizer = Tokenizer(model_path=tokenizer_path)\n        model_args.vocab_size = tokenizer.n_words\n        if reversible_grad:\n            if hasattr(model_args, \"reversible_gradient\"):\n                model_args.reversible_gradient = True\n            else:\n                raise KeyError (f\"{ModelArgs} object has no attribute reversible_gradient\")\n\n        model = Transformer(model_args)\n        self.llma = model\n        for name, param in self.named_parameters():\n            if param.requires_grad:\n               print(f\"Trainable param: {name}, {param.shape}, {param.dtype}\")\n        count = sum(p.numel() for p in self.parameters() if p.requires_grad)\n        print(f\"Parameter count : {count}\")\n\n    def forward(self, examples, labels):\n        output = self.llma(examples)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n           c_loss = output.mean() * 0\n        else:\n           c_loss = self.criterion(output.reshape(-1, 32000), labels.flatten())\n        pred = 0\n        mask = 0\n        return c_loss, c_loss, pred, mask","source_hash":"5ae0a461056ec2a2ee50a5c7efde1e77ef8c57d3751d844da3bc8629b25863a8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.meta.forward","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.meta.forward#L41-L52","kind":"function","name":"forward","path":"gorilla/finetune/model/meta.py","language":"python","start_line":41,"end_line":52,"context_start_line":21,"context_end_line":52,"code":"            params = json.loads(f.read())\n        model_args: ModelArgs = ModelArgs(\n            max_seq_len=2048, max_batch_size=32, **params\n        )\n        tokenizer = Tokenizer(model_path=tokenizer_path)\n        model_args.vocab_size = tokenizer.n_words\n        if reversible_grad:\n            if hasattr(model_args, \"reversible_gradient\"):\n                model_args.reversible_gradient = True\n            else:\n                raise KeyError (f\"{ModelArgs} object has no attribute reversible_gradient\")\n\n        model = Transformer(model_args)\n        self.llma = model\n        for name, param in self.named_parameters():\n            if param.requires_grad:\n               print(f\"Trainable param: {name}, {param.shape}, {param.dtype}\")\n        count = sum(p.numel() for p in self.parameters() if p.requires_grad)\n        print(f\"Parameter count : {count}\")\n\n    def forward(self, examples, labels):\n        output = self.llma(examples)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n           c_loss = output.mean() * 0\n        else:\n           c_loss = self.criterion(output.reshape(-1, 32000), labels.flatten())\n        pred = 0\n        mask = 0\n        return c_loss, c_loss, pred, mask","source_hash":"5ae0a461056ec2a2ee50a5c7efde1e77ef8c57d3751d844da3bc8629b25863a8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.tokenizer","uri":"program://LLaMA-Adapter/module/gorilla.finetune.model.tokenizer#L1-L40","kind":"module","name":"gorilla.finetune.model.tokenizer","path":"gorilla/finetune/model/tokenizer.py","language":"python","start_line":1,"end_line":40,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.tokenizer.Tokenizer","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.tokenizer.Tokenizer#L13-L40","kind":"class","name":"Tokenizer","path":"gorilla/finetune/model/tokenizer.py","language":"python","start_line":13,"end_line":40,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.tokenizer.__init__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.tokenizer.__init__#L14-L28","kind":"function","name":"__init__","path":"gorilla/finetune/model/tokenizer.py","language":"python","start_line":14,"end_line":28,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.tokenizer.encode","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.tokenizer.encode#L30-L37","kind":"function","name":"encode","path":"gorilla/finetune/model/tokenizer.py","language":"python","start_line":30,"end_line":37,"context_start_line":10,"context_end_line":40,"code":"logger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.tokenizer.decode","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.tokenizer.decode#L39-L40","kind":"function","name":"decode","path":"gorilla/finetune/model/tokenizer.py","language":"python","start_line":39,"end_line":40,"context_start_line":19,"context_end_line":40,"code":"\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama","uri":"program://LLaMA-Adapter/module/gorilla.finetune.model.LLM.revllama#L1-L354","kind":"module","name":"gorilla.finetune.model.LLM.revllama","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":1,"end_line":354,"context_start_line":1,"context_end_line":354,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\nimport copy\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\nimport sys\n\nimport torch\nfrom torch import nn\nimport torch.nn.functional as F\nfrom torch.nn import Embedding, Linear\nfrom torch.autograd import Function\n\nfrom apex.normalization import FusedRMSNorm as RMSNorm\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    reversible_gradient: bool = False\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wk = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wv = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wo = Linear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=False\n        )\n\n        self.flash = True\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None\n    ):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        keys = xk\n        values = xv\n\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        if self.flash:\n           output = F.scaled_dot_product_attention(xq, keys, values, attn_mask=None, dropout_p=0.0, is_causal=True)\n        else:\n           scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n           if mask is not None:\n              scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n           scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n           output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(\n                   1, 2\n                   ).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=False\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=False\n        )\n        self.w3 = Linear(\n            dim, hidden_dim, bias=False\n        )\n\n    # @torch.compile\n    def _silu_gating(self, x, y):\n        return F.silu(x) * y\n\n    def forward(self, x):\n        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.seeds = {}\n\n    def set_seed(self, key: str):\n        \"\"\"\n        For activation recompute\n        \"\"\"\n        if hasattr(torch.cuda, \"default_generators\") and len(torch.cuda.default_generators) > 0:\n            device_idx = torch.cuda.current_device()\n            seed = torch.cuda.default_generators[device_idx].seed()\n        else:\n            seed = int(torch.seed() % sys.maxsize)\n        self.seeds[key] = seed\n        torch.manual_seed(seed)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None\n    ):\n        x1, x2 = torch.chunk(x, 2, dim=-1)\n        if self.training:\n            self.set_seed(\"F\")  # seed for the F function\n        f_x2 = self.f_forward(x2, start_pos, freqs_cis, mask, prompt)\n        y1 = x1 + f_x2\n\n        if self.training:\n            self.set_seed(\"G\") # seed for the G function\n        g_y1 = self.g_forward(y1)\n        y2 = x2 + g_y1\n        return torch.cat([y1, y2], dim=-1)\n\n    def f_forward(self, x, start_pos, freqs_cis, mask, prompt):\n        return self.attention(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def g_forward(self, x):\n        return self.feed_forward(self.ffn_norm(x))\n\n    def backward_pass(\n        self,\n        y,\n        dy,\n        start_pos,\n        freqs_cis,\n        mask,\n        prompt\n    ):\n        assert self.training, (\n            \"If you want to train ReversibleModel, make sure to put the model into training mode.\"\n        )\n        y1, y2 = torch.chunk(y, 2, dim=-1)\n        dy1, dy2 = torch.chunk(dy, 2, dim=-1)\n        with torch.enable_grad():\n            y1.requires_grad = True\n            torch.manual_seed(self.seeds[\"G\"])\n            g_y1 = self.g_forward(y1)\n            g_y1.backward(dy2)\n\n        with torch.no_grad():\n            x2 = y2 - g_y1\n            del g_y1\n            dy1 = dy1 + y1.grad\n            y1.grad = None\n\n        with torch.enable_grad():\n            x2.requires_grad = True\n            torch.manual_seed(self.seeds[\"F\"])\n            f_x2 = self.f_forward(x2, start_pos, freqs_cis, mask, prompt)\n            f_x2.backward(dy1)\n\n        with torch.no_grad():\n            x1 = y1 - f_x2\n            del f_x2, y1\n            dy2 = dy2 + x2.grad\n            x2.grad = None\n            x2 = x2.detach()\n\n        return torch.cat([x1, x2], dim=-1), torch.cat([dy1, dy2], dim=-1)\n\n\nclass RevBackProp(Function):\n    @staticmethod\n    def forward(\n        ctx,\n        x,\n        layers,\n        start_pos,\n        freqs_cis,\n        mask,\n        prompt\n    ):\n        with torch.no_grad():\n            for layer in layers:\n                x = layer(x.detach(), start_pos, freqs_cis, mask, prompt)\n\n        ctx.save_for_backward(x.detach())\n        ctx.layers = layers\n        ctx.start_pos = start_pos\n        ctx.freqs_cis = freqs_cis\n        ctx.mask = mask\n        ctx.prompt = prompt\n        return x\n\n    @staticmethod\n    def backward(ctx, dy):\n        y, = ctx.saved_tensors\n        for layer in ctx.layers[::-1]:\n            y, dy = layer.backward_pass(\n                y,\n                dy,\n                ctx.start_pos,\n                ctx.freqs_cis,\n                ctx.mask,\n                ctx.prompt\n            )\n        return dy, None, None, None, None, None\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n        self.reversible_gradient = params.reversible_gradient # If false, use vanilla gradient\n\n    @staticmethod\n    def vanilla_forward(h, layers, start_pos, freqs_cis, mask, prompt):\n        for _, layer in enumerate(layers):\n            h = layer(h, start_pos, freqs_cis, mask, prompt)\n        return h\n\n    def forward(self, examples):\n        _bsz, seqlen = examples.shape\n        h = self.tok_embeddings(examples)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[:seqlen]\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        prompt = None\n        h = torch.cat([h, h], dim=-1)\n\n        if not self.training or not self.reversible_gradient:\n            executing_fn = Transformer.vanilla_forward\n        else:\n            executing_fn = RevBackProp.apply\n        h = executing_fn(h, self.layers, start_pos, freqs_cis, mask, prompt)\n\n        h1, h2 = torch.chunk(h, 2, dim=-1)\n        h = (h1 + h2) / 2.\n        h = self.norm(h)\n        output = self.output(h)\n        return output\n\n\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        h = torch.cat([h, h], dim=-1)\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h1, h2 = torch.chunk(h, 2, dim=-1)\n        h = (h1 + h2) / 2.\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.ModelArgs","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.revllama.ModelArgs#L20-L31","kind":"class","name":"ModelArgs","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":20,"end_line":31,"context_start_line":1,"context_end_line":51,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\nimport copy\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\nimport sys\n\nimport torch\nfrom torch import nn\nimport torch.nn.functional as F\nfrom torch.nn import Embedding, Linear\nfrom torch.autograd import Function\n\nfrom apex.normalization import FusedRMSNorm as RMSNorm\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    reversible_gradient: bool = False\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.precompute_freqs_cis","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.precompute_freqs_cis#L34-L39","kind":"function","name":"precompute_freqs_cis","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":34,"end_line":39,"context_start_line":14,"context_end_line":59,"code":"from torch.autograd import Function\n\nfrom apex.normalization import FusedRMSNorm as RMSNorm\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    reversible_gradient: bool = False\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.reshape_for_broadcast","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.reshape_for_broadcast#L42-L47","kind":"function","name":"reshape_for_broadcast","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":42,"end_line":47,"context_start_line":22,"context_end_line":67,"code":"    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    reversible_gradient: bool = False\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.apply_rotary_emb","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.apply_rotary_emb#L50-L60","kind":"function","name":"apply_rotary_emb","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":50,"end_line":60,"context_start_line":30,"context_end_line":80,"code":"\n    reversible_gradient: bool = False\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wk = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wv = Linear(","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.Attention","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.revllama.Attention#L63-L124","kind":"class","name":"Attention","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":63,"end_line":124,"context_start_line":43,"context_end_line":144,"code":"    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wk = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wv = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wo = Linear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=False\n        )\n\n        self.flash = True\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None\n    ):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        keys = xk\n        values = xv\n\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        if self.flash:\n           output = F.scaled_dot_product_attention(xq, keys, values, attn_mask=None, dropout_p=0.0, is_causal=True)\n        else:\n           scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n           if mask is not None:\n              scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n           scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n           output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(\n                   1, 2\n                   ).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=False\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=False\n        )\n        self.w3 = Linear(","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.FeedForward","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.revllama.FeedForward#L127-L153","kind":"class","name":"FeedForward","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":127,"end_line":153,"context_start_line":107,"context_end_line":173,"code":"\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        if self.flash:\n           output = F.scaled_dot_product_attention(xq, keys, values, attn_mask=None, dropout_p=0.0, is_causal=True)\n        else:\n           scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n           if mask is not None:\n              scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n           scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n           output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(\n                   1, 2\n                   ).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=False\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=False\n        )\n        self.w3 = Linear(\n            dim, hidden_dim, bias=False\n        )\n\n    # @torch.compile\n    def _silu_gating(self, x, y):\n        return F.silu(x) * y\n\n    def forward(self, x):\n        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.seeds = {}\n\n    def set_seed(self, key: str):\n        \"\"\"\n        For activation recompute","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.TransformerBlock","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.revllama.TransformerBlock#L156-L243","kind":"class","name":"TransformerBlock","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":156,"end_line":243,"context_start_line":136,"context_end_line":263,"code":"        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=False\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=False\n        )\n        self.w3 = Linear(\n            dim, hidden_dim, bias=False\n        )\n\n    # @torch.compile\n    def _silu_gating(self, x, y):\n        return F.silu(x) * y\n\n    def forward(self, x):\n        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.seeds = {}\n\n    def set_seed(self, key: str):\n        \"\"\"\n        For activation recompute\n        \"\"\"\n        if hasattr(torch.cuda, \"default_generators\") and len(torch.cuda.default_generators) > 0:\n            device_idx = torch.cuda.current_device()\n            seed = torch.cuda.default_generators[device_idx].seed()\n        else:\n            seed = int(torch.seed() % sys.maxsize)\n        self.seeds[key] = seed\n        torch.manual_seed(seed)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None\n    ):\n        x1, x2 = torch.chunk(x, 2, dim=-1)\n        if self.training:\n            self.set_seed(\"F\")  # seed for the F function\n        f_x2 = self.f_forward(x2, start_pos, freqs_cis, mask, prompt)\n        y1 = x1 + f_x2\n\n        if self.training:\n            self.set_seed(\"G\") # seed for the G function\n        g_y1 = self.g_forward(y1)\n        y2 = x2 + g_y1\n        return torch.cat([y1, y2], dim=-1)\n\n    def f_forward(self, x, start_pos, freqs_cis, mask, prompt):\n        return self.attention(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def g_forward(self, x):\n        return self.feed_forward(self.ffn_norm(x))\n\n    def backward_pass(\n        self,\n        y,\n        dy,\n        start_pos,\n        freqs_cis,\n        mask,\n        prompt\n    ):\n        assert self.training, (\n            \"If you want to train ReversibleModel, make sure to put the model into training mode.\"\n        )\n        y1, y2 = torch.chunk(y, 2, dim=-1)\n        dy1, dy2 = torch.chunk(dy, 2, dim=-1)\n        with torch.enable_grad():\n            y1.requires_grad = True\n            torch.manual_seed(self.seeds[\"G\"])\n            g_y1 = self.g_forward(y1)\n            g_y1.backward(dy2)\n\n        with torch.no_grad():\n            x2 = y2 - g_y1\n            del g_y1\n            dy1 = dy1 + y1.grad\n            y1.grad = None\n\n        with torch.enable_grad():\n            x2.requires_grad = True\n            torch.manual_seed(self.seeds[\"F\"])\n            f_x2 = self.f_forward(x2, start_pos, freqs_cis, mask, prompt)\n            f_x2.backward(dy1)\n\n        with torch.no_grad():\n            x1 = y1 - f_x2\n            del f_x2, y1\n            dy2 = dy2 + x2.grad\n            x2.grad = None\n            x2 = x2.detach()\n\n        return torch.cat([x1, x2], dim=-1), torch.cat([dy1, dy2], dim=-1)\n\n\nclass RevBackProp(Function):\n    @staticmethod\n    def forward(\n        ctx,\n        x,\n        layers,\n        start_pos,\n        freqs_cis,\n        mask,\n        prompt\n    ):\n        with torch.no_grad():\n            for layer in layers:\n                x = layer(x.detach(), start_pos, freqs_cis, mask, prompt)\n\n        ctx.save_for_backward(x.detach())\n        ctx.layers = layers\n        ctx.start_pos = start_pos","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.RevBackProp","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.revllama.RevBackProp#L246-L281","kind":"class","name":"RevBackProp","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":246,"end_line":281,"context_start_line":226,"context_end_line":301,"code":"            del g_y1\n            dy1 = dy1 + y1.grad\n            y1.grad = None\n\n        with torch.enable_grad():\n            x2.requires_grad = True\n            torch.manual_seed(self.seeds[\"F\"])\n            f_x2 = self.f_forward(x2, start_pos, freqs_cis, mask, prompt)\n            f_x2.backward(dy1)\n\n        with torch.no_grad():\n            x1 = y1 - f_x2\n            del f_x2, y1\n            dy2 = dy2 + x2.grad\n            x2.grad = None\n            x2 = x2.detach()\n\n        return torch.cat([x1, x2], dim=-1), torch.cat([dy1, dy2], dim=-1)\n\n\nclass RevBackProp(Function):\n    @staticmethod\n    def forward(\n        ctx,\n        x,\n        layers,\n        start_pos,\n        freqs_cis,\n        mask,\n        prompt\n    ):\n        with torch.no_grad():\n            for layer in layers:\n                x = layer(x.detach(), start_pos, freqs_cis, mask, prompt)\n\n        ctx.save_for_backward(x.detach())\n        ctx.layers = layers\n        ctx.start_pos = start_pos\n        ctx.freqs_cis = freqs_cis\n        ctx.mask = mask\n        ctx.prompt = prompt\n        return x\n\n    @staticmethod\n    def backward(ctx, dy):\n        y, = ctx.saved_tensors\n        for layer in ctx.layers[::-1]:\n            y, dy = layer.backward_pass(\n                y,\n                dy,\n                ctx.start_pos,\n                ctx.freqs_cis,\n                ctx.mask,\n                ctx.prompt\n            )\n        return dy, None, None, None, None, None\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.Transformer","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.revllama.Transformer#L284-L354","kind":"class","name":"Transformer","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":284,"end_line":354,"context_start_line":264,"context_end_line":354,"code":"        ctx.freqs_cis = freqs_cis\n        ctx.mask = mask\n        ctx.prompt = prompt\n        return x\n\n    @staticmethod\n    def backward(ctx, dy):\n        y, = ctx.saved_tensors\n        for layer in ctx.layers[::-1]:\n            y, dy = layer.backward_pass(\n                y,\n                dy,\n                ctx.start_pos,\n                ctx.freqs_cis,\n                ctx.mask,\n                ctx.prompt\n            )\n        return dy, None, None, None, None, None\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n        self.reversible_gradient = params.reversible_gradient # If false, use vanilla gradient\n\n    @staticmethod\n    def vanilla_forward(h, layers, start_pos, freqs_cis, mask, prompt):\n        for _, layer in enumerate(layers):\n            h = layer(h, start_pos, freqs_cis, mask, prompt)\n        return h\n\n    def forward(self, examples):\n        _bsz, seqlen = examples.shape\n        h = self.tok_embeddings(examples)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[:seqlen]\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        prompt = None\n        h = torch.cat([h, h], dim=-1)\n\n        if not self.training or not self.reversible_gradient:\n            executing_fn = Transformer.vanilla_forward\n        else:\n            executing_fn = RevBackProp.apply\n        h = executing_fn(h, self.layers, start_pos, freqs_cis, mask, prompt)\n\n        h1, h2 = torch.chunk(h, 2, dim=-1)\n        h = (h1 + h2) / 2.\n        h = self.norm(h)\n        output = self.output(h)\n        return output\n\n\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        h = torch.cat([h, h], dim=-1)\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h1, h2 = torch.chunk(h, 2, dim=-1)\n        h = (h1 + h2) / 2.\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.__init__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.__init__#L285-L302","kind":"function","name":"__init__","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":285,"end_line":302,"context_start_line":265,"context_end_line":322,"code":"        ctx.mask = mask\n        ctx.prompt = prompt\n        return x\n\n    @staticmethod\n    def backward(ctx, dy):\n        y, = ctx.saved_tensors\n        for layer in ctx.layers[::-1]:\n            y, dy = layer.backward_pass(\n                y,\n                dy,\n                ctx.start_pos,\n                ctx.freqs_cis,\n                ctx.mask,\n                ctx.prompt\n            )\n        return dy, None, None, None, None, None\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n        self.reversible_gradient = params.reversible_gradient # If false, use vanilla gradient\n\n    @staticmethod\n    def vanilla_forward(h, layers, start_pos, freqs_cis, mask, prompt):\n        for _, layer in enumerate(layers):\n            h = layer(h, start_pos, freqs_cis, mask, prompt)\n        return h\n\n    def forward(self, examples):\n        _bsz, seqlen = examples.shape\n        h = self.tok_embeddings(examples)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[:seqlen]\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        prompt = None\n        h = torch.cat([h, h], dim=-1)\n\n        if not self.training or not self.reversible_gradient:\n            executing_fn = Transformer.vanilla_forward","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.forward","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.forward#L310-L331","kind":"function","name":"forward","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":310,"end_line":331,"context_start_line":290,"context_end_line":351,"code":"        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n        self.reversible_gradient = params.reversible_gradient # If false, use vanilla gradient\n\n    @staticmethod\n    def vanilla_forward(h, layers, start_pos, freqs_cis, mask, prompt):\n        for _, layer in enumerate(layers):\n            h = layer(h, start_pos, freqs_cis, mask, prompt)\n        return h\n\n    def forward(self, examples):\n        _bsz, seqlen = examples.shape\n        h = self.tok_embeddings(examples)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[:seqlen]\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        prompt = None\n        h = torch.cat([h, h], dim=-1)\n\n        if not self.training or not self.reversible_gradient:\n            executing_fn = Transformer.vanilla_forward\n        else:\n            executing_fn = RevBackProp.apply\n        h = executing_fn(h, self.layers, start_pos, freqs_cis, mask, prompt)\n\n        h1, h2 = torch.chunk(h, 2, dim=-1)\n        h = (h1 + h2) / 2.\n        h = self.norm(h)\n        output = self.output(h)\n        return output\n\n\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        h = torch.cat([h, h], dim=-1)\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h1, h2 = torch.chunk(h, 2, dim=-1)\n        h = (h1 + h2) / 2.","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama._silu_gating","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama._silu_gating#L149-L150","kind":"function","name":"_silu_gating","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":149,"end_line":150,"context_start_line":129,"context_end_line":170,"code":"        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=False\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=False\n        )\n        self.w3 = Linear(\n            dim, hidden_dim, bias=False\n        )\n\n    # @torch.compile\n    def _silu_gating(self, x, y):\n        return F.silu(x) * y\n\n    def forward(self, x):\n        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.seeds = {}\n","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.set_seed","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.set_seed#L171-L181","kind":"function","name":"set_seed","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":171,"end_line":181,"context_start_line":151,"context_end_line":201,"code":"\n    def forward(self, x):\n        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.seeds = {}\n\n    def set_seed(self, key: str):\n        \"\"\"\n        For activation recompute\n        \"\"\"\n        if hasattr(torch.cuda, \"default_generators\") and len(torch.cuda.default_generators) > 0:\n            device_idx = torch.cuda.current_device()\n            seed = torch.cuda.default_generators[device_idx].seed()\n        else:\n            seed = int(torch.seed() % sys.maxsize)\n        self.seeds[key] = seed\n        torch.manual_seed(seed)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None\n    ):\n        x1, x2 = torch.chunk(x, 2, dim=-1)\n        if self.training:\n            self.set_seed(\"F\")  # seed for the F function\n        f_x2 = self.f_forward(x2, start_pos, freqs_cis, mask, prompt)\n        y1 = x1 + f_x2\n\n        if self.training:\n            self.set_seed(\"G\") # seed for the G function\n        g_y1 = self.g_forward(y1)\n        y2 = x2 + g_y1\n        return torch.cat([y1, y2], dim=-1)\n\n    def f_forward(self, x, start_pos, freqs_cis, mask, prompt):\n        return self.attention(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def g_forward(self, x):","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.f_forward","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.f_forward#L198-L199","kind":"function","name":"f_forward","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":198,"end_line":199,"context_start_line":178,"context_end_line":219,"code":"        else:\n            seed = int(torch.seed() % sys.maxsize)\n        self.seeds[key] = seed\n        torch.manual_seed(seed)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None\n    ):\n        x1, x2 = torch.chunk(x, 2, dim=-1)\n        if self.training:\n            self.set_seed(\"F\")  # seed for the F function\n        f_x2 = self.f_forward(x2, start_pos, freqs_cis, mask, prompt)\n        y1 = x1 + f_x2\n\n        if self.training:\n            self.set_seed(\"G\") # seed for the G function\n        g_y1 = self.g_forward(y1)\n        y2 = x2 + g_y1\n        return torch.cat([y1, y2], dim=-1)\n\n    def f_forward(self, x, start_pos, freqs_cis, mask, prompt):\n        return self.attention(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def g_forward(self, x):\n        return self.feed_forward(self.ffn_norm(x))\n\n    def backward_pass(\n        self,\n        y,\n        dy,\n        start_pos,\n        freqs_cis,\n        mask,\n        prompt\n    ):\n        assert self.training, (\n            \"If you want to train ReversibleModel, make sure to put the model into training mode.\"\n        )\n        y1, y2 = torch.chunk(y, 2, dim=-1)\n        dy1, dy2 = torch.chunk(dy, 2, dim=-1)\n        with torch.enable_grad():\n            y1.requires_grad = True","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.g_forward","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.g_forward#L201-L202","kind":"function","name":"g_forward","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":201,"end_line":202,"context_start_line":181,"context_end_line":222,"code":"        torch.manual_seed(seed)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None\n    ):\n        x1, x2 = torch.chunk(x, 2, dim=-1)\n        if self.training:\n            self.set_seed(\"F\")  # seed for the F function\n        f_x2 = self.f_forward(x2, start_pos, freqs_cis, mask, prompt)\n        y1 = x1 + f_x2\n\n        if self.training:\n            self.set_seed(\"G\") # seed for the G function\n        g_y1 = self.g_forward(y1)\n        y2 = x2 + g_y1\n        return torch.cat([y1, y2], dim=-1)\n\n    def f_forward(self, x, start_pos, freqs_cis, mask, prompt):\n        return self.attention(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def g_forward(self, x):\n        return self.feed_forward(self.ffn_norm(x))\n\n    def backward_pass(\n        self,\n        y,\n        dy,\n        start_pos,\n        freqs_cis,\n        mask,\n        prompt\n    ):\n        assert self.training, (\n            \"If you want to train ReversibleModel, make sure to put the model into training mode.\"\n        )\n        y1, y2 = torch.chunk(y, 2, dim=-1)\n        dy1, dy2 = torch.chunk(dy, 2, dim=-1)\n        with torch.enable_grad():\n            y1.requires_grad = True\n            torch.manual_seed(self.seeds[\"G\"])\n            g_y1 = self.g_forward(y1)\n            g_y1.backward(dy2)","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.backward_pass","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.backward_pass#L204-L243","kind":"function","name":"backward_pass","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":204,"end_line":243,"context_start_line":184,"context_end_line":263,"code":"        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None\n    ):\n        x1, x2 = torch.chunk(x, 2, dim=-1)\n        if self.training:\n            self.set_seed(\"F\")  # seed for the F function\n        f_x2 = self.f_forward(x2, start_pos, freqs_cis, mask, prompt)\n        y1 = x1 + f_x2\n\n        if self.training:\n            self.set_seed(\"G\") # seed for the G function\n        g_y1 = self.g_forward(y1)\n        y2 = x2 + g_y1\n        return torch.cat([y1, y2], dim=-1)\n\n    def f_forward(self, x, start_pos, freqs_cis, mask, prompt):\n        return self.attention(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def g_forward(self, x):\n        return self.feed_forward(self.ffn_norm(x))\n\n    def backward_pass(\n        self,\n        y,\n        dy,\n        start_pos,\n        freqs_cis,\n        mask,\n        prompt\n    ):\n        assert self.training, (\n            \"If you want to train ReversibleModel, make sure to put the model into training mode.\"\n        )\n        y1, y2 = torch.chunk(y, 2, dim=-1)\n        dy1, dy2 = torch.chunk(dy, 2, dim=-1)\n        with torch.enable_grad():\n            y1.requires_grad = True\n            torch.manual_seed(self.seeds[\"G\"])\n            g_y1 = self.g_forward(y1)\n            g_y1.backward(dy2)\n\n        with torch.no_grad():\n            x2 = y2 - g_y1\n            del g_y1\n            dy1 = dy1 + y1.grad\n            y1.grad = None\n\n        with torch.enable_grad():\n            x2.requires_grad = True\n            torch.manual_seed(self.seeds[\"F\"])\n            f_x2 = self.f_forward(x2, start_pos, freqs_cis, mask, prompt)\n            f_x2.backward(dy1)\n\n        with torch.no_grad():\n            x1 = y1 - f_x2\n            del f_x2, y1\n            dy2 = dy2 + x2.grad\n            x2.grad = None\n            x2 = x2.detach()\n\n        return torch.cat([x1, x2], dim=-1), torch.cat([dy1, dy2], dim=-1)\n\n\nclass RevBackProp(Function):\n    @staticmethod\n    def forward(\n        ctx,\n        x,\n        layers,\n        start_pos,\n        freqs_cis,\n        mask,\n        prompt\n    ):\n        with torch.no_grad():\n            for layer in layers:\n                x = layer(x.detach(), start_pos, freqs_cis, mask, prompt)\n\n        ctx.save_for_backward(x.detach())\n        ctx.layers = layers\n        ctx.start_pos = start_pos","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.backward","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.backward#L270-L281","kind":"function","name":"backward","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":270,"end_line":281,"context_start_line":250,"context_end_line":301,"code":"        x,\n        layers,\n        start_pos,\n        freqs_cis,\n        mask,\n        prompt\n    ):\n        with torch.no_grad():\n            for layer in layers:\n                x = layer(x.detach(), start_pos, freqs_cis, mask, prompt)\n\n        ctx.save_for_backward(x.detach())\n        ctx.layers = layers\n        ctx.start_pos = start_pos\n        ctx.freqs_cis = freqs_cis\n        ctx.mask = mask\n        ctx.prompt = prompt\n        return x\n\n    @staticmethod\n    def backward(ctx, dy):\n        y, = ctx.saved_tensors\n        for layer in ctx.layers[::-1]:\n            y, dy = layer.backward_pass(\n                y,\n                dy,\n                ctx.start_pos,\n                ctx.freqs_cis,\n                ctx.mask,\n                ctx.prompt\n            )\n        return dy, None, None, None, None, None\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.vanilla_forward","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.vanilla_forward#L305-L308","kind":"function","name":"vanilla_forward","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":305,"end_line":308,"context_start_line":285,"context_end_line":328,"code":"    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n        self.reversible_gradient = params.reversible_gradient # If false, use vanilla gradient\n\n    @staticmethod\n    def vanilla_forward(h, layers, start_pos, freqs_cis, mask, prompt):\n        for _, layer in enumerate(layers):\n            h = layer(h, start_pos, freqs_cis, mask, prompt)\n        return h\n\n    def forward(self, examples):\n        _bsz, seqlen = examples.shape\n        h = self.tok_embeddings(examples)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[:seqlen]\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        prompt = None\n        h = torch.cat([h, h], dim=-1)\n\n        if not self.training or not self.reversible_gradient:\n            executing_fn = Transformer.vanilla_forward\n        else:\n            executing_fn = RevBackProp.apply\n        h = executing_fn(h, self.layers, start_pos, freqs_cis, mask, prompt)\n\n        h1, h2 = torch.chunk(h, 2, dim=-1)\n        h = (h1 + h2) / 2.","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.revllama.forward_inference","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.revllama.forward_inference#L336-L354","kind":"function","name":"forward_inference","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":336,"end_line":354,"context_start_line":316,"context_end_line":354,"code":"        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        prompt = None\n        h = torch.cat([h, h], dim=-1)\n\n        if not self.training or not self.reversible_gradient:\n            executing_fn = Transformer.vanilla_forward\n        else:\n            executing_fn = RevBackProp.apply\n        h = executing_fn(h, self.layers, start_pos, freqs_cis, mask, prompt)\n\n        h1, h2 = torch.chunk(h, 2, dim=-1)\n        h = (h1 + h2) / 2.\n        h = self.norm(h)\n        output = self.output(h)\n        return output\n\n\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        h = torch.cat([h, h], dim=-1)\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h1, h2 = torch.chunk(h, 2, dim=-1)\n        h = (h1 + h2) / 2.\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama","uri":"program://LLaMA-Adapter/module/gorilla.finetune.model.LLM.llama#L1-L245","kind":"module","name":"gorilla.finetune.model.LLM.llama","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":1,"end_line":245,"context_start_line":1,"context_end_line":245,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\nimport functools\n\nimport torch\nfrom torch import nn\nimport torch.nn.functional as F\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nfrom fairscale.nn.model_parallel.layers import (\n    ParallelEmbedding,\n    RowParallelLinear,\n    ColumnParallelLinear,\n)\n\nfrom apex.normalization import FusedRMSNorm as RMSNorm\n\ndefault_linear_init = functools.partial(nn.init.kaiming_uniform_, a=math.sqrt(5))\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=default_linear_init,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=default_linear_init,\n        )\n        self.wv = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=default_linear_init,\n        )\n        self.wo = RowParallelLinear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=False,\n            input_is_parallel=True,\n            init_method=default_linear_init,\n        )\n\n        self.flash = True\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        keys = xk\n        values = xv\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        if self.flash:\n            output = F.scaled_dot_product_attention(xq, keys, values, attn_mask=None, dropout_p=0.0, is_causal=True)\n        else:\n            scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n            if mask is not None:\n                scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n            output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=default_linear_init,\n        )\n        self.w2 = RowParallelLinear(\n            hidden_dim, dim, bias=False, input_is_parallel=True, init_method=default_linear_init\n        )\n        self.w3 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=default_linear_init\n        )\n\n    # @torch.compile\n    def _silu_gating(self, x, y):\n        return F.silu(x) * y\n\n    def forward(self, x):\n        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def _forward_ffn(self, h):\n        return h + self.feed_forward(self.ffn_norm(h))\n\n    def _forward_attention(self, x, start_pos, freqs_cis, mask, prompt):\n        return x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n        h = self._forward_attention(x, start_pos, freqs_cis, mask, prompt)\n        out = self._forward_ffn(h)\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=nn.init.normal_,\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(\n            params.dim, params.vocab_size, bias=False, init_method=default_linear_init,\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n\n    def forward(self, examples):\n        _bsz, seqlen = examples.shape\n        h = self.tok_embeddings(examples)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[:seqlen]\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h)\n        return output\n\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.ModelArgs","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.llama.ModelArgs#L25-L34","kind":"class","name":"ModelArgs","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":25,"end_line":34,"context_start_line":5,"context_end_line":54,"code":"from dataclasses import dataclass\nimport math\nimport functools\n\nimport torch\nfrom torch import nn\nimport torch.nn.functional as F\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nfrom fairscale.nn.model_parallel.layers import (\n    ParallelEmbedding,\n    RowParallelLinear,\n    ColumnParallelLinear,\n)\n\nfrom apex.normalization import FusedRMSNorm as RMSNorm\n\ndefault_linear_init = functools.partial(nn.init.kaiming_uniform_, a=math.sqrt(5))\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.precompute_freqs_cis","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.llama.precompute_freqs_cis#L37-L42","kind":"function","name":"precompute_freqs_cis","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":37,"end_line":42,"context_start_line":17,"context_end_line":62,"code":"    ColumnParallelLinear,\n)\n\nfrom apex.normalization import FusedRMSNorm as RMSNorm\n\ndefault_linear_init = functools.partial(nn.init.kaiming_uniform_, a=math.sqrt(5))\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.reshape_for_broadcast","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.llama.reshape_for_broadcast#L45-L50","kind":"function","name":"reshape_for_broadcast","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":45,"end_line":50,"context_start_line":25,"context_end_line":70,"code":"class ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.apply_rotary_emb","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.llama.apply_rotary_emb#L53-L63","kind":"function","name":"apply_rotary_emb","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":53,"end_line":63,"context_start_line":33,"context_end_line":83,"code":"    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=default_linear_init,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.Attention","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.llama.Attention#L66-L132","kind":"class","name":"Attention","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":66,"end_line":132,"context_start_line":46,"context_end_line":152,"code":"    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=default_linear_init,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=default_linear_init,\n        )\n        self.wv = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=default_linear_init,\n        )\n        self.wo = RowParallelLinear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=False,\n            input_is_parallel=True,\n            init_method=default_linear_init,\n        )\n\n        self.flash = True\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        keys = xk\n        values = xv\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        if self.flash:\n            output = F.scaled_dot_product_attention(xq, keys, values, attn_mask=None, dropout_p=0.0, is_causal=True)\n        else:\n            scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n            if mask is not None:\n                scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n            output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=default_linear_init,\n        )\n        self.w2 = RowParallelLinear(\n            hidden_dim, dim, bias=False, input_is_parallel=True, init_method=default_linear_init\n        )\n        self.w3 = ColumnParallelLinear(","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.FeedForward","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.llama.FeedForward#L135-L161","kind":"class","name":"FeedForward","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":135,"end_line":161,"context_start_line":115,"context_end_line":181,"code":"        values = xv\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        if self.flash:\n            output = F.scaled_dot_product_attention(xq, keys, values, attn_mask=None, dropout_p=0.0, is_causal=True)\n        else:\n            scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n            if mask is not None:\n                scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n            output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=default_linear_init,\n        )\n        self.w2 = RowParallelLinear(\n            hidden_dim, dim, bias=False, input_is_parallel=True, init_method=default_linear_init\n        )\n        self.w3 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=default_linear_init\n        )\n\n    # @torch.compile\n    def _silu_gating(self, x, y):\n        return F.silu(x) * y\n\n    def forward(self, x):\n        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def _forward_ffn(self, h):\n        return h + self.feed_forward(self.ffn_norm(h))\n\n    def _forward_attention(self, x, start_pos, freqs_cis, mask, prompt):","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.TransformerBlock","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.llama.TransformerBlock#L164-L187","kind":"class","name":"TransformerBlock","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":164,"end_line":187,"context_start_line":144,"context_end_line":207,"code":"        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=default_linear_init,\n        )\n        self.w2 = RowParallelLinear(\n            hidden_dim, dim, bias=False, input_is_parallel=True, init_method=default_linear_init\n        )\n        self.w3 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=default_linear_init\n        )\n\n    # @torch.compile\n    def _silu_gating(self, x, y):\n        return F.silu(x) * y\n\n    def forward(self, x):\n        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def _forward_ffn(self, h):\n        return h + self.feed_forward(self.ffn_norm(h))\n\n    def _forward_attention(self, x, start_pos, freqs_cis, mask, prompt):\n        return x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n        h = self._forward_attention(x, start_pos, freqs_cis, mask, prompt)\n        out = self._forward_ffn(h)\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=nn.init.normal_,\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(\n            params.dim, params.vocab_size, bias=False, init_method=default_linear_init,\n        )","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.Transformer","uri":"program://LLaMA-Adapter/class/gorilla.finetune.model.LLM.llama.Transformer#L190-L245","kind":"class","name":"Transformer","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":190,"end_line":245,"context_start_line":170,"context_end_line":245,"code":"        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def _forward_ffn(self, h):\n        return h + self.feed_forward(self.ffn_norm(h))\n\n    def _forward_attention(self, x, start_pos, freqs_cis, mask, prompt):\n        return x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n        h = self._forward_attention(x, start_pos, freqs_cis, mask, prompt)\n        out = self._forward_ffn(h)\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=nn.init.normal_,\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(\n            params.dim, params.vocab_size, bias=False, init_method=default_linear_init,\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n\n    def forward(self, examples):\n        _bsz, seqlen = examples.shape\n        h = self.tok_embeddings(examples)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[:seqlen]\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h)\n        return output\n\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.__init__","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.llama.__init__#L191-L211","kind":"function","name":"__init__","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":191,"end_line":211,"context_start_line":171,"context_end_line":231,"code":"        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def _forward_ffn(self, h):\n        return h + self.feed_forward(self.ffn_norm(h))\n\n    def _forward_attention(self, x, start_pos, freqs_cis, mask, prompt):\n        return x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n        h = self._forward_attention(x, start_pos, freqs_cis, mask, prompt)\n        out = self._forward_ffn(h)\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=nn.init.normal_,\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(\n            params.dim, params.vocab_size, bias=False, init_method=default_linear_init,\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n\n    def forward(self, examples):\n        _bsz, seqlen = examples.shape\n        h = self.tok_embeddings(examples)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[:seqlen]\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h)\n        return output\n\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.forward","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.llama.forward#L214-L226","kind":"function","name":"forward","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":214,"end_line":226,"context_start_line":194,"context_end_line":245,"code":"        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=nn.init.normal_,\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(\n            params.dim, params.vocab_size, bias=False, init_method=default_linear_init,\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n\n    def forward(self, examples):\n        _bsz, seqlen = examples.shape\n        h = self.tok_embeddings(examples)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[:seqlen]\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h)\n        return output\n\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama._silu_gating","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.llama._silu_gating#L157-L158","kind":"function","name":"_silu_gating","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":157,"end_line":158,"context_start_line":137,"context_end_line":178,"code":"        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=default_linear_init,\n        )\n        self.w2 = RowParallelLinear(\n            hidden_dim, dim, bias=False, input_is_parallel=True, init_method=default_linear_init\n        )\n        self.w3 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=default_linear_init\n        )\n\n    # @torch.compile\n    def _silu_gating(self, x, y):\n        return F.silu(x) * y\n\n    def forward(self, x):\n        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def _forward_ffn(self, h):","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama._forward_ffn","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.llama._forward_ffn#L178-L179","kind":"function","name":"_forward_ffn","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":178,"end_line":179,"context_start_line":158,"context_end_line":199,"code":"        return F.silu(x) * y\n\n    def forward(self, x):\n        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def _forward_ffn(self, h):\n        return h + self.feed_forward(self.ffn_norm(h))\n\n    def _forward_attention(self, x, start_pos, freqs_cis, mask, prompt):\n        return x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n        h = self._forward_attention(x, start_pos, freqs_cis, mask, prompt)\n        out = self._forward_ffn(h)\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=nn.init.normal_,\n        )\n","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama._forward_attention","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.llama._forward_attention#L181-L182","kind":"function","name":"_forward_attention","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":181,"end_line":182,"context_start_line":161,"context_end_line":202,"code":"        return self.w2(self._silu_gating(self.w1(x), self.w3(x)))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def _forward_ffn(self, h):\n        return h + self.feed_forward(self.ffn_norm(h))\n\n    def _forward_attention(self, x, start_pos, freqs_cis, mask, prompt):\n        return x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n        h = self._forward_attention(x, start_pos, freqs_cis, mask, prompt)\n        out = self._forward_ffn(h)\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=nn.init.normal_,\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.model.LLM.llama.forward_inference","uri":"program://LLaMA-Adapter/function/gorilla.finetune.model.LLM.llama.forward_inference#L230-L245","kind":"function","name":"forward_inference","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":230,"end_line":245,"context_start_line":210,"context_end_line":245,"code":"            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n\n    def forward(self, examples):\n        _bsz, seqlen = examples.shape\n        h = self.tok_embeddings(examples)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[:seqlen]\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h)\n        return output\n\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.tools.get_consolidated_ckpt","uri":"program://LLaMA-Adapter/module/gorilla.finetune.tools.get_consolidated_ckpt#L1-L196","kind":"module","name":"gorilla.finetune.tools.get_consolidated_ckpt","path":"gorilla/finetune/tools/get_consolidated_ckpt.py","language":"python","start_line":1,"end_line":196,"context_start_line":1,"context_end_line":196,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport sys\nimport os\nsys.path.append(os.path.abspath(__file__).rsplit('/', 2)[0])\n\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport time\nfrom pathlib import Path\nimport functools\nfrom functools import partial\n\nimport torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.distributed.fsdp import (\n    FullyShardedDataParallel as FSDP,\n    MixedPrecision,\n    ShardingStrategy,\n)\nfrom torch.distributed.algorithms._checkpoint.checkpoint_wrapper import (\n    checkpoint_wrapper,\n    CheckpointImpl,\n    apply_activation_checkpointing,\n)\nfrom torch.distributed.fsdp import (\n    FullyShardedDataParallel as FSDP,\n    StateDictType,\n    FullStateDictConfig,\n    ShardedStateDictConfig\n)\nfrom torch.distributed.fsdp.wrap import (\n    transformer_auto_wrap_policy,\n)\n\nfrom fairscale.nn.model_parallel import initialize as fs_init\n\nfrom apex.optimizers import FusedAdam\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom model.meta import MetaModel\nfrom model.LLM.llama import Attention, FeedForward\nfrom engine_finetune import train_one_epoch, val_one_epoch\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('MAE pre-training', add_help=False)\n    # Model parameters\n    parser.add_argument('--llama_type', default='llama', type=str, metavar='MODEL', choices=['llama', 'revllama'],\n                        help='Name of model to train')\n    parser.add_argument('--reversible_grad', action='store_true', default=False,\n                        help='Whether to use reversible grad')\n\n    parser.add_argument('--llama_config', default='params.json', type=str,\n                        help='Path to llama model config')\n\n    parser.add_argument('--load_dir', default='/path/to/sharded', type=str,\n                        help='path to sharded')\n    parser.add_argument('--save_dir', default='/path/to/full', type=str,\n                        help='path to full')\n\n    parser.add_argument('--lr', type=float, default=0.001, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--weight_decay', type=float, default=0.02,\n                        help='weight decay (default: 0.05)')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--model_parallel_size', type=int, default=1)\n    parser.add_argument('--data_parallel', type=str, choices=['ddp', 'sdp', 'fsdp'], default='sdp')\n    parser.add_argument('--precision', type=str, choices=['fp16', 'bf16', 'tf32'], default='bf16')\n    parser.add_argument('--checkpointing', action=\"store_true\", default=False,\n                        help=\"enable gradient checkopointing\")\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n    fs_init.initialize_model_parallel(args.model_parallel_size)\n    if args.precision == \"tf32\":\n        torch.backends.cuda.matmul.allow_tf32 = True\n        torch.backends.cudnn.allow_tf32 = True\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device('cuda')\n\n    global_rank = misc.get_rank()\n    mp_rank = fs_init.get_model_parallel_rank()\n    mp_world_size = fs_init.get_model_parallel_world_size()\n    dp_rank = fs_init.get_data_parallel_rank()\n    dp_world_size = fs_init.get_data_parallel_world_size()\n    dp_group = fs_init.get_data_parallel_group()\n\n    # define the model\n    model = MetaModel(args.llama_type, args.reversible_grad, args.llama_config)\n\n    mixed_precision_dtype = {\n        \"fp16\": torch.float16,\n        \"bf16\": torch.bfloat16,\n        \"tf32\": torch.float32,\n    }[args.precision]\n    model = FSDP(\n        model,\n        process_group=fs_init.get_data_parallel_group(),\n        auto_wrap_policy=functools.partial(\n            transformer_auto_wrap_policy,\n            transformer_layer_cls=[Attention, FeedForward],\n        ),\n        limit_all_gathers=True,\n        use_orig_params=True,\n        sync_module_states=True,\n        mixed_precision=MixedPrecision(\n            param_dtype=mixed_precision_dtype,\n            reduce_dtype=mixed_precision_dtype,\n            buffer_dtype=mixed_precision_dtype,\n        ),\n        sharding_strategy={\n            \"sdp\": ShardingStrategy.SHARD_GRAD_OP,\n            \"ddp\": ShardingStrategy.NO_SHARD,\n            \"fsdp\": ShardingStrategy.FULL_SHARD,\n        }[args.data_parallel],\n        device_id=device\n    )\n\n    param_groups = misc.add_weight_decay(model, args.weight_decay)\n    optimizer = FusedAdam(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n\n    with FSDP.state_dict_type(model, StateDictType.SHARDED_STATE_DICT,\n                              state_dict_config=ShardedStateDictConfig(offload_to_cpu=True)):\n        load_path = os.path.join(\n            args.load_dir,\n            f\"checkpoint.{global_rank:05d}-of-{misc.get_world_size():05d}.pth\",\n        )\n        state_dict = torch.load(load_path, map_location='cpu')\n        model.load_state_dict(state_dict['model'])\n        optimizer.load_state_dict(state_dict['optimizer'])\n\n\n    with FSDP.state_dict_type(\n            model,\n            StateDictType.FULL_STATE_DICT,\n            FullStateDictConfig(rank0_only=True, offload_to_cpu=True),\n    ):\n        consolidated_model_state_dict = model.state_dict()\n        optim_state_dict = FSDP.optim_state_dict(model, optimizer)\n\n        to_save = {\n            \"model\": consolidated_model_state_dict,\n            \"optimizer\": optim_state_dict,\n        }\n        consolidated_model_save_path = os.path.join(\n            args.save_dir,\n            f\"consolidated.{mp_rank:02d}-of-{mp_world_size:02d}.pth\",\n        )\n        if dp_rank == 0:\n            torch.save(to_save, consolidated_model_save_path)\n\n    # others common\n    keys = [\"epoch\", \"iter\", \"scaler\", \"args\", \"dataset_state\"]\n    other_to_save = {\n        k: state_dict[k] for k in keys if k in state_dict\n    }\n    other_save_path = os.path.join(\n            args.save_dir,\n            f\"other-{misc.get_rank():02d}-of-{misc.get_world_size():02d}.pth\",\n        )\n    torch.save(other_to_save, other_save_path)\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    main(args)","source_hash":"9ec2fc9de26ef16ab90e936f2ae235df21297a616bde0935aca78b9e9a92421c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.tools.get_consolidated_ckpt.get_args_parser","uri":"program://LLaMA-Adapter/function/gorilla.finetune.tools.get_consolidated_ckpt.get_args_parser#L59-L94","kind":"function","name":"get_args_parser","path":"gorilla/finetune/tools/get_consolidated_ckpt.py","language":"python","start_line":59,"end_line":94,"context_start_line":39,"context_end_line":114,"code":"    FullyShardedDataParallel as FSDP,\n    StateDictType,\n    FullStateDictConfig,\n    ShardedStateDictConfig\n)\nfrom torch.distributed.fsdp.wrap import (\n    transformer_auto_wrap_policy,\n)\n\nfrom fairscale.nn.model_parallel import initialize as fs_init\n\nfrom apex.optimizers import FusedAdam\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom model.meta import MetaModel\nfrom model.LLM.llama import Attention, FeedForward\nfrom engine_finetune import train_one_epoch, val_one_epoch\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('MAE pre-training', add_help=False)\n    # Model parameters\n    parser.add_argument('--llama_type', default='llama', type=str, metavar='MODEL', choices=['llama', 'revllama'],\n                        help='Name of model to train')\n    parser.add_argument('--reversible_grad', action='store_true', default=False,\n                        help='Whether to use reversible grad')\n\n    parser.add_argument('--llama_config', default='params.json', type=str,\n                        help='Path to llama model config')\n\n    parser.add_argument('--load_dir', default='/path/to/sharded', type=str,\n                        help='path to sharded')\n    parser.add_argument('--save_dir', default='/path/to/full', type=str,\n                        help='path to full')\n\n    parser.add_argument('--lr', type=float, default=0.001, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--weight_decay', type=float, default=0.02,\n                        help='weight decay (default: 0.05)')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--model_parallel_size', type=int, default=1)\n    parser.add_argument('--data_parallel', type=str, choices=['ddp', 'sdp', 'fsdp'], default='sdp')\n    parser.add_argument('--precision', type=str, choices=['fp16', 'bf16', 'tf32'], default='bf16')\n    parser.add_argument('--checkpointing', action=\"store_true\", default=False,\n                        help=\"enable gradient checkopointing\")\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n    fs_init.initialize_model_parallel(args.model_parallel_size)\n    if args.precision == \"tf32\":\n        torch.backends.cuda.matmul.allow_tf32 = True\n        torch.backends.cudnn.allow_tf32 = True\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device('cuda')\n\n    global_rank = misc.get_rank()\n    mp_rank = fs_init.get_model_parallel_rank()\n    mp_world_size = fs_init.get_model_parallel_world_size()\n    dp_rank = fs_init.get_data_parallel_rank()\n    dp_world_size = fs_init.get_data_parallel_world_size()\n    dp_group = fs_init.get_data_parallel_group()","source_hash":"9ec2fc9de26ef16ab90e936f2ae235df21297a616bde0935aca78b9e9a92421c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.finetune.tools.get_consolidated_ckpt.main","uri":"program://LLaMA-Adapter/function/gorilla.finetune.tools.get_consolidated_ckpt.main#L97-L190","kind":"function","name":"main","path":"gorilla/finetune/tools/get_consolidated_ckpt.py","language":"python","start_line":97,"end_line":190,"context_start_line":77,"context_end_line":196,"code":"    parser.add_argument('--weight_decay', type=float, default=0.02,\n                        help='weight decay (default: 0.05)')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--model_parallel_size', type=int, default=1)\n    parser.add_argument('--data_parallel', type=str, choices=['ddp', 'sdp', 'fsdp'], default='sdp')\n    parser.add_argument('--precision', type=str, choices=['fp16', 'bf16', 'tf32'], default='bf16')\n    parser.add_argument('--checkpointing', action=\"store_true\", default=False,\n                        help=\"enable gradient checkopointing\")\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n    fs_init.initialize_model_parallel(args.model_parallel_size)\n    if args.precision == \"tf32\":\n        torch.backends.cuda.matmul.allow_tf32 = True\n        torch.backends.cudnn.allow_tf32 = True\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device('cuda')\n\n    global_rank = misc.get_rank()\n    mp_rank = fs_init.get_model_parallel_rank()\n    mp_world_size = fs_init.get_model_parallel_world_size()\n    dp_rank = fs_init.get_data_parallel_rank()\n    dp_world_size = fs_init.get_data_parallel_world_size()\n    dp_group = fs_init.get_data_parallel_group()\n\n    # define the model\n    model = MetaModel(args.llama_type, args.reversible_grad, args.llama_config)\n\n    mixed_precision_dtype = {\n        \"fp16\": torch.float16,\n        \"bf16\": torch.bfloat16,\n        \"tf32\": torch.float32,\n    }[args.precision]\n    model = FSDP(\n        model,\n        process_group=fs_init.get_data_parallel_group(),\n        auto_wrap_policy=functools.partial(\n            transformer_auto_wrap_policy,\n            transformer_layer_cls=[Attention, FeedForward],\n        ),\n        limit_all_gathers=True,\n        use_orig_params=True,\n        sync_module_states=True,\n        mixed_precision=MixedPrecision(\n            param_dtype=mixed_precision_dtype,\n            reduce_dtype=mixed_precision_dtype,\n            buffer_dtype=mixed_precision_dtype,\n        ),\n        sharding_strategy={\n            \"sdp\": ShardingStrategy.SHARD_GRAD_OP,\n            \"ddp\": ShardingStrategy.NO_SHARD,\n            \"fsdp\": ShardingStrategy.FULL_SHARD,\n        }[args.data_parallel],\n        device_id=device\n    )\n\n    param_groups = misc.add_weight_decay(model, args.weight_decay)\n    optimizer = FusedAdam(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n\n    with FSDP.state_dict_type(model, StateDictType.SHARDED_STATE_DICT,\n                              state_dict_config=ShardedStateDictConfig(offload_to_cpu=True)):\n        load_path = os.path.join(\n            args.load_dir,\n            f\"checkpoint.{global_rank:05d}-of-{misc.get_world_size():05d}.pth\",\n        )\n        state_dict = torch.load(load_path, map_location='cpu')\n        model.load_state_dict(state_dict['model'])\n        optimizer.load_state_dict(state_dict['optimizer'])\n\n\n    with FSDP.state_dict_type(\n            model,\n            StateDictType.FULL_STATE_DICT,\n            FullStateDictConfig(rank0_only=True, offload_to_cpu=True),\n    ):\n        consolidated_model_state_dict = model.state_dict()\n        optim_state_dict = FSDP.optim_state_dict(model, optimizer)\n\n        to_save = {\n            \"model\": consolidated_model_state_dict,\n            \"optimizer\": optim_state_dict,\n        }\n        consolidated_model_save_path = os.path.join(\n            args.save_dir,\n            f\"consolidated.{mp_rank:02d}-of-{mp_world_size:02d}.pth\",\n        )\n        if dp_rank == 0:\n            torch.save(to_save, consolidated_model_save_path)\n\n    # others common\n    keys = [\"epoch\", \"iter\", \"scaler\", \"args\", \"dataset_state\"]\n    other_to_save = {\n        k: state_dict[k] for k in keys if k in state_dict\n    }\n    other_save_path = os.path.join(\n            args.save_dir,\n            f\"other-{misc.get_rank():02d}-of-{misc.get_world_size():02d}.pth\",\n        )\n    torch.save(other_to_save, other_save_path)\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    main(args)","source_hash":"9ec2fc9de26ef16ab90e936f2ae235df21297a616bde0935aca78b9e9a92421c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.setup","uri":"program://LLaMA-Adapter/module/gorilla.inference.setup#L1-L6","kind":"module","name":"gorilla.inference.setup","path":"gorilla/inference/setup.py","language":"python","start_line":1,"end_line":6,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom setuptools import setup, find_packages\n\nsetup(name=\"llama\", version=\"0.0.0\", packages=find_packages())","source_hash":"550bf5cf8205e38353070d61d3a01af1e7174f35a40c625c1e976d1f1f4133e8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.example","uri":"program://LLaMA-Adapter/module/gorilla.inference.example#L1-L153","kind":"module","name":"gorilla.inference.example","path":"gorilla/inference/example.py","language":"python","start_line":1,"end_line":153,"context_start_line":1,"context_end_line":153,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Tuple\nimport os\nimport sys\nimport torch\nimport fire\nimport time\nimport json\n\nfrom pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")['model']\n    checkpoint = {key[5:]:val for key, val in checkpoint.items() if key.startswith('llma.')}\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params\n    )\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    load_result = model.load_state_dict(checkpoint, strict=False)\n    print(load_result)\n\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    temperature: float = 0.8,\n    top_p: float = 0.95,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    generator = load(\n        ckpt_dir, tokenizer_path, local_rank, world_size, max_seq_len, max_batch_size\n    )\n\n    prompts = [\n        # For these prompts, the expected answer is the natural continuation of the prompt\n        \"I believe the meaning of life is\",\n        \"Simply put, the theory of relativity states that \",\n        \"Building a website can be done in 10 simple steps:\\n\",\n        # Few shot prompts: https://huggingface.co/blog/few-shot-learning-gpt-neo-and-inference-api\n        \"\"\"Tweet: \"I hate it when my phone battery dies.\"\nSentiment: Negative\n###\nTweet: \"My day has been 👍\"\nSentiment: Positive\n###\nTweet: \"This is the link to the article\"\nSentiment: Neutral\n###\nTweet: \"This new music video was incredibile\"\nSentiment:\"\"\",\n        \"\"\"Translate English to French:\n\nsea otter => loutre de mer\n\npeppermint => menthe poivrée\n\nplush girafe => girafe peluche\n\ncheese =>\"\"\",\n    ]\n    results = generator.generate(\n        prompts, max_gen_len=256, temperature=temperature, top_p=top_p\n    )\n\n    for result in results:\n        print(result)\n        print(\"\\n==================================\\n\")\n\n\n    while True:\n        instruction = input(\"Please type in your instruction \\n\")\n        text_in = input(\"please type in your input (optional) \\n\")\n        prompt = format_prompt(instruction, text_in)\n        results = generator.generate(\n            [prompt], max_gen_len=256, temperature=temperature, top_p=top_p\n        )\n        print(f\"result is: {results}\")\n\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\n\n\n\nif __name__ == \"__main__\":\n    fire.Fire(main)","source_hash":"fe4f54077b8d1bfb1c05199e82ffcabb0dcae6414514e6447b296190105e9e7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.example.setup_model_parallel","uri":"program://LLaMA-Adapter/function/gorilla.inference.example.setup_model_parallel#L19-L29","kind":"function","name":"setup_model_parallel","path":"gorilla/inference/example.py","language":"python","start_line":19,"end_line":29,"context_start_line":1,"context_end_line":49,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Tuple\nimport os\nimport sys\nimport torch\nimport fire\nimport time\nimport json\n\nfrom pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")['model']\n    checkpoint = {key[5:]:val for key, val in checkpoint.items() if key.startswith('llma.')}\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:","source_hash":"fe4f54077b8d1bfb1c05199e82ffcabb0dcae6414514e6447b296190105e9e7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.example.load","uri":"program://LLaMA-Adapter/function/gorilla.inference.example.load#L32-L65","kind":"function","name":"load","path":"gorilla/inference/example.py","language":"python","start_line":32,"end_line":65,"context_start_line":12,"context_end_line":85,"code":"from pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")['model']\n    checkpoint = {key[5:]:val for key, val in checkpoint.items() if key.startswith('llma.')}\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params\n    )\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    load_result = model.load_state_dict(checkpoint, strict=False)\n    print(load_result)\n\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    temperature: float = 0.8,\n    top_p: float = 0.95,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    generator = load(\n        ckpt_dir, tokenizer_path, local_rank, world_size, max_seq_len, max_batch_size\n    )\n\n    prompts = [\n        # For these prompts, the expected answer is the natural continuation of the prompt","source_hash":"fe4f54077b8d1bfb1c05199e82ffcabb0dcae6414514e6447b296190105e9e7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.example.main","uri":"program://LLaMA-Adapter/function/gorilla.inference.example.main#L68-L127","kind":"function","name":"main","path":"gorilla/inference/example.py","language":"python","start_line":68,"end_line":127,"context_start_line":48,"context_end_line":147,"code":"    checkpoint = {key[5:]:val for key, val in checkpoint.items() if key.startswith('llma.')}\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params\n    )\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    load_result = model.load_state_dict(checkpoint, strict=False)\n    print(load_result)\n\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    temperature: float = 0.8,\n    top_p: float = 0.95,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    generator = load(\n        ckpt_dir, tokenizer_path, local_rank, world_size, max_seq_len, max_batch_size\n    )\n\n    prompts = [\n        # For these prompts, the expected answer is the natural continuation of the prompt\n        \"I believe the meaning of life is\",\n        \"Simply put, the theory of relativity states that \",\n        \"Building a website can be done in 10 simple steps:\\n\",\n        # Few shot prompts: https://huggingface.co/blog/few-shot-learning-gpt-neo-and-inference-api\n        \"\"\"Tweet: \"I hate it when my phone battery dies.\"\nSentiment: Negative\n###\nTweet: \"My day has been 👍\"\nSentiment: Positive\n###\nTweet: \"This is the link to the article\"\nSentiment: Neutral\n###\nTweet: \"This new music video was incredibile\"\nSentiment:\"\"\",\n        \"\"\"Translate English to French:\n\nsea otter => loutre de mer\n\npeppermint => menthe poivrée\n\nplush girafe => girafe peluche\n\ncheese =>\"\"\",\n    ]\n    results = generator.generate(\n        prompts, max_gen_len=256, temperature=temperature, top_p=top_p\n    )\n\n    for result in results:\n        print(result)\n        print(\"\\n==================================\\n\")\n\n\n    while True:\n        instruction = input(\"Please type in your instruction \\n\")\n        text_in = input(\"please type in your input (optional) \\n\")\n        prompt = format_prompt(instruction, text_in)\n        results = generator.generate(\n            [prompt], max_gen_len=256, temperature=temperature, top_p=top_p\n        )\n        print(f\"result is: {results}\")\n\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})","source_hash":"fe4f54077b8d1bfb1c05199e82ffcabb0dcae6414514e6447b296190105e9e7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.example.format_prompt","uri":"program://LLaMA-Adapter/function/gorilla.inference.example.format_prompt#L131-L147","kind":"function","name":"format_prompt","path":"gorilla/inference/example.py","language":"python","start_line":131,"end_line":147,"context_start_line":111,"context_end_line":153,"code":"    results = generator.generate(\n        prompts, max_gen_len=256, temperature=temperature, top_p=top_p\n    )\n\n    for result in results:\n        print(result)\n        print(\"\\n==================================\\n\")\n\n\n    while True:\n        instruction = input(\"Please type in your instruction \\n\")\n        text_in = input(\"please type in your input (optional) \\n\")\n        prompt = format_prompt(instruction, text_in)\n        results = generator.generate(\n            [prompt], max_gen_len=256, temperature=temperature, top_p=top_p\n        )\n        print(f\"result is: {results}\")\n\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\n\n\n\nif __name__ == \"__main__\":\n    fire.Fire(main)","source_hash":"fe4f54077b8d1bfb1c05199e82ffcabb0dcae6414514e6447b296190105e9e7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_llama_adapter_v1","uri":"program://LLaMA-Adapter/module/gorilla.inference.gorilla_inference_llama_adapter_v1#L1-L180","kind":"module","name":"gorilla.inference.gorilla_inference_llama_adapter_v1","path":"gorilla/inference/gorilla_inference_llama_adapter_v1.py","language":"python","start_line":1,"end_line":180,"context_start_line":1,"context_end_line":180,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Tuple\nimport os\nimport sys\nimport torch\nimport fire\nimport time\nimport json\nfrom tqdm import tqdm\n\nfrom pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama_for_adapter import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n    # quantizer: bool=False,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")\n    adapter_checkpoint = torch.load(adapter_path, map_location=\"cpu\")\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    # model_args: ModelArgs = ModelArgs(max_seq_len=max_seq_len, max_batch_size=max_batch_size, quantizer=quantizer, **params)\n    model_args: ModelArgs = ModelArgs(max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params)\n    model_args.adapter_layer = int(adapter_checkpoint[\"adapter_query.weight\"].shape[0] / model_args.adapter_len)\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    print(model)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    model.load_state_dict(checkpoint, strict=False)\n    model.load_state_dict(adapter_checkpoint, strict=False)\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    temperature: float = 0.1,\n    top_p: float = 0.75,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n    # quantizer: bool = False,\n    dataset_path: str = '../gorilla-main/eval/eval-data/questions/{tensorflowhub, huggingface, torchhub}/questions_{tensorflowhub, huggingface, torchhub}_0_shot.jsonl',\n    inference_batch_size = 32,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    # generator = load(ckpt_dir, tokenizer_path, adapter_path, local_rank, world_size, max_seq_len, max_batch_size, quantizer)\n    generator = load(ckpt_dir, tokenizer_path, adapter_path, local_rank, world_size, max_seq_len, max_batch_size)\n\n    questions_json = get_questions(dataset_path)\n\n    # ans_jsons = []\n    # for i, line in enumerate(tqdm(questions_json)):\n    #     ques_json = json.loads(line)\n    #     idx = ques_json[\"question_id\"]\n    #     prompt = ques_json[\"text\"]\n    #     formated_prompt = format_prompt(prompt)\n    #     results = generator.generate(\n    #         [formated_prompt], max_gen_len=256, temperature=temperature, top_p=top_p\n    #     )\n    #     ans_jsons.append(\n    #         {\n    #             \"question_id\": idx,\n    #             \"questions\": prompt,\n    #             \"text\": results[0],\n    #         }\n    #     )\n\n    ans_jsons = []\n    batch_idx = []\n    batch_prompt = []\n    batch_formated_prompt = []\n    question_num = len(questions_json)\n    for i, line in enumerate(tqdm(questions_json)):\n        ques_json = json.loads(line)\n        idx = ques_json[\"question_id\"]\n        prompt = ques_json[\"text\"]\n        formated_prompt = format_prompt(prompt)\n        batch_idx.append(idx)\n        batch_prompt.append(prompt)\n        batch_formated_prompt.append(formated_prompt)\n        if (i+1) % inference_batch_size == 0 or i == question_num:\n            results = generator.generate(\n                batch_formated_prompt, max_gen_len=256, temperature=temperature, top_p=top_p\n            )\n            for i in range(len(batch_idx)):\n                ans_jsons.append(\n                    {\n                        \"question_id\": batch_idx[i],\n                        \"questions\": batch_prompt[i],\n                        \"text\": results[i],\n                    }\n                )\n            batch_idx = []\n            batch_prompt = []\n            batch_formated_prompt = []\n\n\n    # Write output to file\n    with open(os.path.join(os.path.dirname(adapter_path), 'model_prediction_results.jsonl'), \"w\") as ans_file:\n        for line in ans_jsons:\n            ans_file.write(json.dumps(line) + \"\\n\")\n\n\n\n\n\n\nif __name__ == \"__main__\":\n    fire.Fire(main)","source_hash":"d75f8b7173d048be89a4f376b8c4d45f18e70f843861578d687ac52a8192d849","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_llama_adapter_v1.setup_model_parallel","uri":"program://LLaMA-Adapter/function/gorilla.inference.gorilla_inference_llama_adapter_v1.setup_model_parallel#L20-L30","kind":"function","name":"setup_model_parallel","path":"gorilla/inference/gorilla_inference_llama_adapter_v1.py","language":"python","start_line":20,"end_line":30,"context_start_line":1,"context_end_line":50,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Tuple\nimport os\nimport sys\nimport torch\nimport fire\nimport time\nimport json\nfrom tqdm import tqdm\n\nfrom pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama_for_adapter import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n","source_hash":"d75f8b7173d048be89a4f376b8c4d45f18e70f843861578d687ac52a8192d849","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_llama_adapter_v1.format_prompt","uri":"program://LLaMA-Adapter/function/gorilla.inference.gorilla_inference_llama_adapter_v1.format_prompt#L33-L49","kind":"function","name":"format_prompt","path":"gorilla/inference/gorilla_inference_llama_adapter_v1.py","language":"python","start_line":33,"end_line":49,"context_start_line":13,"context_end_line":69,"code":"from pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama_for_adapter import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,","source_hash":"d75f8b7173d048be89a4f376b8c4d45f18e70f843861578d687ac52a8192d849","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_llama_adapter_v1.get_questions","uri":"program://LLaMA-Adapter/function/gorilla.inference.gorilla_inference_llama_adapter_v1.get_questions#L51-L59","kind":"function","name":"get_questions","path":"gorilla/inference/gorilla_inference_llama_adapter_v1.py","language":"python","start_line":51,"end_line":59,"context_start_line":31,"context_end_line":79,"code":"\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n    # quantizer: bool=False,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")","source_hash":"d75f8b7173d048be89a4f376b8c4d45f18e70f843861578d687ac52a8192d849","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_llama_adapter_v1.load","uri":"program://LLaMA-Adapter/function/gorilla.inference.gorilla_inference_llama_adapter_v1.load#L62-L97","kind":"function","name":"load","path":"gorilla/inference/gorilla_inference_llama_adapter_v1.py","language":"python","start_line":62,"end_line":97,"context_start_line":42,"context_end_line":117,"code":"            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n    # quantizer: bool=False,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")\n    adapter_checkpoint = torch.load(adapter_path, map_location=\"cpu\")\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    # model_args: ModelArgs = ModelArgs(max_seq_len=max_seq_len, max_batch_size=max_batch_size, quantizer=quantizer, **params)\n    model_args: ModelArgs = ModelArgs(max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params)\n    model_args.adapter_layer = int(adapter_checkpoint[\"adapter_query.weight\"].shape[0] / model_args.adapter_len)\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    print(model)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    model.load_state_dict(checkpoint, strict=False)\n    model.load_state_dict(adapter_checkpoint, strict=False)\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    temperature: float = 0.1,\n    top_p: float = 0.75,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n    # quantizer: bool = False,\n    dataset_path: str = '../gorilla-main/eval/eval-data/questions/{tensorflowhub, huggingface, torchhub}/questions_{tensorflowhub, huggingface, torchhub}_0_shot.jsonl',\n    inference_batch_size = 32,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    # generator = load(ckpt_dir, tokenizer_path, adapter_path, local_rank, world_size, max_seq_len, max_batch_size, quantizer)","source_hash":"d75f8b7173d048be89a4f376b8c4d45f18e70f843861578d687ac52a8192d849","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_llama_adapter_v1.main","uri":"program://LLaMA-Adapter/function/gorilla.inference.gorilla_inference_llama_adapter_v1.main#L101-L172","kind":"function","name":"main","path":"gorilla/inference/gorilla_inference_llama_adapter_v1.py","language":"python","start_line":101,"end_line":172,"context_start_line":81,"context_end_line":180,"code":"    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    # model_args: ModelArgs = ModelArgs(max_seq_len=max_seq_len, max_batch_size=max_batch_size, quantizer=quantizer, **params)\n    model_args: ModelArgs = ModelArgs(max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params)\n    model_args.adapter_layer = int(adapter_checkpoint[\"adapter_query.weight\"].shape[0] / model_args.adapter_len)\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    print(model)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    model.load_state_dict(checkpoint, strict=False)\n    model.load_state_dict(adapter_checkpoint, strict=False)\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    adapter_path: str,\n    temperature: float = 0.1,\n    top_p: float = 0.75,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n    # quantizer: bool = False,\n    dataset_path: str = '../gorilla-main/eval/eval-data/questions/{tensorflowhub, huggingface, torchhub}/questions_{tensorflowhub, huggingface, torchhub}_0_shot.jsonl',\n    inference_batch_size = 32,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    # generator = load(ckpt_dir, tokenizer_path, adapter_path, local_rank, world_size, max_seq_len, max_batch_size, quantizer)\n    generator = load(ckpt_dir, tokenizer_path, adapter_path, local_rank, world_size, max_seq_len, max_batch_size)\n\n    questions_json = get_questions(dataset_path)\n\n    # ans_jsons = []\n    # for i, line in enumerate(tqdm(questions_json)):\n    #     ques_json = json.loads(line)\n    #     idx = ques_json[\"question_id\"]\n    #     prompt = ques_json[\"text\"]\n    #     formated_prompt = format_prompt(prompt)\n    #     results = generator.generate(\n    #         [formated_prompt], max_gen_len=256, temperature=temperature, top_p=top_p\n    #     )\n    #     ans_jsons.append(\n    #         {\n    #             \"question_id\": idx,\n    #             \"questions\": prompt,\n    #             \"text\": results[0],\n    #         }\n    #     )\n\n    ans_jsons = []\n    batch_idx = []\n    batch_prompt = []\n    batch_formated_prompt = []\n    question_num = len(questions_json)\n    for i, line in enumerate(tqdm(questions_json)):\n        ques_json = json.loads(line)\n        idx = ques_json[\"question_id\"]\n        prompt = ques_json[\"text\"]\n        formated_prompt = format_prompt(prompt)\n        batch_idx.append(idx)\n        batch_prompt.append(prompt)\n        batch_formated_prompt.append(formated_prompt)\n        if (i+1) % inference_batch_size == 0 or i == question_num:\n            results = generator.generate(\n                batch_formated_prompt, max_gen_len=256, temperature=temperature, top_p=top_p\n            )\n            for i in range(len(batch_idx)):\n                ans_jsons.append(\n                    {\n                        \"question_id\": batch_idx[i],\n                        \"questions\": batch_prompt[i],\n                        \"text\": results[i],\n                    }\n                )\n            batch_idx = []\n            batch_prompt = []\n            batch_formated_prompt = []\n\n\n    # Write output to file\n    with open(os.path.join(os.path.dirname(adapter_path), 'model_prediction_results.jsonl'), \"w\") as ans_file:\n        for line in ans_jsons:\n            ans_file.write(json.dumps(line) + \"\\n\")\n\n\n\n\n\n\nif __name__ == \"__main__\":\n    fire.Fire(main)","source_hash":"d75f8b7173d048be89a4f376b8c4d45f18e70f843861578d687ac52a8192d849","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_full_finetune","uri":"program://LLaMA-Adapter/module/gorilla.inference.gorilla_inference_full_finetune#L1-L175","kind":"module","name":"gorilla.inference.gorilla_inference_full_finetune","path":"gorilla/inference/gorilla_inference_full_finetune.py","language":"python","start_line":1,"end_line":175,"context_start_line":1,"context_end_line":175,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Tuple\nimport os\nimport sys\nimport torch\nimport fire\nimport time\nimport json\nfrom tqdm import tqdm\n\nfrom pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"consolidated*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")\n    checkpoint = {key[5:]:val for key, val in checkpoint.items() if key.startswith('llma.')}\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params\n    )\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    load_result = model.load_state_dict(checkpoint, strict=False)\n    print(load_result)\n\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    temperature: float = 0.8,\n    top_p: float = 0.95,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n    dataset_path: str = '../gorilla-main/eval/eval-data/questions/{tensorflowhub, huggingface, torchhub}/questions_{tensorflowhub, huggingface, torchhub}_0_shot.jsonl',\n    inference_batch_size = 32,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    generator = load(\n        ckpt_dir, tokenizer_path, local_rank, world_size, max_seq_len, max_batch_size\n    )\n\n    questions_json = get_questions(dataset_path)\n\n    # ans_jsons = []\n    # for i, line in enumerate(tqdm(questions_json)):\n    #     ques_json = json.loads(line)\n    #     idx = ques_json[\"question_id\"]\n    #     prompt = ques_json[\"text\"]\n    #     formated_prompt = format_prompt(prompt)\n    #     results = generator.generate(\n    #         [formated_prompt], max_gen_len=256, temperature=temperature, top_p=top_p\n    #     )\n    #     ans_jsons.append(\n    #         {\n    #             \"question_id\": idx,\n    #             \"questions\": prompt,\n    #             \"text\": results[0],\n    #         }\n    #     )\n\n    ans_jsons = []\n    batch_idx = []\n    batch_prompt = []\n    batch_formated_prompt = []\n    question_num = len(questions_json)\n    for i, line in enumerate(tqdm(questions_json)):\n        ques_json = json.loads(line)\n        idx = ques_json[\"question_id\"]\n        prompt = ques_json[\"text\"]\n        formated_prompt = format_prompt(prompt)\n        batch_idx.append(idx)\n        batch_prompt.append(prompt)\n        batch_formated_prompt.append(formated_prompt)\n        if (i+1) % inference_batch_size == 0 or i == question_num:\n            results = generator.generate(\n                batch_formated_prompt, max_gen_len=256, temperature=temperature, top_p=top_p\n            )\n            for i in range(len(batch_idx)):\n                ans_jsons.append(\n                    {\n                        \"question_id\": batch_idx[i],\n                        \"questions\": batch_prompt[i],\n                        \"text\": results[i],\n                    }\n                )\n            batch_idx = []\n            batch_prompt = []\n            batch_formated_prompt = []\n\n    # Write output to file\n    with open(os.path.join(ckpt_dir, 'model_prediction_results.jsonl'), \"w\") as ans_file:\n        for line in ans_jsons:\n            ans_file.write(json.dumps(line) + \"\\n\")\n\n\n\n\n\n\nif __name__ == \"__main__\":\n    fire.Fire(main)","source_hash":"657c95387fb4aa14fdcfcf7f8422a42dd1e59b5f6f72e347c10196fcb98cbe01","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_full_finetune.setup_model_parallel","uri":"program://LLaMA-Adapter/function/gorilla.inference.gorilla_inference_full_finetune.setup_model_parallel#L20-L30","kind":"function","name":"setup_model_parallel","path":"gorilla/inference/gorilla_inference_full_finetune.py","language":"python","start_line":20,"end_line":30,"context_start_line":1,"context_end_line":50,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Tuple\nimport os\nimport sys\nimport torch\nimport fire\nimport time\nimport json\nfrom tqdm import tqdm\n\nfrom pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n","source_hash":"657c95387fb4aa14fdcfcf7f8422a42dd1e59b5f6f72e347c10196fcb98cbe01","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_full_finetune.format_prompt","uri":"program://LLaMA-Adapter/function/gorilla.inference.gorilla_inference_full_finetune.format_prompt#L33-L49","kind":"function","name":"format_prompt","path":"gorilla/inference/gorilla_inference_full_finetune.py","language":"python","start_line":33,"end_line":49,"context_start_line":13,"context_end_line":69,"code":"from pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))\n\n    torch.distributed.init_process_group(\"nccl\")\n    initialize_model_parallel(world_size)\n    torch.cuda.set_device(local_rank)\n\n    # seed must be the same in all processes\n    torch.manual_seed(1)\n    return local_rank, world_size\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n) -> LLaMA:","source_hash":"657c95387fb4aa14fdcfcf7f8422a42dd1e59b5f6f72e347c10196fcb98cbe01","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_full_finetune.get_questions","uri":"program://LLaMA-Adapter/function/gorilla.inference.gorilla_inference_full_finetune.get_questions#L51-L59","kind":"function","name":"get_questions","path":"gorilla/inference/gorilla_inference_full_finetune.py","language":"python","start_line":51,"end_line":59,"context_start_line":31,"context_end_line":79,"code":"\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"consolidated*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")\n    checkpoint = {key[5:]:val for key, val in checkpoint.items() if key.startswith('llma.')}\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:","source_hash":"657c95387fb4aa14fdcfcf7f8422a42dd1e59b5f6f72e347c10196fcb98cbe01","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_full_finetune.load","uri":"program://LLaMA-Adapter/function/gorilla.inference.gorilla_inference_full_finetune.load#L62-L95","kind":"function","name":"load","path":"gorilla/inference/gorilla_inference_full_finetune.py","language":"python","start_line":62,"end_line":95,"context_start_line":42,"context_end_line":115,"code":"            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\n\ndef load(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    local_rank: int,\n    world_size: int,\n    max_seq_len: int,\n    max_batch_size: int,\n) -> LLaMA:\n    start_time = time.time()\n    checkpoints = sorted(Path(ckpt_dir).glob(\"consolidated*.pth\"))\n    assert world_size == len(\n        checkpoints\n    ), f\"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}\"\n    ckpt_path = checkpoints[local_rank]\n    print(\"Loading\")\n    checkpoint = torch.load(ckpt_path, map_location=\"cpu\")\n    checkpoint = {key[5:]:val for key, val in checkpoint.items() if key.startswith('llma.')}\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params\n    )\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    load_result = model.load_state_dict(checkpoint, strict=False)\n    print(load_result)\n\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    temperature: float = 0.8,\n    top_p: float = 0.95,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n    dataset_path: str = '../gorilla-main/eval/eval-data/questions/{tensorflowhub, huggingface, torchhub}/questions_{tensorflowhub, huggingface, torchhub}_0_shot.jsonl',\n    inference_batch_size = 32,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    generator = load(\n        ckpt_dir, tokenizer_path, local_rank, world_size, max_seq_len, max_batch_size\n    )\n","source_hash":"657c95387fb4aa14fdcfcf7f8422a42dd1e59b5f6f72e347c10196fcb98cbe01","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.gorilla_inference_full_finetune.main","uri":"program://LLaMA-Adapter/function/gorilla.inference.gorilla_inference_full_finetune.main#L98-L167","kind":"function","name":"main","path":"gorilla/inference/gorilla_inference_full_finetune.py","language":"python","start_line":98,"end_line":167,"context_start_line":78,"context_end_line":175,"code":"    checkpoint = {key[5:]:val for key, val in checkpoint.items() if key.startswith('llma.')}\n    with open(Path(ckpt_dir) / \"params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=max_seq_len, max_batch_size=max_batch_size, **params\n    )\n    tokenizer = Tokenizer(model_path=tokenizer_path)\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    load_result = model.load_state_dict(checkpoint, strict=False)\n    print(load_result)\n\n    generator = LLaMA(model, tokenizer)\n    print(f\"Loaded in {time.time() - start_time:.2f} seconds\")\n    return generator\n\n\ndef main(\n    ckpt_dir: str,\n    tokenizer_path: str,\n    temperature: float = 0.8,\n    top_p: float = 0.95,\n    max_seq_len: int = 512,\n    max_batch_size: int = 32,\n    dataset_path: str = '../gorilla-main/eval/eval-data/questions/{tensorflowhub, huggingface, torchhub}/questions_{tensorflowhub, huggingface, torchhub}_0_shot.jsonl',\n    inference_batch_size = 32,\n):\n    local_rank, world_size = setup_model_parallel()\n    if local_rank > 0:\n        sys.stdout = open(os.devnull, \"w\")\n\n    generator = load(\n        ckpt_dir, tokenizer_path, local_rank, world_size, max_seq_len, max_batch_size\n    )\n\n    questions_json = get_questions(dataset_path)\n\n    # ans_jsons = []\n    # for i, line in enumerate(tqdm(questions_json)):\n    #     ques_json = json.loads(line)\n    #     idx = ques_json[\"question_id\"]\n    #     prompt = ques_json[\"text\"]\n    #     formated_prompt = format_prompt(prompt)\n    #     results = generator.generate(\n    #         [formated_prompt], max_gen_len=256, temperature=temperature, top_p=top_p\n    #     )\n    #     ans_jsons.append(\n    #         {\n    #             \"question_id\": idx,\n    #             \"questions\": prompt,\n    #             \"text\": results[0],\n    #         }\n    #     )\n\n    ans_jsons = []\n    batch_idx = []\n    batch_prompt = []\n    batch_formated_prompt = []\n    question_num = len(questions_json)\n    for i, line in enumerate(tqdm(questions_json)):\n        ques_json = json.loads(line)\n        idx = ques_json[\"question_id\"]\n        prompt = ques_json[\"text\"]\n        formated_prompt = format_prompt(prompt)\n        batch_idx.append(idx)\n        batch_prompt.append(prompt)\n        batch_formated_prompt.append(formated_prompt)\n        if (i+1) % inference_batch_size == 0 or i == question_num:\n            results = generator.generate(\n                batch_formated_prompt, max_gen_len=256, temperature=temperature, top_p=top_p\n            )\n            for i in range(len(batch_idx)):\n                ans_jsons.append(\n                    {\n                        \"question_id\": batch_idx[i],\n                        \"questions\": batch_prompt[i],\n                        \"text\": results[i],\n                    }\n                )\n            batch_idx = []\n            batch_prompt = []\n            batch_formated_prompt = []\n\n    # Write output to file\n    with open(os.path.join(ckpt_dir, 'model_prediction_results.jsonl'), \"w\") as ans_file:\n        for line in ans_jsons:\n            ans_file.write(json.dumps(line) + \"\\n\")\n\n\n\n\n\n\nif __name__ == \"__main__\":\n    fire.Fire(main)","source_hash":"657c95387fb4aa14fdcfcf7f8422a42dd1e59b5f6f72e347c10196fcb98cbe01","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.generation","uri":"program://LLaMA-Adapter/module/gorilla.inference.llama_for_adapter.generation#L1-L75","kind":"module","name":"gorilla.inference.llama_for_adapter.generation","path":"gorilla/inference/llama_for_adapter/generation.py","language":"python","start_line":1,"end_line":75,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.generation.LLaMA","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama_for_adapter.generation.LLaMA#L12-L64","kind":"class","name":"LLaMA","path":"gorilla/inference/llama_for_adapter/generation.py","language":"python","start_line":12,"end_line":64,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.generation.sample_top_p","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.generation.sample_top_p#L67-L75","kind":"function","name":"sample_top_p","path":"gorilla/inference/llama_for_adapter/generation.py","language":"python","start_line":67,"end_line":75,"context_start_line":47,"context_end_line":75,"code":"                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.generation.__init__","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.generation.__init__#L13-L15","kind":"function","name":"__init__","path":"gorilla/inference/llama_for_adapter/generation.py","language":"python","start_line":13,"end_line":15,"context_start_line":1,"context_end_line":35,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.generation.generate","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.generation.generate#L17-L64","kind":"function","name":"generate","path":"gorilla/inference/llama_for_adapter/generation.py","language":"python","start_line":17,"end_line":64,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model","uri":"program://LLaMA-Adapter/module/gorilla.inference.llama_for_adapter.model#L1-L240","kind":"module","name":"gorilla.inference.llama_for_adapter.model","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":1,"end_line":240,"context_start_line":1,"context_end_line":240,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.nn.functional as F\nfrom fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    adapter_len: int = 10\n    adapter_layer: int = 8\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wv = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wo = RowParallelLinear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=False,\n            input_is_parallel=True,\n            init_method=lambda x: x,\n        )\n\n        self.cache_k = torch.zeros((args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)).cuda()\n        self.cache_v = torch.zeros((args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)).cuda()\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        self.cache_k = self.cache_k.to(xq)\n        self.cache_v = self.cache_v.to(xq)\n\n        self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk\n        self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n        keys = self.cache_k[:bsz, : start_pos + seqlen]\n        values = self.cache_v[:bsz, : start_pos + seqlen]\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_k = self.wk(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_v = self.wv(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_k = adapter_k.transpose(1, 2)\n            adapter_v = adapter_v.transpose(1, 2)\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        if adapter is not None:\n            adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n            adapter_scores = self.gate * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n            output = output + torch.matmul(adapter_scores, adapter_v)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n        self.w2 = RowParallelLinear(hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x)\n        self.w3 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        prompt = self.adapter_query.weight.reshape(\n            self.params.adapter_layer, self.params.adapter_len, self.params.dim\n        ).unsqueeze(1)\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers[: -1 * self.params.adapter_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        layer_index = 0\n        for layer in self.layers[-1 * self.params.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, prompt[layer_index])\n            layer_index = layer_index + 1\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.ModelArgs","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama_for_adapter.model.ModelArgs#L16-L28","kind":"class","name":"ModelArgs","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":16,"end_line":28,"context_start_line":1,"context_end_line":48,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.nn.functional as F\nfrom fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    adapter_len: int = 10\n    adapter_layer: int = 8\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.RMSNorm","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama_for_adapter.model.RMSNorm#L31-L42","kind":"class","name":"RMSNorm","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":31,"end_line":42,"context_start_line":11,"context_end_line":62,"code":"from fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    adapter_len: int = 10\n    adapter_layer: int = 8\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.precompute_freqs_cis","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.model.precompute_freqs_cis#L45-L50","kind":"function","name":"precompute_freqs_cis","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":45,"end_line":50,"context_start_line":25,"context_end_line":70,"code":"    max_seq_len: int = 2048\n\n    adapter_len: int = 10\n    adapter_layer: int = 8\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.reshape_for_broadcast","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.model.reshape_for_broadcast#L53-L58","kind":"function","name":"reshape_for_broadcast","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":53,"end_line":58,"context_start_line":33,"context_end_line":78,"code":"        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.apply_rotary_emb","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.model.apply_rotary_emb#L61-L71","kind":"function","name":"apply_rotary_emb","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":61,"end_line":71,"context_start_line":41,"context_end_line":91,"code":"        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.Attention","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama_for_adapter.model.Attention#L74-L155","kind":"class","name":"Attention","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":74,"end_line":155,"context_start_line":54,"context_end_line":175,"code":"    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wv = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wo = RowParallelLinear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=False,\n            input_is_parallel=True,\n            init_method=lambda x: x,\n        )\n\n        self.cache_k = torch.zeros((args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)).cuda()\n        self.cache_v = torch.zeros((args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)).cuda()\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        self.cache_k = self.cache_k.to(xq)\n        self.cache_v = self.cache_v.to(xq)\n\n        self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk\n        self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n        keys = self.cache_k[:bsz, : start_pos + seqlen]\n        values = self.cache_v[:bsz, : start_pos + seqlen]\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_k = self.wk(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_v = self.wv(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_k = adapter_k.transpose(1, 2)\n            adapter_v = adapter_v.transpose(1, 2)\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        if adapter is not None:\n            adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n            adapter_scores = self.gate * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n            output = output + torch.matmul(adapter_scores, adapter_v)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n        self.w2 = RowParallelLinear(hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x)\n        self.w3 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.FeedForward","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama_for_adapter.model.FeedForward#L158-L174","kind":"class","name":"FeedForward","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":158,"end_line":174,"context_start_line":138,"context_end_line":194,"code":"            adapter_v = self.wv(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_k = adapter_k.transpose(1, 2)\n            adapter_v = adapter_v.transpose(1, 2)\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        if adapter is not None:\n            adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n            adapter_scores = self.gate * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n            output = output + torch.matmul(adapter_scores, adapter_v)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n        self.w2 = RowParallelLinear(hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x)\n        self.w3 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.TransformerBlock","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama_for_adapter.model.TransformerBlock#L177-L194","kind":"class","name":"TransformerBlock","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":177,"end_line":194,"context_start_line":157,"context_end_line":214,"code":"\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n        self.w2 = RowParallelLinear(hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x)\n        self.w3 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.Transformer","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama_for_adapter.model.Transformer#L197-L240","kind":"class","name":"Transformer","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":197,"end_line":240,"context_start_line":177,"context_end_line":240,"code":"class TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        prompt = self.adapter_query.weight.reshape(\n            self.params.adapter_layer, self.params.adapter_len, self.params.dim\n        ).unsqueeze(1)\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers[: -1 * self.params.adapter_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        layer_index = 0\n        for layer in self.layers[-1 * self.params.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, prompt[layer_index])\n            layer_index = layer_index + 1\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.__init__","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.model.__init__#L198-L216","kind":"function","name":"__init__","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":198,"end_line":216,"context_start_line":178,"context_end_line":236,"code":"    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        prompt = self.adapter_query.weight.reshape(\n            self.params.adapter_layer, self.params.adapter_len, self.params.dim\n        ).unsqueeze(1)\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers[: -1 * self.params.adapter_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        layer_index = 0\n        for layer in self.layers[-1 * self.params.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, prompt[layer_index])","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model._norm","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.model._norm#L37-L38","kind":"function","name":"_norm","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":37,"end_line":38,"context_start_line":17,"context_end_line":58,"code":"    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    adapter_len: int = 10\n    adapter_layer: int = 8\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.model.forward","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.model.forward#L219-L240","kind":"function","name":"forward","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":219,"end_line":240,"context_start_line":199,"context_end_line":240,"code":"        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        prompt = self.adapter_query.weight.reshape(\n            self.params.adapter_layer, self.params.adapter_len, self.params.dim\n        ).unsqueeze(1)\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers[: -1 * self.params.adapter_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        layer_index = 0\n        for layer in self.layers[-1 * self.params.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, prompt[layer_index])\n            layer_index = layer_index + 1\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.tokenizer","uri":"program://LLaMA-Adapter/module/gorilla.inference.llama_for_adapter.tokenizer#L1-L38","kind":"module","name":"gorilla.inference.llama_for_adapter.tokenizer","path":"gorilla/inference/llama_for_adapter/tokenizer.py","language":"python","start_line":1,"end_line":38,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.tokenizer.Tokenizer","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama_for_adapter.tokenizer.Tokenizer#L13-L38","kind":"class","name":"Tokenizer","path":"gorilla/inference/llama_for_adapter/tokenizer.py","language":"python","start_line":13,"end_line":38,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.tokenizer.__init__","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.tokenizer.__init__#L14-L26","kind":"function","name":"__init__","path":"gorilla/inference/llama_for_adapter/tokenizer.py","language":"python","start_line":14,"end_line":26,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.tokenizer.encode","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.tokenizer.encode#L28-L35","kind":"function","name":"encode","path":"gorilla/inference/llama_for_adapter/tokenizer.py","language":"python","start_line":28,"end_line":35,"context_start_line":8,"context_end_line":38,"code":"from sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama_for_adapter.tokenizer.decode","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama_for_adapter.tokenizer.decode#L37-L38","kind":"function","name":"decode","path":"gorilla/inference/llama_for_adapter/tokenizer.py","language":"python","start_line":37,"end_line":38,"context_start_line":17,"context_end_line":38,"code":"        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.generation","uri":"program://LLaMA-Adapter/module/gorilla.inference.llama.generation#L1-L77","kind":"module","name":"gorilla.inference.llama.generation","path":"gorilla/inference/llama/generation.py","language":"python","start_line":1,"end_line":77,"context_start_line":1,"context_end_line":77,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.tokenizer import Tokenizer\nfrom llama.model import Transformer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"ccf9877e509c2f5fa350117dd9385fce9f055329d36d43a5f4fe70a7b7b24b81","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.generation.LLaMA","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama.generation.LLaMA#L12-L66","kind":"class","name":"LLaMA","path":"gorilla/inference/llama/generation.py","language":"python","start_line":12,"end_line":66,"context_start_line":1,"context_end_line":77,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.tokenizer import Tokenizer\nfrom llama.model import Transformer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"ccf9877e509c2f5fa350117dd9385fce9f055329d36d43a5f4fe70a7b7b24b81","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.generation.sample_top_p","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.generation.sample_top_p#L69-L77","kind":"function","name":"sample_top_p","path":"gorilla/inference/llama/generation.py","language":"python","start_line":69,"end_line":77,"context_start_line":49,"context_end_line":77,"code":"            # only replace token if prompt has already been generated\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"ccf9877e509c2f5fa350117dd9385fce9f055329d36d43a5f4fe70a7b7b24b81","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.generation.__init__","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.generation.__init__#L13-L15","kind":"function","name":"__init__","path":"gorilla/inference/llama/generation.py","language":"python","start_line":13,"end_line":15,"context_start_line":1,"context_end_line":35,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.tokenizer import Tokenizer\nfrom llama.model import Transformer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()","source_hash":"ccf9877e509c2f5fa350117dd9385fce9f055329d36d43a5f4fe70a7b7b24b81","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.generation.generate","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.generation.generate#L17-L66","kind":"function","name":"generate","path":"gorilla/inference/llama/generation.py","language":"python","start_line":17,"end_line":66,"context_start_line":1,"context_end_line":77,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.tokenizer import Tokenizer\nfrom llama.model import Transformer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"ccf9877e509c2f5fa350117dd9385fce9f055329d36d43a5f4fe70a7b7b24b81","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model","uri":"program://LLaMA-Adapter/module/gorilla.inference.llama.model#L1-L238","kind":"module","name":"gorilla.inference.llama.model","path":"gorilla/inference/llama/model.py","language":"python","start_line":1,"end_line":238,"context_start_line":1,"context_end_line":238,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\n\nimport torch\nfrom torch import nn\nimport torch.nn.functional as F\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nfrom fairscale.nn.model_parallel.layers import (\n    ParallelEmbedding,\n    RowParallelLinear,\n    ColumnParallelLinear,\n)\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wv = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wo = RowParallelLinear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=False,\n            input_is_parallel=True,\n            init_method=lambda x: x,\n        )\n\n        self.cache_k = torch.zeros(\n            (args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)\n        ).cuda()\n        self.cache_v = torch.zeros(\n            (args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)\n        ).cuda()\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor]):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        self.cache_k = self.cache_k.to(xq)\n        self.cache_v = self.cache_v.to(xq)\n\n        self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk\n        self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n        keys = self.cache_k[:bsz, : start_pos + seqlen]\n        values = self.cache_v[:bsz, : start_pos + seqlen]\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x\n        )\n        self.w2 = RowParallelLinear(\n            hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x\n        )\n        self.w3 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x\n        )\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor]):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=lambda x: x\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(\n            params.dim, params.vocab_size, bias=False, init_method=lambda x: x\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.ModelArgs","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama.model.ModelArgs#L21-L30","kind":"class","name":"ModelArgs","path":"gorilla/inference/llama/model.py","language":"python","start_line":21,"end_line":30,"context_start_line":1,"context_end_line":50,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\n\nimport torch\nfrom torch import nn\nimport torch.nn.functional as F\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nfrom fairscale.nn.model_parallel.layers import (\n    ParallelEmbedding,\n    RowParallelLinear,\n    ColumnParallelLinear,\n)\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.RMSNorm","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama.model.RMSNorm#L33-L44","kind":"class","name":"RMSNorm","path":"gorilla/inference/llama/model.py","language":"python","start_line":33,"end_line":44,"context_start_line":13,"context_end_line":64,"code":"from fairscale.nn.model_parallel.layers import (\n    ParallelEmbedding,\n    RowParallelLinear,\n    ColumnParallelLinear,\n)\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.precompute_freqs_cis","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.model.precompute_freqs_cis#L47-L52","kind":"function","name":"precompute_freqs_cis","path":"gorilla/inference/llama/model.py","language":"python","start_line":47,"end_line":52,"context_start_line":27,"context_end_line":72,"code":"    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.reshape_for_broadcast","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.model.reshape_for_broadcast#L55-L60","kind":"function","name":"reshape_for_broadcast","path":"gorilla/inference/llama/model.py","language":"python","start_line":55,"end_line":60,"context_start_line":35,"context_end_line":80,"code":"        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.apply_rotary_emb","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.model.apply_rotary_emb#L63-L73","kind":"function","name":"apply_rotary_emb","path":"gorilla/inference/llama/model.py","language":"python","start_line":63,"end_line":73,"context_start_line":43,"context_end_line":93,"code":"        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.Attention","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama.model.Attention#L76-L150","kind":"class","name":"Attention","path":"gorilla/inference/llama/model.py","language":"python","start_line":76,"end_line":150,"context_start_line":56,"context_end_line":170,"code":"    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wv = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wo = RowParallelLinear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=False,\n            input_is_parallel=True,\n            init_method=lambda x: x,\n        )\n\n        self.cache_k = torch.zeros(\n            (args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)\n        ).cuda()\n        self.cache_v = torch.zeros(\n            (args.max_batch_size, args.max_seq_len, self.n_local_heads, self.head_dim)\n        ).cuda()\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor]):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        self.cache_k = self.cache_k.to(xq)\n        self.cache_v = self.cache_v.to(xq)\n\n        self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk\n        self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n        keys = self.cache_k[:bsz, : start_pos + seqlen]\n        values = self.cache_v[:bsz, : start_pos + seqlen]\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x\n        )\n        self.w2 = RowParallelLinear(\n            hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x\n        )\n        self.w3 = ColumnParallelLinear(","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.FeedForward","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama.model.FeedForward#L153-L175","kind":"class","name":"FeedForward","path":"gorilla/inference/llama/model.py","language":"python","start_line":153,"end_line":175,"context_start_line":133,"context_end_line":195,"code":"        self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n        keys = self.cache_k[:bsz, : start_pos + seqlen]\n        values = self.cache_v[:bsz, : start_pos + seqlen]\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x\n        )\n        self.w2 = RowParallelLinear(\n            hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x\n        )\n        self.w3 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x\n        )\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor]):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.TransformerBlock","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama.model.TransformerBlock#L178-L195","kind":"class","name":"TransformerBlock","path":"gorilla/inference/llama/model.py","language":"python","start_line":178,"end_line":195,"context_start_line":158,"context_end_line":215,"code":"        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x\n        )\n        self.w2 = RowParallelLinear(\n            hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x\n        )\n        self.w3 = ColumnParallelLinear(\n            dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x\n        )\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor]):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=lambda x: x\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(\n            params.dim, params.vocab_size, bias=False, init_method=lambda x: x","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.Transformer","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama.model.Transformer#L198-L238","kind":"class","name":"Transformer","path":"gorilla/inference/llama/model.py","language":"python","start_line":198,"end_line":238,"context_start_line":178,"context_end_line":238,"code":"class TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor]):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=lambda x: x\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(\n            params.dim, params.vocab_size, bias=False, init_method=lambda x: x\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.__init__","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.model.__init__#L199-L220","kind":"function","name":"__init__","path":"gorilla/inference/llama/model.py","language":"python","start_line":199,"end_line":220,"context_start_line":179,"context_end_line":238,"code":"    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor]):\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=lambda x: x\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(\n            params.dim, params.vocab_size, bias=False, init_method=lambda x: x\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model._norm","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.model._norm#L39-L40","kind":"function","name":"_norm","path":"gorilla/inference/llama/model.py","language":"python","start_line":39,"end_line":40,"context_start_line":19,"context_end_line":60,"code":"\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.model.forward","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.model.forward#L223-L238","kind":"function","name":"forward","path":"gorilla/inference/llama/model.py","language":"python","start_line":223,"end_line":238,"context_start_line":203,"context_end_line":238,"code":"        self.n_layers = params.n_layers\n\n        self.tok_embeddings = ParallelEmbedding(\n            params.vocab_size, params.dim, init_method=lambda x: x\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(\n            params.dim, params.vocab_size, bias=False, init_method=lambda x: x\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.tokenizer","uri":"program://LLaMA-Adapter/module/gorilla.inference.llama.tokenizer#L1-L40","kind":"module","name":"gorilla.inference.llama.tokenizer","path":"gorilla/inference/llama/tokenizer.py","language":"python","start_line":1,"end_line":40,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.tokenizer.Tokenizer","uri":"program://LLaMA-Adapter/class/gorilla.inference.llama.tokenizer.Tokenizer#L13-L40","kind":"class","name":"Tokenizer","path":"gorilla/inference/llama/tokenizer.py","language":"python","start_line":13,"end_line":40,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.tokenizer.__init__","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.tokenizer.__init__#L14-L28","kind":"function","name":"__init__","path":"gorilla/inference/llama/tokenizer.py","language":"python","start_line":14,"end_line":28,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.tokenizer.encode","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.tokenizer.encode#L30-L37","kind":"function","name":"encode","path":"gorilla/inference/llama/tokenizer.py","language":"python","start_line":30,"end_line":37,"context_start_line":10,"context_end_line":40,"code":"logger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.inference.llama.tokenizer.decode","uri":"program://LLaMA-Adapter/function/gorilla.inference.llama.tokenizer.decode#L39-L40","kind":"function","name":"decode","path":"gorilla/inference/llama/tokenizer.py","language":"python","start_line":39,"end_line":40,"context_start_line":19,"context_end_line":40,"code":"\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.extract_adapter_from_checkpoint","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.extract_adapter_from_checkpoint#L1-L23","kind":"module","name":"gorilla.alpaca_finetuning_v1.extract_adapter_from_checkpoint","path":"gorilla/alpaca_finetuning_v1/extract_adapter_from_checkpoint.py","language":"python","start_line":1,"end_line":23,"context_start_line":1,"context_end_line":23,"code":"import torch\nimport argparse\n\nargs = argparse.ArgumentParser(\"extract\", add_help=False)\n\nargs.add_argument(\"--model_path\", type=str)\n\nargs = args.parse_args()\n\nmodel = torch.load(args.model_path, map_location=\"cpu\")\nnew_model = dict()\nweight_list = [\"layers.\" + str(i) + \".attention.gate\" for i in range(32)]\nold_weight_list = [\"layers.\" + str(i) + \".attention.gate\" for i in range(32)]\nweight_list = weight_list + [\"adapter_query.weight\"]\n\nprint(weight_list)\nprint(model[\"model\"][\"adapter_query.weight\"].shape)\n\nfor i in range(len(weight_list)):\n    new_model[weight_list[i]] = model[\"model\"][weight_list[i]]\n\nsave_path = args.model_path.replace('.pth', '-adapter.pth')\ntorch.save(new_model, save_path)","source_hash":"df2af3a9757f423ca8e8f05de35694c99a88a4a334f8c787e5481766a6e0f7f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.models_llama_adapter","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.models_llama_adapter#L1-L50","kind":"module","name":"gorilla.alpaca_finetuning_v1.models_llama_adapter","path":"gorilla/alpaca_finetuning_v1/models_llama_adapter.py","language":"python","start_line":1,"end_line":50,"context_start_line":1,"context_end_line":50,"code":"import json\n\nimport torch\n\nfrom llama import ModelArgs, Tokenizer, Transformer\n\n\ndef Llama7B_adapter(args, **kwargs):\n\n    llama_model_path = args.llama_model_path\n    model_name = \"7B\"\n\n    checkpoint = torch.load(llama_model_path + \"/consolidated.00.pth\", map_location=\"cpu\")\n    print(llama_model_path + \"/consolidated.00.pth\")\n\n    with open(llama_model_path + \"/params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=args.max_seq_len,\n        max_batch_size=32,\n        adapter_len=args.adapter_len,\n        adapter_layer=args.adapter_layer,\n        **params\n    )\n    tokenizer = Tokenizer(model_path=llama_model_path + \"/tokenizer.model\")\n\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model_llama_adapter = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    model_llama_adapter.load_state_dict(checkpoint, strict=False)\n\n    for name, param in model_llama_adapter.named_parameters():\n        if \"adapter\" not in name:\n            param.requires_grad = False\n        else:\n            param.requires_grad = True\n            param.data = param.data.float()\n\n    for name, param in model_llama_adapter.layers[-1 * args.adapter_layer :].named_parameters():\n        if \"gate\" in name or \"adapter\" in name:\n            param.data = param.data.float()\n            param.requires_grad = True\n\n    return model_llama_adapter\n\n\n# set recommended archs\nLlama7B_adapter = Llama7B_adapter","source_hash":"4e63b524896d6fd5c98dda7936b21db83f656d8eab48fbc3261b6709eaf69261","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.models_llama_adapter.Llama7B_adapter","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.models_llama_adapter.Llama7B_adapter#L8-L46","kind":"function","name":"Llama7B_adapter","path":"gorilla/alpaca_finetuning_v1/models_llama_adapter.py","language":"python","start_line":8,"end_line":46,"context_start_line":1,"context_end_line":50,"code":"import json\n\nimport torch\n\nfrom llama import ModelArgs, Tokenizer, Transformer\n\n\ndef Llama7B_adapter(args, **kwargs):\n\n    llama_model_path = args.llama_model_path\n    model_name = \"7B\"\n\n    checkpoint = torch.load(llama_model_path + \"/consolidated.00.pth\", map_location=\"cpu\")\n    print(llama_model_path + \"/consolidated.00.pth\")\n\n    with open(llama_model_path + \"/params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=args.max_seq_len,\n        max_batch_size=32,\n        adapter_len=args.adapter_len,\n        adapter_layer=args.adapter_layer,\n        **params\n    )\n    tokenizer = Tokenizer(model_path=llama_model_path + \"/tokenizer.model\")\n\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model_llama_adapter = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    model_llama_adapter.load_state_dict(checkpoint, strict=False)\n\n    for name, param in model_llama_adapter.named_parameters():\n        if \"adapter\" not in name:\n            param.requires_grad = False\n        else:\n            param.requires_grad = True\n            param.data = param.data.float()\n\n    for name, param in model_llama_adapter.layers[-1 * args.adapter_layer :].named_parameters():\n        if \"gate\" in name or \"adapter\" in name:\n            param.data = param.data.float()\n            param.requires_grad = True\n\n    return model_llama_adapter\n\n\n# set recommended archs\nLlama7B_adapter = Llama7B_adapter","source_hash":"4e63b524896d6fd5c98dda7936b21db83f656d8eab48fbc3261b6709eaf69261","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.engine_finetuning","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.engine_finetuning#L1-L132","kind":"module","name":"gorilla.alpaca_finetuning_v1.engine_finetuning","path":"gorilla/alpaca_finetuning_v1/engine_finetuning.py","language":"python","start_line":1,"end_line":132,"context_start_line":1,"context_end_line":132,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\nimport util.lr_sched as lr_sched\nimport util.misc as misc\n\n\ndef train_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n\n    model.train(True)\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter(\"lr\", misc.SmoothedValue(window_size=1, fmt=\"{value:.6f}\"))\n    header = \"Epoch: [{}]\".format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print(\"log_dir: {}\".format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header)\n    ):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        c_loss = model(examples, labels)\n        loss = c_loss\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n\n        loss_scaler(loss, optimizer, parameters=model.parameters(), update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\"We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar(\"c_train_loss\", c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar(\"lr\", lr, epoch_1000x)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n\ndef val_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter(\"lr\", misc.SmoothedValue(window_size=1, fmt=\"{value:.6f}\"))\n    header = \"Epoch: [{}]\".format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    if log_writer is not None:\n        print(\"log_dir: {}\".format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header)\n    ):\n\n        with torch.no_grad():\n            c_loss = model(examples, labels)\n        loss = c_loss\n        loss_value = loss.item()\n\n        c_loss_value = c_loss.item()\n\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        metric_logger.update(closs=c_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\"We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar(\"c_train_loss\", c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar(\"lr\", lr, epoch_1000x)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"e90ffdf70e0a5bc1217c32533638a96c91df1ec263f2805e7eecde3789ceb5a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.engine_finetuning.train_one_epoch","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.engine_finetuning.train_one_epoch#L10-L76","kind":"function","name":"train_one_epoch","path":"gorilla/alpaca_finetuning_v1/engine_finetuning.py","language":"python","start_line":10,"end_line":76,"context_start_line":1,"context_end_line":96,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\nimport util.lr_sched as lr_sched\nimport util.misc as misc\n\n\ndef train_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n\n    model.train(True)\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter(\"lr\", misc.SmoothedValue(window_size=1, fmt=\"{value:.6f}\"))\n    header = \"Epoch: [{}]\".format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print(\"log_dir: {}\".format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header)\n    ):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        c_loss = model(examples, labels)\n        loss = c_loss\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n\n        loss_scaler(loss, optimizer, parameters=model.parameters(), update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\"We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar(\"c_train_loss\", c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar(\"lr\", lr, epoch_1000x)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n\ndef val_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter(\"lr\", misc.SmoothedValue(window_size=1, fmt=\"{value:.6f}\"))\n    header = \"Epoch: [{}]\".format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n","source_hash":"e90ffdf70e0a5bc1217c32533638a96c91df1ec263f2805e7eecde3789ceb5a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.engine_finetuning.val_one_epoch","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.engine_finetuning.val_one_epoch#L79-L132","kind":"function","name":"val_one_epoch","path":"gorilla/alpaca_finetuning_v1/engine_finetuning.py","language":"python","start_line":79,"end_line":132,"context_start_line":59,"context_end_line":132,"code":"        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\"We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar(\"c_train_loss\", c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar(\"lr\", lr, epoch_1000x)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}\n\n\ndef val_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n    model.eval()\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter(\"lr\", misc.SmoothedValue(window_size=1, fmt=\"{value:.6f}\"))\n    header = \"Epoch: [{}]\".format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    if log_writer is not None:\n        print(\"log_dir: {}\".format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask) in enumerate(\n        metric_logger.log_every(data_loader, print_freq, header)\n    ):\n\n        with torch.no_grad():\n            c_loss = model(examples, labels)\n        loss = c_loss\n        loss_value = loss.item()\n\n        c_loss_value = c_loss.item()\n\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        metric_logger.update(closs=c_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\"We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar(\"c_train_loss\", c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar(\"lr\", lr, epoch_1000x)\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"e90ffdf70e0a5bc1217c32533638a96c91df1ec263f2805e7eecde3789ceb5a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.finetuning","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.finetuning#L1-L306","kind":"module","name":"gorilla.alpaca_finetuning_v1.finetuning","path":"gorilla/alpaca_finetuning_v1/finetuning.py","language":"python","start_line":1,"end_line":306,"context_start_line":1,"context_end_line":306,"code":"import argparse\nimport copy\nimport datetime\nimport json\nimport os\nimport time\nfrom pathlib import Path\n\nimport models_llama_adapter\nimport numpy as np\nimport timm.optim.optim_factory as optim_factory\nimport torch\nimport torch.backends.cudnn as cudnn\nimport util.misc as misc\nfrom engine_finetuning import train_one_epoch, val_one_epoch\nfrom torch.utils.data import Dataset\nfrom torch.utils.tensorboard import SummaryWriter\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\n\nfrom llama import Tokenizer\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\nclass InstructionDataset(Dataset):\n    def __init__(self, data_path, model_path, max_words=30, partition=\"train\"):\n        ann = []\n        data_lists = list(open(data_path))\n        for data in data_lists:\n            data = json.loads(data)['code']\n            if '###Instruction.' in data:\n                data = data.replace('###Instruction.', '###Instruction:')\n            if 'Output:' in data:\n               ann.append({'instruction':data.split('Output:')[0].split('Instruction:')[1].split('###')[0],\n                           'input': '',\n                           'output': data.split('Output:')[1]})\n        self.ann = ann\n\n        # self.ann = json.load(open(data_path))\n        if partition == \"train\":\n            self.ann = self.ann\n        else:\n            self.ann = self.ann[:200]\n\n        self.max_words = max_words\n        tokenizer = Tokenizer(model_path=model_path + \"./tokenizer.model\")\n        self.tokenizer1 = tokenizer\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n\n        ann = self.ann[index]\n        if ann.get(\"input\", \"\") == \"\":\n            prompt = PROMPT_DICT[\"prompt_no_input\"].format_map(ann)\n        else:\n            prompt = PROMPT_DICT[\"prompt_input\"].format_map(ann)\n        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)\n        labels[: len(prompt)] = -1\n        example_mask = example.ge(0)\n        label_mask = labels.ge(0)\n        example[~example_mask] = 0\n        labels[~label_mask] = 0\n        example_mask = example_mask.float()\n        label_mask = label_mask.float()\n\n        return example, labels, example_mask\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser(\"MAE pre-training\", add_help=False)\n    parser.add_argument(\n        \"--batch_size\",\n        default=64,\n        type=int,\n        help=\"Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus\",\n    )\n    parser.add_argument(\"--epochs\", default=400, type=int)\n    parser.add_argument(\n        \"--accum_iter\",\n        default=1,\n        type=int,\n        help=\"Accumulate gradient iterations (for increasing the effective batch size under memory constraints)\",\n    )\n\n    # Model parameters\n    parser.add_argument(\"--llama_model_path\", default=\"./llama\", type=str, help=\"path of llama model\")\n    parser.add_argument(\"--model\", default=\"llama7B_adapter\", type=str, metavar=\"MODEL\", help=\"Name of model to train\")\n\n    parser.add_argument(\"--adapter_layer\", type=int, default=30, metavar=\"LENGTH\", help=\"the number of adapter layer\")\n\n    parser.add_argument(\"--adapter_len\", type=int, default=10, metavar=\"LENGTH\", help=\"the adapter length\")\n\n    parser.add_argument(\"--max_seq_len\", type=int, default=512, metavar=\"LENGTH\", help=\"the maximum sequence length\")\n\n    # Optimizer parameters\n    parser.add_argument(\"--weight_decay\", type=float, default=0.05, help=\"weight decay (default: 0.05)\")\n\n    parser.add_argument(\"--lr\", type=float, default=None, metavar=\"LR\", help=\"learning rate (absolute lr)\")\n    parser.add_argument(\n        \"--blr\",\n        type=float,\n        default=1e-3,\n        metavar=\"LR\",\n        help=\"base learning rate: absolute_lr = base_lr * total_batch_size / 256\",\n    )\n    parser.add_argument(\n        \"--min_lr\", type=float, default=0.0, metavar=\"LR\", help=\"lower lr bound for cyclic schedulers that hit 0\"\n    )\n\n    parser.add_argument(\"--warmup_epochs\", type=int, default=40, metavar=\"N\", help=\"epochs to warmup LR\")\n\n    # Dataset parameters\n    parser.add_argument(\"--data_path\", default=\"/instruction_dataset/\", type=str, help=\"dataset path\")\n\n    parser.add_argument(\"--output_dir\", default=\"./output_dir\", help=\"path where to save, empty for no saving\")\n    parser.add_argument(\"--log_dir\", default=\"./output_dir\", help=\"path where to tensorboard log\")\n    parser.add_argument(\"--device\", default=\"cuda\", help=\"device to use for training / testing\")\n    parser.add_argument(\"--seed\", default=0, type=int)\n    parser.add_argument(\"--resume\", default=\"\", help=\"resume from checkpoint\")\n\n    parser.add_argument(\"--start_epoch\", default=0, type=int, metavar=\"N\", help=\"start epoch\")\n    parser.add_argument(\"--num_workers\", default=10, type=int)\n    parser.add_argument(\n        \"--pin_mem\",\n        action=\"store_true\",\n        help=\"Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.\",\n    )\n    parser.add_argument(\"--no_pin_mem\", action=\"store_false\", dest=\"pin_mem\")\n    parser.set_defaults(pin_mem=True)\n\n    # distributed training parameters\n    parser.add_argument(\"--world_size\", default=1, type=int, help=\"number of distributed processes\")\n    parser.add_argument(\"--local_rank\", default=-1, type=int)\n    parser.add_argument(\"--dist_on_itp\", action=\"store_true\")\n    parser.add_argument(\"--dist_url\", default=\"env://\", help=\"url used to set up distributed training\")\n\n    return parser\n\n\ndef main(args):\n\n    misc.init_distributed_mode(args)\n\n    print(\"job dir: {}\".format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(\", \", \",\\n\"))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n\n    cudnn.benchmark = True\n\n    dataset_train = InstructionDataset(\n        data_path=args.data_path, model_path=args.llama_model_path, max_words=args.max_seq_len, partition=\"train\"\n    )\n    dataset_val = InstructionDataset(\n        data_path=args.data_path, model_path=args.llama_model_path, max_words=args.max_seq_len, partition=\"val\"\n    )\n\n    print(dataset_train)\n    print(dataset_val)\n\n    if True:  # args.distributed:\n        num_tasks = misc.get_world_size()\n        global_rank = misc.get_rank()\n        sampler_train = torch.utils.data.DistributedSampler(\n            dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True\n        )\n\n        sampler_val = torch.utils.data.DistributedSampler(\n            dataset_val, num_replicas=num_tasks, rank=global_rank, shuffle=True\n        )\n\n        print(\"Sampler_train = %s\" % str(sampler_train))\n    else:\n        sampler_train = torch.utils.data.RandomSampler(dataset_train)\n\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train,\n        sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    data_loader_val = torch.utils.data.DataLoader(\n        dataset_val,\n        sampler=sampler_val,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # define the model\n    model = models_llama_adapter.__dict__[args.model](args)\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = optim_factory.param_groups_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n    misc.load_model(args=args, model_without_ddp=model_without_ddp, optimizer=optimizer, loss_scaler=loss_scaler)\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n            data_loader_val.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train, optimizer, device, epoch, loss_scaler, log_writer=log_writer, args=args\n        )\n\n        val_stats = val_one_epoch(\n            model, data_loader_val, optimizer, device, epoch, loss_scaler, log_writer=log_writer, args=args\n        )\n\n        if args.output_dir and (epoch % 8 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args,\n                model=model,\n                model_without_ddp=model_without_ddp,\n                optimizer=optimizer,\n                loss_scaler=loss_scaler,\n                epoch=epoch,\n            )\n\n        log_stats = {\n            **{f\"train_{k}\": v for k, v in train_stats.items()},\n            \"epoch\": epoch,\n            **{f\"val_{k}\": v for k, v in val_stats.items()},\n        }\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print(\"Training time {}\".format(total_time_str))\n\n\nif __name__ == \"__main__\":\n\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"9b86155ada31fbcad2a44259b26833df69bf9b5674e2f9924e077e6e06fc5474","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.finetuning.InstructionDataset","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.finetuning.InstructionDataset#L36-L87","kind":"class","name":"InstructionDataset","path":"gorilla/alpaca_finetuning_v1/finetuning.py","language":"python","start_line":36,"end_line":87,"context_start_line":16,"context_end_line":107,"code":"from torch.utils.data import Dataset\nfrom torch.utils.tensorboard import SummaryWriter\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\n\nfrom llama import Tokenizer\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\nclass InstructionDataset(Dataset):\n    def __init__(self, data_path, model_path, max_words=30, partition=\"train\"):\n        ann = []\n        data_lists = list(open(data_path))\n        for data in data_lists:\n            data = json.loads(data)['code']\n            if '###Instruction.' in data:\n                data = data.replace('###Instruction.', '###Instruction:')\n            if 'Output:' in data:\n               ann.append({'instruction':data.split('Output:')[0].split('Instruction:')[1].split('###')[0],\n                           'input': '',\n                           'output': data.split('Output:')[1]})\n        self.ann = ann\n\n        # self.ann = json.load(open(data_path))\n        if partition == \"train\":\n            self.ann = self.ann\n        else:\n            self.ann = self.ann[:200]\n\n        self.max_words = max_words\n        tokenizer = Tokenizer(model_path=model_path + \"./tokenizer.model\")\n        self.tokenizer1 = tokenizer\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n\n        ann = self.ann[index]\n        if ann.get(\"input\", \"\") == \"\":\n            prompt = PROMPT_DICT[\"prompt_no_input\"].format_map(ann)\n        else:\n            prompt = PROMPT_DICT[\"prompt_input\"].format_map(ann)\n        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)\n        labels[: len(prompt)] = -1\n        example_mask = example.ge(0)\n        label_mask = labels.ge(0)\n        example[~example_mask] = 0\n        labels[~label_mask] = 0\n        example_mask = example_mask.float()\n        label_mask = label_mask.float()\n\n        return example, labels, example_mask\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser(\"MAE pre-training\", add_help=False)\n    parser.add_argument(\n        \"--batch_size\",\n        default=64,\n        type=int,\n        help=\"Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus\",\n    )\n    parser.add_argument(\"--epochs\", default=400, type=int)\n    parser.add_argument(\n        \"--accum_iter\",\n        default=1,\n        type=int,\n        help=\"Accumulate gradient iterations (for increasing the effective batch size under memory constraints)\",\n    )\n\n    # Model parameters\n    parser.add_argument(\"--llama_model_path\", default=\"./llama\", type=str, help=\"path of llama model\")","source_hash":"9b86155ada31fbcad2a44259b26833df69bf9b5674e2f9924e077e6e06fc5474","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.finetuning.get_args_parser","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.finetuning.get_args_parser#L90-L158","kind":"function","name":"get_args_parser","path":"gorilla/alpaca_finetuning_v1/finetuning.py","language":"python","start_line":90,"end_line":158,"context_start_line":70,"context_end_line":178,"code":"        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)\n        labels[: len(prompt)] = -1\n        example_mask = example.ge(0)\n        label_mask = labels.ge(0)\n        example[~example_mask] = 0\n        labels[~label_mask] = 0\n        example_mask = example_mask.float()\n        label_mask = label_mask.float()\n\n        return example, labels, example_mask\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser(\"MAE pre-training\", add_help=False)\n    parser.add_argument(\n        \"--batch_size\",\n        default=64,\n        type=int,\n        help=\"Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus\",\n    )\n    parser.add_argument(\"--epochs\", default=400, type=int)\n    parser.add_argument(\n        \"--accum_iter\",\n        default=1,\n        type=int,\n        help=\"Accumulate gradient iterations (for increasing the effective batch size under memory constraints)\",\n    )\n\n    # Model parameters\n    parser.add_argument(\"--llama_model_path\", default=\"./llama\", type=str, help=\"path of llama model\")\n    parser.add_argument(\"--model\", default=\"llama7B_adapter\", type=str, metavar=\"MODEL\", help=\"Name of model to train\")\n\n    parser.add_argument(\"--adapter_layer\", type=int, default=30, metavar=\"LENGTH\", help=\"the number of adapter layer\")\n\n    parser.add_argument(\"--adapter_len\", type=int, default=10, metavar=\"LENGTH\", help=\"the adapter length\")\n\n    parser.add_argument(\"--max_seq_len\", type=int, default=512, metavar=\"LENGTH\", help=\"the maximum sequence length\")\n\n    # Optimizer parameters\n    parser.add_argument(\"--weight_decay\", type=float, default=0.05, help=\"weight decay (default: 0.05)\")\n\n    parser.add_argument(\"--lr\", type=float, default=None, metavar=\"LR\", help=\"learning rate (absolute lr)\")\n    parser.add_argument(\n        \"--blr\",\n        type=float,\n        default=1e-3,\n        metavar=\"LR\",\n        help=\"base learning rate: absolute_lr = base_lr * total_batch_size / 256\",\n    )\n    parser.add_argument(\n        \"--min_lr\", type=float, default=0.0, metavar=\"LR\", help=\"lower lr bound for cyclic schedulers that hit 0\"\n    )\n\n    parser.add_argument(\"--warmup_epochs\", type=int, default=40, metavar=\"N\", help=\"epochs to warmup LR\")\n\n    # Dataset parameters\n    parser.add_argument(\"--data_path\", default=\"/instruction_dataset/\", type=str, help=\"dataset path\")\n\n    parser.add_argument(\"--output_dir\", default=\"./output_dir\", help=\"path where to save, empty for no saving\")\n    parser.add_argument(\"--log_dir\", default=\"./output_dir\", help=\"path where to tensorboard log\")\n    parser.add_argument(\"--device\", default=\"cuda\", help=\"device to use for training / testing\")\n    parser.add_argument(\"--seed\", default=0, type=int)\n    parser.add_argument(\"--resume\", default=\"\", help=\"resume from checkpoint\")\n\n    parser.add_argument(\"--start_epoch\", default=0, type=int, metavar=\"N\", help=\"start epoch\")\n    parser.add_argument(\"--num_workers\", default=10, type=int)\n    parser.add_argument(\n        \"--pin_mem\",\n        action=\"store_true\",\n        help=\"Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.\",\n    )\n    parser.add_argument(\"--no_pin_mem\", action=\"store_false\", dest=\"pin_mem\")\n    parser.set_defaults(pin_mem=True)\n\n    # distributed training parameters\n    parser.add_argument(\"--world_size\", default=1, type=int, help=\"number of distributed processes\")\n    parser.add_argument(\"--local_rank\", default=-1, type=int)\n    parser.add_argument(\"--dist_on_itp\", action=\"store_true\")\n    parser.add_argument(\"--dist_url\", default=\"env://\", help=\"url used to set up distributed training\")\n\n    return parser\n\n\ndef main(args):\n\n    misc.init_distributed_mode(args)\n\n    print(\"job dir: {}\".format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(\", \", \",\\n\"))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n\n    cudnn.benchmark = True\n\n    dataset_train = InstructionDataset(\n        data_path=args.data_path, model_path=args.llama_model_path, max_words=args.max_seq_len, partition=\"train\"","source_hash":"9b86155ada31fbcad2a44259b26833df69bf9b5674e2f9924e077e6e06fc5474","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.finetuning.main","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.finetuning.main#L161-L297","kind":"function","name":"main","path":"gorilla/alpaca_finetuning_v1/finetuning.py","language":"python","start_line":161,"end_line":297,"context_start_line":141,"context_end_line":306,"code":"\n    parser.add_argument(\"--start_epoch\", default=0, type=int, metavar=\"N\", help=\"start epoch\")\n    parser.add_argument(\"--num_workers\", default=10, type=int)\n    parser.add_argument(\n        \"--pin_mem\",\n        action=\"store_true\",\n        help=\"Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.\",\n    )\n    parser.add_argument(\"--no_pin_mem\", action=\"store_false\", dest=\"pin_mem\")\n    parser.set_defaults(pin_mem=True)\n\n    # distributed training parameters\n    parser.add_argument(\"--world_size\", default=1, type=int, help=\"number of distributed processes\")\n    parser.add_argument(\"--local_rank\", default=-1, type=int)\n    parser.add_argument(\"--dist_on_itp\", action=\"store_true\")\n    parser.add_argument(\"--dist_url\", default=\"env://\", help=\"url used to set up distributed training\")\n\n    return parser\n\n\ndef main(args):\n\n    misc.init_distributed_mode(args)\n\n    print(\"job dir: {}\".format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(\", \", \",\\n\"))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n\n    cudnn.benchmark = True\n\n    dataset_train = InstructionDataset(\n        data_path=args.data_path, model_path=args.llama_model_path, max_words=args.max_seq_len, partition=\"train\"\n    )\n    dataset_val = InstructionDataset(\n        data_path=args.data_path, model_path=args.llama_model_path, max_words=args.max_seq_len, partition=\"val\"\n    )\n\n    print(dataset_train)\n    print(dataset_val)\n\n    if True:  # args.distributed:\n        num_tasks = misc.get_world_size()\n        global_rank = misc.get_rank()\n        sampler_train = torch.utils.data.DistributedSampler(\n            dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True\n        )\n\n        sampler_val = torch.utils.data.DistributedSampler(\n            dataset_val, num_replicas=num_tasks, rank=global_rank, shuffle=True\n        )\n\n        print(\"Sampler_train = %s\" % str(sampler_train))\n    else:\n        sampler_train = torch.utils.data.RandomSampler(dataset_train)\n\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train,\n        sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    data_loader_val = torch.utils.data.DataLoader(\n        dataset_val,\n        sampler=sampler_val,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # define the model\n    model = models_llama_adapter.__dict__[args.model](args)\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = optim_factory.param_groups_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n    misc.load_model(args=args, model_without_ddp=model_without_ddp, optimizer=optimizer, loss_scaler=loss_scaler)\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n            data_loader_val.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train, optimizer, device, epoch, loss_scaler, log_writer=log_writer, args=args\n        )\n\n        val_stats = val_one_epoch(\n            model, data_loader_val, optimizer, device, epoch, loss_scaler, log_writer=log_writer, args=args\n        )\n\n        if args.output_dir and (epoch % 8 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args,\n                model=model,\n                model_without_ddp=model_without_ddp,\n                optimizer=optimizer,\n                loss_scaler=loss_scaler,\n                epoch=epoch,\n            )\n\n        log_stats = {\n            **{f\"train_{k}\": v for k, v in train_stats.items()},\n            \"epoch\": epoch,\n            **{f\"val_{k}\": v for k, v in val_stats.items()},\n        }\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print(\"Training time {}\".format(total_time_str))\n\n\nif __name__ == \"__main__\":\n\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"9b86155ada31fbcad2a44259b26833df69bf9b5674e2f9924e077e6e06fc5474","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.finetuning.__init__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.finetuning.__init__#L37-L58","kind":"function","name":"__init__","path":"gorilla/alpaca_finetuning_v1/finetuning.py","language":"python","start_line":37,"end_line":58,"context_start_line":17,"context_end_line":78,"code":"from torch.utils.tensorboard import SummaryWriter\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\n\nfrom llama import Tokenizer\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n\nclass InstructionDataset(Dataset):\n    def __init__(self, data_path, model_path, max_words=30, partition=\"train\"):\n        ann = []\n        data_lists = list(open(data_path))\n        for data in data_lists:\n            data = json.loads(data)['code']\n            if '###Instruction.' in data:\n                data = data.replace('###Instruction.', '###Instruction:')\n            if 'Output:' in data:\n               ann.append({'instruction':data.split('Output:')[0].split('Instruction:')[1].split('###')[0],\n                           'input': '',\n                           'output': data.split('Output:')[1]})\n        self.ann = ann\n\n        # self.ann = json.load(open(data_path))\n        if partition == \"train\":\n            self.ann = self.ann\n        else:\n            self.ann = self.ann[:200]\n\n        self.max_words = max_words\n        tokenizer = Tokenizer(model_path=model_path + \"./tokenizer.model\")\n        self.tokenizer1 = tokenizer\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n\n        ann = self.ann[index]\n        if ann.get(\"input\", \"\") == \"\":\n            prompt = PROMPT_DICT[\"prompt_no_input\"].format_map(ann)\n        else:\n            prompt = PROMPT_DICT[\"prompt_input\"].format_map(ann)\n        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)","source_hash":"9b86155ada31fbcad2a44259b26833df69bf9b5674e2f9924e077e6e06fc5474","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.finetuning.__len__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.finetuning.__len__#L60-L61","kind":"function","name":"__len__","path":"gorilla/alpaca_finetuning_v1/finetuning.py","language":"python","start_line":60,"end_line":61,"context_start_line":40,"context_end_line":81,"code":"        for data in data_lists:\n            data = json.loads(data)['code']\n            if '###Instruction.' in data:\n                data = data.replace('###Instruction.', '###Instruction:')\n            if 'Output:' in data:\n               ann.append({'instruction':data.split('Output:')[0].split('Instruction:')[1].split('###')[0],\n                           'input': '',\n                           'output': data.split('Output:')[1]})\n        self.ann = ann\n\n        # self.ann = json.load(open(data_path))\n        if partition == \"train\":\n            self.ann = self.ann\n        else:\n            self.ann = self.ann[:200]\n\n        self.max_words = max_words\n        tokenizer = Tokenizer(model_path=model_path + \"./tokenizer.model\")\n        self.tokenizer1 = tokenizer\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n\n        ann = self.ann[index]\n        if ann.get(\"input\", \"\") == \"\":\n            prompt = PROMPT_DICT[\"prompt_no_input\"].format_map(ann)\n        else:\n            prompt = PROMPT_DICT[\"prompt_input\"].format_map(ann)\n        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)\n        labels[: len(prompt)] = -1\n        example_mask = example.ge(0)\n        label_mask = labels.ge(0)","source_hash":"9b86155ada31fbcad2a44259b26833df69bf9b5674e2f9924e077e6e06fc5474","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.finetuning.__getitem__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.finetuning.__getitem__#L63-L87","kind":"function","name":"__getitem__","path":"gorilla/alpaca_finetuning_v1/finetuning.py","language":"python","start_line":63,"end_line":87,"context_start_line":43,"context_end_line":107,"code":"                data = data.replace('###Instruction.', '###Instruction:')\n            if 'Output:' in data:\n               ann.append({'instruction':data.split('Output:')[0].split('Instruction:')[1].split('###')[0],\n                           'input': '',\n                           'output': data.split('Output:')[1]})\n        self.ann = ann\n\n        # self.ann = json.load(open(data_path))\n        if partition == \"train\":\n            self.ann = self.ann\n        else:\n            self.ann = self.ann[:200]\n\n        self.max_words = max_words\n        tokenizer = Tokenizer(model_path=model_path + \"./tokenizer.model\")\n        self.tokenizer1 = tokenizer\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n\n        ann = self.ann[index]\n        if ann.get(\"input\", \"\") == \"\":\n            prompt = PROMPT_DICT[\"prompt_no_input\"].format_map(ann)\n        else:\n            prompt = PROMPT_DICT[\"prompt_input\"].format_map(ann)\n        example = prompt + ann[\"output\"]\n        prompt = torch.tensor(self.tokenizer1.encode(prompt, bos=True, eos=False), dtype=torch.int64)\n        example = torch.tensor(self.tokenizer1.encode(example, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - example.shape[0]\n        if padding > 0:\n            example = torch.cat((example, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            example = example[: self.max_words]\n        labels = copy.deepcopy(example)\n        labels[: len(prompt)] = -1\n        example_mask = example.ge(0)\n        label_mask = labels.ge(0)\n        example[~example_mask] = 0\n        labels[~label_mask] = 0\n        example_mask = example_mask.float()\n        label_mask = label_mask.float()\n\n        return example, labels, example_mask\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser(\"MAE pre-training\", add_help=False)\n    parser.add_argument(\n        \"--batch_size\",\n        default=64,\n        type=int,\n        help=\"Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus\",\n    )\n    parser.add_argument(\"--epochs\", default=400, type=int)\n    parser.add_argument(\n        \"--accum_iter\",\n        default=1,\n        type=int,\n        help=\"Accumulate gradient iterations (for increasing the effective batch size under memory constraints)\",\n    )\n\n    # Model parameters\n    parser.add_argument(\"--llama_model_path\", default=\"./llama\", type=str, help=\"path of llama model\")","source_hash":"9b86155ada31fbcad2a44259b26833df69bf9b5674e2f9924e077e6e06fc5474","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.util.misc#L1-L340","kind":"module","name":"gorilla.alpaca_finetuning_v1.util.misc","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":1,"end_line":340,"context_start_line":1,"context_end_line":340,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\n\nimport torch\nimport torch.distributed as dist\nfrom torch import inf\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.resume:\n        if args.resume.startswith('https'):\n            checkpoint = torch.hub.load_state_dict_from_url(\n                args.resume, map_location='cpu', check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location='cpu')\n        model_without_ddp.load_state_dict(checkpoint['model'])\n        print(\"Resume checkpoint %s\" % args.resume)\n        if 'optimizer' in checkpoint and 'epoch' in checkpoint and not (hasattr(args, 'eval') and args.eval):\n            optimizer.load_state_dict(checkpoint['optimizer'])\n            args.start_epoch = checkpoint['epoch'] + 1\n            if 'scaler' in checkpoint:\n                loss_scaler.load_state_dict(checkpoint['scaler'])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.SmoothedValue","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.util.misc.SmoothedValue#L24-L83","kind":"class","name":"SmoothedValue","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":24,"end_line":83,"context_start_line":4,"context_end_line":103,"code":"# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\n\nimport torch\nimport torch.distributed as dist\nfrom torch import inf\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.MetricLogger","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.util.misc.MetricLogger#L86-L167","kind":"class","name":"MetricLogger","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":86,"end_line":167,"context_start_line":66,"context_end_line":187,"code":"    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.setup_for_distributed","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.setup_for_distributed#L170-L184","kind":"function","name":"setup_for_distributed","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":170,"end_line":184,"context_start_line":150,"context_end_line":204,"code":"                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.is_dist_avail_and_initialized","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.is_dist_avail_and_initialized#L187-L192","kind":"function","name":"is_dist_avail_and_initialized","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":187,"end_line":192,"context_start_line":167,"context_end_line":212,"code":"            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.get_world_size","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.get_world_size#L195-L198","kind":"function","name":"get_world_size","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":195,"end_line":198,"context_start_line":175,"context_end_line":218,"code":"\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.get_rank","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.get_rank#L201-L204","kind":"function","name":"get_rank","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":201,"end_line":204,"context_start_line":181,"context_end_line":224,"code":"            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.is_main_process","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.is_main_process#L207-L208","kind":"function","name":"is_main_process","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":207,"end_line":208,"context_start_line":187,"context_end_line":228,"code":"def is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.save_on_master","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.save_on_master#L211-L213","kind":"function","name":"save_on_master","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":211,"end_line":213,"context_start_line":191,"context_end_line":233,"code":"        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.init_distributed_mode","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.init_distributed_mode#L216-L248","kind":"function","name":"init_distributed_mode","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":216,"end_line":248,"context_start_line":196,"context_end_line":268,"code":"    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.NativeScalerWithGradNormCount","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.util.misc.NativeScalerWithGradNormCount#L251-L277","kind":"class","name":"NativeScalerWithGradNormCount","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":251,"end_line":277,"context_start_line":231,"context_end_line":297,"code":"        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.get_grad_norm_","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.get_grad_norm_#L280-L292","kind":"function","name":"get_grad_norm_","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":280,"end_line":292,"context_start_line":260,"context_end_line":312,"code":"            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.save_model","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.save_model#L295-L312","kind":"function","name":"save_model","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":295,"end_line":312,"context_start_line":275,"context_end_line":332,"code":"\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.resume:\n        if args.resume.startswith('https'):\n            checkpoint = torch.hub.load_state_dict_from_url(\n                args.resume, map_location='cpu', check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location='cpu')\n        model_without_ddp.load_state_dict(checkpoint['model'])\n        print(\"Resume checkpoint %s\" % args.resume)\n        if 'optimizer' in checkpoint and 'epoch' in checkpoint and not (hasattr(args, 'eval') and args.eval):\n            optimizer.load_state_dict(checkpoint['optimizer'])\n            args.start_epoch = checkpoint['epoch'] + 1\n            if 'scaler' in checkpoint:\n                loss_scaler.load_state_dict(checkpoint['scaler'])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.load_model","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.load_model#L315-L329","kind":"function","name":"load_model","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":315,"end_line":329,"context_start_line":295,"context_end_line":340,"code":"def save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.resume:\n        if args.resume.startswith('https'):\n            checkpoint = torch.hub.load_state_dict_from_url(\n                args.resume, map_location='cpu', check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location='cpu')\n        model_without_ddp.load_state_dict(checkpoint['model'])\n        print(\"Resume checkpoint %s\" % args.resume)\n        if 'optimizer' in checkpoint and 'epoch' in checkpoint and not (hasattr(args, 'eval') and args.eval):\n            optimizer.load_state_dict(checkpoint['optimizer'])\n            args.start_epoch = checkpoint['epoch'] + 1\n            if 'scaler' in checkpoint:\n                loss_scaler.load_state_dict(checkpoint['scaler'])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.all_reduce_mean","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.all_reduce_mean#L332-L340","kind":"function","name":"all_reduce_mean","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":332,"end_line":340,"context_start_line":312,"context_end_line":340,"code":"        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.resume:\n        if args.resume.startswith('https'):\n            checkpoint = torch.hub.load_state_dict_from_url(\n                args.resume, map_location='cpu', check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location='cpu')\n        model_without_ddp.load_state_dict(checkpoint['model'])\n        print(\"Resume checkpoint %s\" % args.resume)\n        if 'optimizer' in checkpoint and 'epoch' in checkpoint and not (hasattr(args, 'eval') and args.eval):\n            optimizer.load_state_dict(checkpoint['optimizer'])\n            args.start_epoch = checkpoint['epoch'] + 1\n            if 'scaler' in checkpoint:\n                loss_scaler.load_state_dict(checkpoint['scaler'])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.__init__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.__init__#L254-L255","kind":"function","name":"__init__","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":254,"end_line":255,"context_start_line":234,"context_end_line":275,"code":"        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.update","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.update#L91-L98","kind":"function","name":"update","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":91,"end_line":98,"context_start_line":71,"context_end_line":118,"code":"        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.synchronize_between_processes","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.synchronize_between_processes#L116-L118","kind":"function","name":"synchronize_between_processes","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":116,"end_line":118,"context_start_line":96,"context_end_line":138,"code":"                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.median","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.median#L56-L58","kind":"function","name":"median","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":56,"end_line":58,"context_start_line":36,"context_end_line":78,"code":"\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.avg","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.avg#L61-L63","kind":"function","name":"avg","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":61,"end_line":63,"context_start_line":41,"context_end_line":83,"code":"\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.global_avg","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.global_avg#L66-L67","kind":"function","name":"global_avg","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":66,"end_line":67,"context_start_line":46,"context_end_line":87,"code":"        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.max","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.max#L70-L71","kind":"function","name":"max","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":70,"end_line":71,"context_start_line":50,"context_end_line":91,"code":"        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.value","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.value#L74-L75","kind":"function","name":"value","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":74,"end_line":75,"context_start_line":54,"context_end_line":95,"code":"\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.__str__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.__str__#L108-L114","kind":"function","name":"__str__","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":108,"end_line":114,"context_start_line":88,"context_end_line":134,"code":"        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.__getattr__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.__getattr__#L100-L106","kind":"function","name":"__getattr__","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":100,"end_line":106,"context_start_line":80,"context_end_line":126,"code":"            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.add_meter","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.add_meter#L120-L121","kind":"function","name":"add_meter","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":120,"end_line":121,"context_start_line":100,"context_end_line":141,"code":"    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.log_every","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.log_every#L123-L167","kind":"function","name":"log_every","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":123,"end_line":167,"context_start_line":103,"context_end_line":187,"code":"        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.print","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.print#L176-L182","kind":"function","name":"print","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":176,"end_line":182,"context_start_line":156,"context_end_line":202,"code":"                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.__call__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.__call__#L257-L271","kind":"function","name":"__call__","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":257,"end_line":271,"context_start_line":237,"context_end_line":291,"code":"        return\n\n    args.distributed = True\n\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.state_dict","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.state_dict#L273-L274","kind":"function","name":"state_dict","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":273,"end_line":274,"context_start_line":253,"context_end_line":294,"code":"\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.misc.load_state_dict","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.misc.load_state_dict#L276-L277","kind":"function","name":"load_state_dict","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":276,"end_line":277,"context_start_line":256,"context_end_line":297,"code":"\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.datasets","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.util.datasets#L1-L64","kind":"module","name":"gorilla.alpaca_finetuning_v1.util.datasets","path":"gorilla/alpaca_finetuning_v1/util/datasets.py","language":"python","start_line":1,"end_line":64,"context_start_line":1,"context_end_line":64,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\n\nimport os\n\nimport PIL\nfrom timm.data import create_transform\nfrom timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD\nfrom torchvision import datasets, transforms\n\n\ndef build_dataset(is_train, args):\n    transform = build_transform(is_train, args)\n\n    root = os.path.join(args.data_path, \"train\" if is_train else \"val\")\n    dataset = datasets.ImageFolder(root, transform=transform)\n\n    print(dataset)\n\n    return dataset\n\n\ndef build_transform(is_train, args):\n    mean = IMAGENET_DEFAULT_MEAN\n    std = IMAGENET_DEFAULT_STD\n    # train transform\n    if is_train:\n        # this should always dispatch to transforms_imagenet_train\n        transform = create_transform(\n            input_size=args.input_size,\n            is_training=True,\n            color_jitter=args.color_jitter,\n            auto_augment=args.aa,\n            interpolation=\"bicubic\",\n            re_prob=args.reprob,\n            re_mode=args.remode,\n            re_count=args.recount,\n            mean=mean,\n            std=std,\n        )\n        return transform\n\n    # eval transform\n    t = []\n    if args.input_size <= 224:\n        crop_pct = 224 / 256\n    else:\n        crop_pct = 1.0\n    size = int(args.input_size / crop_pct)\n    t.append(\n        transforms.Resize(size, interpolation=PIL.Image.BICUBIC),  # to maintain same ratio w.r.t. 224 images\n    )\n    t.append(transforms.CenterCrop(args.input_size))\n\n    t.append(transforms.ToTensor())\n    t.append(transforms.Normalize(mean, std))\n    return transforms.Compose(t)","source_hash":"b40d68d3bfae8ebdf0b0c35884f609d3d8a46dc73c5e6397d8f537a389ee9b1d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.datasets.build_dataset","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.datasets.build_dataset#L19-L27","kind":"function","name":"build_dataset","path":"gorilla/alpaca_finetuning_v1/util/datasets.py","language":"python","start_line":19,"end_line":27,"context_start_line":1,"context_end_line":47,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\n\nimport os\n\nimport PIL\nfrom timm.data import create_transform\nfrom timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD\nfrom torchvision import datasets, transforms\n\n\ndef build_dataset(is_train, args):\n    transform = build_transform(is_train, args)\n\n    root = os.path.join(args.data_path, \"train\" if is_train else \"val\")\n    dataset = datasets.ImageFolder(root, transform=transform)\n\n    print(dataset)\n\n    return dataset\n\n\ndef build_transform(is_train, args):\n    mean = IMAGENET_DEFAULT_MEAN\n    std = IMAGENET_DEFAULT_STD\n    # train transform\n    if is_train:\n        # this should always dispatch to transforms_imagenet_train\n        transform = create_transform(\n            input_size=args.input_size,\n            is_training=True,\n            color_jitter=args.color_jitter,\n            auto_augment=args.aa,\n            interpolation=\"bicubic\",\n            re_prob=args.reprob,\n            re_mode=args.remode,\n            re_count=args.recount,\n            mean=mean,\n            std=std,\n        )","source_hash":"b40d68d3bfae8ebdf0b0c35884f609d3d8a46dc73c5e6397d8f537a389ee9b1d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.datasets.build_transform","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.datasets.build_transform#L30-L64","kind":"function","name":"build_transform","path":"gorilla/alpaca_finetuning_v1/util/datasets.py","language":"python","start_line":30,"end_line":64,"context_start_line":10,"context_end_line":64,"code":"\nimport os\n\nimport PIL\nfrom timm.data import create_transform\nfrom timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD\nfrom torchvision import datasets, transforms\n\n\ndef build_dataset(is_train, args):\n    transform = build_transform(is_train, args)\n\n    root = os.path.join(args.data_path, \"train\" if is_train else \"val\")\n    dataset = datasets.ImageFolder(root, transform=transform)\n\n    print(dataset)\n\n    return dataset\n\n\ndef build_transform(is_train, args):\n    mean = IMAGENET_DEFAULT_MEAN\n    std = IMAGENET_DEFAULT_STD\n    # train transform\n    if is_train:\n        # this should always dispatch to transforms_imagenet_train\n        transform = create_transform(\n            input_size=args.input_size,\n            is_training=True,\n            color_jitter=args.color_jitter,\n            auto_augment=args.aa,\n            interpolation=\"bicubic\",\n            re_prob=args.reprob,\n            re_mode=args.remode,\n            re_count=args.recount,\n            mean=mean,\n            std=std,\n        )\n        return transform\n\n    # eval transform\n    t = []\n    if args.input_size <= 224:\n        crop_pct = 224 / 256\n    else:\n        crop_pct = 1.0\n    size = int(args.input_size / crop_pct)\n    t.append(\n        transforms.Resize(size, interpolation=PIL.Image.BICUBIC),  # to maintain same ratio w.r.t. 224 images\n    )\n    t.append(transforms.CenterCrop(args.input_size))\n\n    t.append(transforms.ToTensor())\n    t.append(transforms.Normalize(mean, std))\n    return transforms.Compose(t)","source_hash":"b40d68d3bfae8ebdf0b0c35884f609d3d8a46dc73c5e6397d8f537a389ee9b1d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.lr_decay","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.util.lr_decay#L1-L76","kind":"module","name":"gorilla.alpaca_finetuning_v1.util.lr_decay","path":"gorilla/alpaca_finetuning_v1/util/lr_decay.py","language":"python","start_line":1,"end_line":76,"context_start_line":1,"context_end_line":76,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# ELECTRA https://github.com/google-research/electra\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport json\n\n\ndef param_groups_lrd(model, weight_decay=0.05, no_weight_decay_list=[], layer_decay=.75):\n    \"\"\"\n    Parameter groups for layer-wise lr decay\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L58\n    \"\"\"\n    param_group_names = {}\n    param_groups = {}\n\n    num_layers = len(model.blocks) + 1\n\n    layer_scales = list(layer_decay ** (num_layers - i) for i in range(num_layers + 1))\n\n    for n, p in model.named_parameters():\n        if not p.requires_grad:\n            continue\n\n        # no decay: all 1D parameters and model specific ones\n        if p.ndim == 1 or n in no_weight_decay_list:\n            g_decay = \"no_decay\"\n            this_decay = 0.\n        else:\n            g_decay = \"decay\"\n            this_decay = weight_decay\n            \n        layer_id = get_layer_id_for_vit(n, num_layers)\n        group_name = \"layer_%d_%s\" % (layer_id, g_decay)\n\n        if group_name not in param_group_names:\n            this_scale = layer_scales[layer_id]\n\n            param_group_names[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n            param_groups[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n\n        param_group_names[group_name][\"params\"].append(n)\n        param_groups[group_name][\"params\"].append(p)\n\n    # print(\"parameter groups: \\n%s\" % json.dumps(param_group_names, indent=2))\n\n    return list(param_groups.values())\n\n\ndef get_layer_id_for_vit(name, num_layers):\n    \"\"\"\n    Assign a parameter with its layer id\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L33\n    \"\"\"\n    if name in ['cls_token', 'pos_embed']:\n        return 0\n    elif name.startswith('patch_embed'):\n        return 0\n    elif name.startswith('blocks'):\n        return int(name.split('.')[1]) + 1\n    else:\n        return num_layers","source_hash":"c21f7ac4070e7dc3ef9fe3a904b9ab59b80082d136b70d99cf7853c5ceb62dba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.lr_decay.param_groups_lrd","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.lr_decay.param_groups_lrd#L15-L61","kind":"function","name":"param_groups_lrd","path":"gorilla/alpaca_finetuning_v1/util/lr_decay.py","language":"python","start_line":15,"end_line":61,"context_start_line":1,"context_end_line":76,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# ELECTRA https://github.com/google-research/electra\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport json\n\n\ndef param_groups_lrd(model, weight_decay=0.05, no_weight_decay_list=[], layer_decay=.75):\n    \"\"\"\n    Parameter groups for layer-wise lr decay\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L58\n    \"\"\"\n    param_group_names = {}\n    param_groups = {}\n\n    num_layers = len(model.blocks) + 1\n\n    layer_scales = list(layer_decay ** (num_layers - i) for i in range(num_layers + 1))\n\n    for n, p in model.named_parameters():\n        if not p.requires_grad:\n            continue\n\n        # no decay: all 1D parameters and model specific ones\n        if p.ndim == 1 or n in no_weight_decay_list:\n            g_decay = \"no_decay\"\n            this_decay = 0.\n        else:\n            g_decay = \"decay\"\n            this_decay = weight_decay\n            \n        layer_id = get_layer_id_for_vit(n, num_layers)\n        group_name = \"layer_%d_%s\" % (layer_id, g_decay)\n\n        if group_name not in param_group_names:\n            this_scale = layer_scales[layer_id]\n\n            param_group_names[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n            param_groups[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n\n        param_group_names[group_name][\"params\"].append(n)\n        param_groups[group_name][\"params\"].append(p)\n\n    # print(\"parameter groups: \\n%s\" % json.dumps(param_group_names, indent=2))\n\n    return list(param_groups.values())\n\n\ndef get_layer_id_for_vit(name, num_layers):\n    \"\"\"\n    Assign a parameter with its layer id\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L33\n    \"\"\"\n    if name in ['cls_token', 'pos_embed']:\n        return 0\n    elif name.startswith('patch_embed'):\n        return 0\n    elif name.startswith('blocks'):\n        return int(name.split('.')[1]) + 1\n    else:\n        return num_layers","source_hash":"c21f7ac4070e7dc3ef9fe3a904b9ab59b80082d136b70d99cf7853c5ceb62dba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.lr_decay.get_layer_id_for_vit","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.lr_decay.get_layer_id_for_vit#L64-L76","kind":"function","name":"get_layer_id_for_vit","path":"gorilla/alpaca_finetuning_v1/util/lr_decay.py","language":"python","start_line":64,"end_line":76,"context_start_line":44,"context_end_line":76,"code":"\n            param_group_names[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n            param_groups[group_name] = {\n                \"lr_scale\": this_scale,\n                \"weight_decay\": this_decay,\n                \"params\": [],\n            }\n\n        param_group_names[group_name][\"params\"].append(n)\n        param_groups[group_name][\"params\"].append(p)\n\n    # print(\"parameter groups: \\n%s\" % json.dumps(param_group_names, indent=2))\n\n    return list(param_groups.values())\n\n\ndef get_layer_id_for_vit(name, num_layers):\n    \"\"\"\n    Assign a parameter with its layer id\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L33\n    \"\"\"\n    if name in ['cls_token', 'pos_embed']:\n        return 0\n    elif name.startswith('patch_embed'):\n        return 0\n    elif name.startswith('blocks'):\n        return int(name.split('.')[1]) + 1\n    else:\n        return num_layers","source_hash":"c21f7ac4070e7dc3ef9fe3a904b9ab59b80082d136b70d99cf7853c5ceb62dba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.lr_sched","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.util.lr_sched#L1-L23","kind":"module","name":"gorilla.alpaca_finetuning_v1.util.lr_sched","path":"gorilla/alpaca_finetuning_v1/util/lr_sched.py","language":"python","start_line":1,"end_line":23,"context_start_line":1,"context_end_line":23,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs\n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * (\n            1.0 + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs))\n        )\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"136a8dba00b53af7934e7377cc6a0dce1e84fef604e55673ce8460b109aa6d0d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.lr_sched.adjust_learning_rate","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.lr_sched.adjust_learning_rate#L10-L23","kind":"function","name":"adjust_learning_rate","path":"gorilla/alpaca_finetuning_v1/util/lr_sched.py","language":"python","start_line":10,"end_line":23,"context_start_line":1,"context_end_line":23,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs\n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * (\n            1.0 + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs))\n        )\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"136a8dba00b53af7934e7377cc6a0dce1e84fef604e55673ce8460b109aa6d0d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.pos_embed","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.util.pos_embed#L1-L97","kind":"module","name":"gorilla.alpaca_finetuning_v1.util.pos_embed","path":"gorilla/alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":1,"end_line":97,"context_start_line":1,"context_end_line":97,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# Position embedding utils\n# --------------------------------------------------------\n\nimport numpy as np\nimport torch\n\n\n# --------------------------------------------------------\n# 2D sine-cosine position embedding\n# References:\n# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py\n# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"\n    grid_size: int of the grid height and width\n    return:\n    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)\n    \"\"\"\n    grid_h = np.arange(grid_size, dtype=np.float32)\n    grid_w = np.arange(grid_size, dtype=np.float32)\n    grid = np.meshgrid(grid_w, grid_h)  # here w goes first\n    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.0\n    omega = 1.0 / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum(\"m,d->md\", pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out)  # (M, D/2)\n    emb_cos = np.cos(out)  # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)\n    return emb\n\n\n# --------------------------------------------------------\n# Interpolate position embeddings for high-resolution\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\ndef interpolate_pos_embed(model, checkpoint_model):\n    if \"pos_embed\" in checkpoint_model:\n        pos_embed_checkpoint = checkpoint_model[\"pos_embed\"]\n        embedding_size = pos_embed_checkpoint.shape[-1]\n        num_patches = model.patch_embed.num_patches\n        num_extra_tokens = model.pos_embed.shape[-2] - num_patches\n        # height (== width) for the checkpoint position embedding\n        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)\n        # height (== width) for the new position embedding\n        new_size = int(num_patches**0.5)\n        # class_token and dist_token are kept unchanged\n        if orig_size != new_size:\n            print(\"Position interpolate from %dx%d to %dx%d\" % (orig_size, orig_size, new_size, new_size))\n            extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]\n            # only the position tokens are interpolated\n            pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:]\n            pos_tokens = pos_tokens.reshape(-1, orig_size, orig_size, embedding_size).permute(0, 3, 1, 2)\n            pos_tokens = torch.nn.functional.interpolate(\n                pos_tokens, size=(new_size, new_size), mode=\"bicubic\", align_corners=False\n            )\n            pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)\n            new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)\n            checkpoint_model[\"pos_embed\"] = new_pos_embed","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.pos_embed.get_2d_sincos_pos_embed","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.pos_embed.get_2d_sincos_pos_embed#L20-L35","kind":"function","name":"get_2d_sincos_pos_embed","path":"gorilla/alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":20,"end_line":35,"context_start_line":1,"context_end_line":55,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# Position embedding utils\n# --------------------------------------------------------\n\nimport numpy as np\nimport torch\n\n\n# --------------------------------------------------------\n# 2D sine-cosine position embedding\n# References:\n# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py\n# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"\n    grid_size: int of the grid height and width\n    return:\n    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)\n    \"\"\"\n    grid_h = np.arange(grid_size, dtype=np.float32)\n    grid_w = np.arange(grid_size, dtype=np.float32)\n    grid = np.meshgrid(grid_w, grid_h)  # here w goes first\n    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.pos_embed.get_2d_sincos_pos_embed_from_grid","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.pos_embed.get_2d_sincos_pos_embed_from_grid#L38-L46","kind":"function","name":"get_2d_sincos_pos_embed_from_grid","path":"gorilla/alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":38,"end_line":46,"context_start_line":18,"context_end_line":66,"code":"# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"\n    grid_size: int of the grid height and width\n    return:\n    pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or w/o cls_token)\n    \"\"\"\n    grid_h = np.arange(grid_size, dtype=np.float32)\n    grid_w = np.arange(grid_size, dtype=np.float32)\n    grid = np.meshgrid(grid_w, grid_h)  # here w goes first\n    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.0\n    omega = 1.0 / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum(\"m,d->md\", pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out)  # (M, D/2)\n    emb_cos = np.cos(out)  # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.pos_embed.get_1d_sincos_pos_embed_from_grid","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.pos_embed.get_1d_sincos_pos_embed_from_grid#L49-L67","kind":"function","name":"get_1d_sincos_pos_embed_from_grid","path":"gorilla/alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":49,"end_line":67,"context_start_line":29,"context_end_line":87,"code":"    grid = np.stack(grid, axis=0)\n\n    grid = grid.reshape([2, 1, grid_size, grid_size])\n    pos_embed = get_2d_sincos_pos_embed_from_grid(embed_dim, grid)\n    if cls_token:\n        pos_embed = np.concatenate([np.zeros([1, embed_dim]), pos_embed], axis=0)\n    return pos_embed\n\n\ndef get_2d_sincos_pos_embed_from_grid(embed_dim, grid):\n    assert embed_dim % 2 == 0\n\n    # use half of dimensions to encode grid_h\n    emb_h = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[0])  # (H*W, D/2)\n    emb_w = get_1d_sincos_pos_embed_from_grid(embed_dim // 2, grid[1])  # (H*W, D/2)\n\n    emb = np.concatenate([emb_h, emb_w], axis=1)  # (H*W, D)\n    return emb\n\n\ndef get_1d_sincos_pos_embed_from_grid(embed_dim, pos):\n    \"\"\"\n    embed_dim: output dimension for each position\n    pos: a list of positions to be encoded: size (M,)\n    out: (M, D)\n    \"\"\"\n    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.0\n    omega = 1.0 / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum(\"m,d->md\", pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out)  # (M, D/2)\n    emb_cos = np.cos(out)  # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)\n    return emb\n\n\n# --------------------------------------------------------\n# Interpolate position embeddings for high-resolution\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\ndef interpolate_pos_embed(model, checkpoint_model):\n    if \"pos_embed\" in checkpoint_model:\n        pos_embed_checkpoint = checkpoint_model[\"pos_embed\"]\n        embedding_size = pos_embed_checkpoint.shape[-1]\n        num_patches = model.patch_embed.num_patches\n        num_extra_tokens = model.pos_embed.shape[-2] - num_patches\n        # height (== width) for the checkpoint position embedding\n        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)\n        # height (== width) for the new position embedding\n        new_size = int(num_patches**0.5)\n        # class_token and dist_token are kept unchanged\n        if orig_size != new_size:\n            print(\"Position interpolate from %dx%d to %dx%d\" % (orig_size, orig_size, new_size, new_size))","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.pos_embed.interpolate_pos_embed","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.pos_embed.interpolate_pos_embed#L75-L97","kind":"function","name":"interpolate_pos_embed","path":"gorilla/alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":75,"end_line":97,"context_start_line":55,"context_end_line":97,"code":"    assert embed_dim % 2 == 0\n    omega = np.arange(embed_dim // 2, dtype=np.float)\n    omega /= embed_dim / 2.0\n    omega = 1.0 / 10000**omega  # (D/2,)\n\n    pos = pos.reshape(-1)  # (M,)\n    out = np.einsum(\"m,d->md\", pos, omega)  # (M, D/2), outer product\n\n    emb_sin = np.sin(out)  # (M, D/2)\n    emb_cos = np.cos(out)  # (M, D/2)\n\n    emb = np.concatenate([emb_sin, emb_cos], axis=1)  # (M, D)\n    return emb\n\n\n# --------------------------------------------------------\n# Interpolate position embeddings for high-resolution\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\ndef interpolate_pos_embed(model, checkpoint_model):\n    if \"pos_embed\" in checkpoint_model:\n        pos_embed_checkpoint = checkpoint_model[\"pos_embed\"]\n        embedding_size = pos_embed_checkpoint.shape[-1]\n        num_patches = model.patch_embed.num_patches\n        num_extra_tokens = model.pos_embed.shape[-2] - num_patches\n        # height (== width) for the checkpoint position embedding\n        orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5)\n        # height (== width) for the new position embedding\n        new_size = int(num_patches**0.5)\n        # class_token and dist_token are kept unchanged\n        if orig_size != new_size:\n            print(\"Position interpolate from %dx%d to %dx%d\" % (orig_size, orig_size, new_size, new_size))\n            extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens]\n            # only the position tokens are interpolated\n            pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:]\n            pos_tokens = pos_tokens.reshape(-1, orig_size, orig_size, embedding_size).permute(0, 3, 1, 2)\n            pos_tokens = torch.nn.functional.interpolate(\n                pos_tokens, size=(new_size, new_size), mode=\"bicubic\", align_corners=False\n            )\n            pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)\n            new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)\n            checkpoint_model[\"pos_embed\"] = new_pos_embed","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.lars","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.util.lars#L1-L47","kind":"module","name":"gorilla.alpaca_finetuning_v1.util.lars","path":"gorilla/alpaca_finetuning_v1/util/lars.py","language":"python","start_line":1,"end_line":47,"context_start_line":1,"context_end_line":47,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# LARS optimizer, implementation from MoCo v3:\n# https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\n\nimport torch\n\n\nclass LARS(torch.optim.Optimizer):\n    \"\"\"\n    LARS optimizer, no rate scaling or weight decay for parameters <= 1D.\n    \"\"\"\n    def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, trust_coefficient=0.001):\n        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, trust_coefficient=trust_coefficient)\n        super().__init__(params, defaults)\n\n    @torch.no_grad()\n    def step(self):\n        for g in self.param_groups:\n            for p in g['params']:\n                dp = p.grad\n\n                if dp is None:\n                    continue\n\n                if p.ndim > 1: # if not normalization gamma/beta or bias\n                    dp = dp.add(p, alpha=g['weight_decay'])\n                    param_norm = torch.norm(p)\n                    update_norm = torch.norm(dp)\n                    one = torch.ones_like(param_norm)\n                    q = torch.where(param_norm > 0.,\n                                    torch.where(update_norm > 0,\n                                    (g['trust_coefficient'] * param_norm / update_norm), one),\n                                    one)\n                    dp = dp.mul(q)\n\n                param_state = self.state[p]\n                if 'mu' not in param_state:\n                    param_state['mu'] = torch.zeros_like(p)\n                mu = param_state['mu']\n                mu.mul_(g['momentum']).add_(dp)\n                p.add_(mu, alpha=-g['lr'])","source_hash":"1bb75085eecf93b8d7ed7972c507495586e71a8f483bcdcd4f37c354f88a0fc0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.lars.LARS","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.util.lars.LARS#L14-L47","kind":"class","name":"LARS","path":"gorilla/alpaca_finetuning_v1/util/lars.py","language":"python","start_line":14,"end_line":47,"context_start_line":1,"context_end_line":47,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# LARS optimizer, implementation from MoCo v3:\n# https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\n\nimport torch\n\n\nclass LARS(torch.optim.Optimizer):\n    \"\"\"\n    LARS optimizer, no rate scaling or weight decay for parameters <= 1D.\n    \"\"\"\n    def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, trust_coefficient=0.001):\n        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, trust_coefficient=trust_coefficient)\n        super().__init__(params, defaults)\n\n    @torch.no_grad()\n    def step(self):\n        for g in self.param_groups:\n            for p in g['params']:\n                dp = p.grad\n\n                if dp is None:\n                    continue\n\n                if p.ndim > 1: # if not normalization gamma/beta or bias\n                    dp = dp.add(p, alpha=g['weight_decay'])\n                    param_norm = torch.norm(p)\n                    update_norm = torch.norm(dp)\n                    one = torch.ones_like(param_norm)\n                    q = torch.where(param_norm > 0.,\n                                    torch.where(update_norm > 0,\n                                    (g['trust_coefficient'] * param_norm / update_norm), one),\n                                    one)\n                    dp = dp.mul(q)\n\n                param_state = self.state[p]\n                if 'mu' not in param_state:\n                    param_state['mu'] = torch.zeros_like(p)\n                mu = param_state['mu']\n                mu.mul_(g['momentum']).add_(dp)\n                p.add_(mu, alpha=-g['lr'])","source_hash":"1bb75085eecf93b8d7ed7972c507495586e71a8f483bcdcd4f37c354f88a0fc0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.lars.__init__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.lars.__init__#L18-L20","kind":"function","name":"__init__","path":"gorilla/alpaca_finetuning_v1/util/lars.py","language":"python","start_line":18,"end_line":20,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# LARS optimizer, implementation from MoCo v3:\n# https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\n\nimport torch\n\n\nclass LARS(torch.optim.Optimizer):\n    \"\"\"\n    LARS optimizer, no rate scaling or weight decay for parameters <= 1D.\n    \"\"\"\n    def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, trust_coefficient=0.001):\n        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, trust_coefficient=trust_coefficient)\n        super().__init__(params, defaults)\n\n    @torch.no_grad()\n    def step(self):\n        for g in self.param_groups:\n            for p in g['params']:\n                dp = p.grad\n\n                if dp is None:\n                    continue\n\n                if p.ndim > 1: # if not normalization gamma/beta or bias\n                    dp = dp.add(p, alpha=g['weight_decay'])\n                    param_norm = torch.norm(p)\n                    update_norm = torch.norm(dp)\n                    one = torch.ones_like(param_norm)\n                    q = torch.where(param_norm > 0.,\n                                    torch.where(update_norm > 0,\n                                    (g['trust_coefficient'] * param_norm / update_norm), one),\n                                    one)\n                    dp = dp.mul(q)","source_hash":"1bb75085eecf93b8d7ed7972c507495586e71a8f483bcdcd4f37c354f88a0fc0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.util.lars.step","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.util.lars.step#L23-L47","kind":"function","name":"step","path":"gorilla/alpaca_finetuning_v1/util/lars.py","language":"python","start_line":23,"end_line":47,"context_start_line":3,"context_end_line":47,"code":"\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# LARS optimizer, implementation from MoCo v3:\n# https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\n\nimport torch\n\n\nclass LARS(torch.optim.Optimizer):\n    \"\"\"\n    LARS optimizer, no rate scaling or weight decay for parameters <= 1D.\n    \"\"\"\n    def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, trust_coefficient=0.001):\n        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, trust_coefficient=trust_coefficient)\n        super().__init__(params, defaults)\n\n    @torch.no_grad()\n    def step(self):\n        for g in self.param_groups:\n            for p in g['params']:\n                dp = p.grad\n\n                if dp is None:\n                    continue\n\n                if p.ndim > 1: # if not normalization gamma/beta or bias\n                    dp = dp.add(p, alpha=g['weight_decay'])\n                    param_norm = torch.norm(p)\n                    update_norm = torch.norm(dp)\n                    one = torch.ones_like(param_norm)\n                    q = torch.where(param_norm > 0.,\n                                    torch.where(update_norm > 0,\n                                    (g['trust_coefficient'] * param_norm / update_norm), one),\n                                    one)\n                    dp = dp.mul(q)\n\n                param_state = self.state[p]\n                if 'mu' not in param_state:\n                    param_state['mu'] = torch.zeros_like(p)\n                mu = param_state['mu']\n                mu.mul_(g['momentum']).add_(dp)\n                p.add_(mu, alpha=-g['lr'])","source_hash":"1bb75085eecf93b8d7ed7972c507495586e71a8f483bcdcd4f37c354f88a0fc0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.generation","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.llama.generation#L1-L75","kind":"module","name":"gorilla.alpaca_finetuning_v1.llama.generation","path":"gorilla/alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":1,"end_line":75,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward_only(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.generation.LLaMA","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.llama.generation.LLaMA#L12-L64","kind":"class","name":"LLaMA","path":"gorilla/alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":12,"end_line":64,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward_only(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.generation.sample_top_p","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.generation.sample_top_p#L67-L75","kind":"function","name":"sample_top_p","path":"gorilla/alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":67,"end_line":75,"context_start_line":47,"context_end_line":75,"code":"                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.generation.__init__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.generation.__init__#L13-L15","kind":"function","name":"__init__","path":"gorilla/alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":13,"end_line":15,"context_start_line":1,"context_end_line":35,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.generation.generate","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.generation.generate#L17-L64","kind":"function","name":"generate","path":"gorilla/alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":17,"end_line":64,"context_start_line":1,"context_end_line":75,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward_only(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[: len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.llama.model#L1-L222","kind":"module","name":"gorilla.alpaca_finetuning_v1.llama.model","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":1,"end_line":222,"context_start_line":1,"context_end_line":222,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport torch\nimport torch.nn.functional as F\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n    adapter_len: int = 10\n    adapter_layer: int = 30\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wk = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wv = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wo = Linear(args.n_heads * self.head_dim, args.dim, bias=False)\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_k = self.wk(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_v = self.wv(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            xk = torch.cat([adapter_k, xk], dim=1)\n            xv = torch.cat([adapter_v, xv], dim=1)\n            extra_mask = torch.zeros(1, 1, seqlen, adapter_len).to(mask)\n            mask = torch.cat([extra_mask, mask], dim=-1)\n        keys = xk\n        values = xv\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        if adapter is not None:\n            scores = torch.cat(\n                [\n                    self.gate.tanh().half() * F.softmax(scores[:, :, :, :adapter_len].float(), dim=-1).type_as(xq),\n                    F.softmax(scores[:, :, :, adapter_len:].float(), dim=-1).type_as(xq),\n                ],\n                dim=-1,\n            )\n        else:\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(dim, hidden_dim, bias=False)\n        self.w2 = Linear(hidden_dim, dim, bias=False)\n        self.w3 = Linear(dim, hidden_dim, bias=False)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(params.vocab_size, params.dim)\n\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n\n        _bsz, seqlen = examples.shape\n\n        with torch.no_grad():\n            h = self.tok_embeddings(examples)\n            freqs_cis = self.freqs_cis.to(h.device)\n            freqs_cis = freqs_cis[:seqlen]\n            mask = None\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n            mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n            start_pos = 0\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter_index = 0\n        adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, 4096).unsqueeze(1)\n        for layer in self.layers[-1 * self.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n            adapter_index = adapter_index + 1\n\n        h = self.norm(h)\n        output = self.output(h)\n        output = output[:, :-1, :].reshape(-1, self.vocab_size)\n        labels = labels[:, 1:].flatten()\n\n        c_loss = self.criterion(output, labels)\n        return c_loss","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.ModelArgs","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.llama.model.ModelArgs#L15-L26","kind":"class","name":"ModelArgs","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":15,"end_line":26,"context_start_line":1,"context_end_line":46,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport torch\nimport torch.nn.functional as F\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n    adapter_len: int = 10\n    adapter_layer: int = 30\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.RMSNorm","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.llama.model.RMSNorm#L29-L40","kind":"class","name":"RMSNorm","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":29,"end_line":40,"context_start_line":9,"context_end_line":60,"code":"import torch.nn.functional as F\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n    adapter_len: int = 10\n    adapter_layer: int = 30\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.precompute_freqs_cis","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.model.precompute_freqs_cis#L43-L48","kind":"function","name":"precompute_freqs_cis","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":43,"end_line":48,"context_start_line":23,"context_end_line":68,"code":"    max_batch_size: int = 32\n    max_seq_len: int = 2048\n    adapter_len: int = 10\n    adapter_layer: int = 30\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.reshape_for_broadcast","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.model.reshape_for_broadcast#L51-L56","kind":"function","name":"reshape_for_broadcast","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":51,"end_line":56,"context_start_line":31,"context_end_line":76,"code":"        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.apply_rotary_emb","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.model.apply_rotary_emb#L59-L69","kind":"function","name":"apply_rotary_emb","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":59,"end_line":69,"context_start_line":39,"context_end_line":89,"code":"        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wk = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wv = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wo = Linear(args.n_heads * self.head_dim, args.dim, bias=False)\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.Attention","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.llama.model.Attention#L72-L129","kind":"class","name":"Attention","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":72,"end_line":129,"context_start_line":52,"context_end_line":149,"code":"    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wk = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wv = Linear(args.dim, args.n_heads * self.head_dim, bias=False)\n        self.wo = Linear(args.n_heads * self.head_dim, args.dim, bias=False)\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_k = self.wk(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_v = self.wv(adapter).view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            xk = torch.cat([adapter_k, xk], dim=1)\n            xv = torch.cat([adapter_v, xv], dim=1)\n            extra_mask = torch.zeros(1, 1, seqlen, adapter_len).to(mask)\n            mask = torch.cat([extra_mask, mask], dim=-1)\n        keys = xk\n        values = xv\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        if adapter is not None:\n            scores = torch.cat(\n                [\n                    self.gate.tanh().half() * F.softmax(scores[:, :, :, :adapter_len].float(), dim=-1).type_as(xq),\n                    F.softmax(scores[:, :, :, adapter_len:].float(), dim=-1).type_as(xq),\n                ],\n                dim=-1,\n            )\n        else:\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(dim, hidden_dim, bias=False)\n        self.w2 = Linear(hidden_dim, dim, bias=False)\n        self.w3 = Linear(dim, hidden_dim, bias=False)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.FeedForward","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.llama.model.FeedForward#L132-L148","kind":"class","name":"FeedForward","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":132,"end_line":148,"context_start_line":112,"context_end_line":168,"code":"        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n        if adapter is not None:\n            scores = torch.cat(\n                [\n                    self.gate.tanh().half() * F.softmax(scores[:, :, :, :adapter_len].float(), dim=-1).type_as(xq),\n                    F.softmax(scores[:, :, :, adapter_len:].float(), dim=-1).type_as(xq),\n                ],\n                dim=-1,\n            )\n        else:\n            scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(dim, hidden_dim, bias=False)\n        self.w2 = Linear(hidden_dim, dim, bias=False)\n        self.w3 = Linear(dim, hidden_dim, bias=False)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.TransformerBlock","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.llama.model.TransformerBlock#L151-L169","kind":"class","name":"TransformerBlock","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":151,"end_line":169,"context_start_line":131,"context_end_line":189,"code":"\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(dim, hidden_dim, bias=False)\n        self.w2 = Linear(hidden_dim, dim, bias=False)\n        self.w3 = Linear(dim, hidden_dim, bias=False)\n\n    def forward(self, x):\n        return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(params.vocab_size, params.dim)\n\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.Transformer","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.llama.model.Transformer#L172-L222","kind":"class","name":"Transformer","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":172,"end_line":222,"context_start_line":152,"context_end_line":222,"code":"    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(params.vocab_size, params.dim)\n\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n\n        _bsz, seqlen = examples.shape\n\n        with torch.no_grad():\n            h = self.tok_embeddings(examples)\n            freqs_cis = self.freqs_cis.to(h.device)\n            freqs_cis = freqs_cis[:seqlen]\n            mask = None\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n            mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n            start_pos = 0\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter_index = 0\n        adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, 4096).unsqueeze(1)\n        for layer in self.layers[-1 * self.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n            adapter_index = adapter_index + 1\n\n        h = self.norm(h)\n        output = self.output(h)\n        output = output[:, :-1, :].reshape(-1, self.vocab_size)\n        labels = labels[:, 1:].flatten()\n\n        c_loss = self.criterion(output, labels)\n        return c_loss","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.__init__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.model.__init__#L173-L193","kind":"function","name":"__init__","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":173,"end_line":193,"context_start_line":153,"context_end_line":213,"code":"        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of)\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(params.vocab_size, params.dim)\n\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n\n        _bsz, seqlen = examples.shape\n\n        with torch.no_grad():\n            h = self.tok_embeddings(examples)\n            freqs_cis = self.freqs_cis.to(h.device)\n            freqs_cis = freqs_cis[:seqlen]\n            mask = None\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n            mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n            start_pos = 0\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter_index = 0\n        adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, 4096).unsqueeze(1)\n        for layer in self.layers[-1 * self.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model._norm","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.model._norm#L35-L36","kind":"function","name":"_norm","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":35,"end_line":36,"context_start_line":15,"context_end_line":56,"code":"class ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n    adapter_len: int = 10\n    adapter_layer: int = 30\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.model.forward","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.model.forward#L195-L222","kind":"function","name":"forward","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":195,"end_line":222,"context_start_line":175,"context_end_line":222,"code":"        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(params.vocab_size, params.dim)\n\n        self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(params.dim, params.vocab_size, bias=False)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n\n        _bsz, seqlen = examples.shape\n\n        with torch.no_grad():\n            h = self.tok_embeddings(examples)\n            freqs_cis = self.freqs_cis.to(h.device)\n            freqs_cis = freqs_cis[:seqlen]\n            mask = None\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n            mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n            start_pos = 0\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        adapter_index = 0\n        adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, 4096).unsqueeze(1)\n        for layer in self.layers[-1 * self.adapter_layer :]:\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n            adapter_index = adapter_index + 1\n\n        h = self.norm(h)\n        output = self.output(h)\n        output = output[:, :-1, :].reshape(-1, self.vocab_size)\n        labels = labels[:, 1:].flatten()\n\n        c_loss = self.criterion(output, labels)\n        return c_loss","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.tokenizer","uri":"program://LLaMA-Adapter/module/gorilla.alpaca_finetuning_v1.llama.tokenizer#L1-L38","kind":"module","name":"gorilla.alpaca_finetuning_v1.llama.tokenizer","path":"gorilla/alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":1,"end_line":38,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.tokenizer.Tokenizer","uri":"program://LLaMA-Adapter/class/gorilla.alpaca_finetuning_v1.llama.tokenizer.Tokenizer#L13-L38","kind":"class","name":"Tokenizer","path":"gorilla/alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":13,"end_line":38,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.tokenizer.__init__","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.tokenizer.__init__#L14-L26","kind":"function","name":"__init__","path":"gorilla/alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":14,"end_line":26,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.tokenizer.encode","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.tokenizer.encode#L28-L35","kind":"function","name":"encode","path":"gorilla/alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":28,"end_line":35,"context_start_line":8,"context_end_line":38,"code":"from sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.alpaca_finetuning_v1.llama.tokenizer.decode","uri":"program://LLaMA-Adapter/function/gorilla.alpaca_finetuning_v1.llama.tokenizer.decode#L37-L38","kind":"function","name":"decode","path":"gorilla/alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":37,"end_line":38,"context_start_line":17,"context_end_line":38,"code":"        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.apply_delta","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.inference.apply_delta#L1-L167","kind":"module","name":"gorilla.gorilla-main.inference.apply_delta","path":"gorilla/gorilla-main/inference/apply_delta.py","language":"python","start_line":1,"end_line":167,"context_start_line":1,"context_end_line":167,"code":"\"\"\"\nApply the delta weights on top of a base model.\n\nUsage:\npython3 apply_delta.py --base-model-path path/to/hf_llama/ --target-model-path path/to/gorilla-7b-hf-v0 --delta-path path/to/models--gorilla-llm--gorilla-7b-hf-delta-v0\n\nThanks to LMSYS for the template of this code.\n\"\"\"\nimport argparse\nimport gc\nimport glob\nimport json\nimport os\nimport shutil\nimport tempfile\n\nfrom huggingface_hub import snapshot_download\nimport torch\nfrom torch import nn\nfrom tqdm import tqdm\nfrom transformers import AutoTokenizer, AutoModelForCausalLM, AutoConfig\n\n\nGB = 1 << 30\n\n\ndef split_files(model_path, tmp_path, split_size):\n    if not os.path.exists(model_path):\n        model_path = snapshot_download(repo_id=model_path)\n    if not os.path.exists(tmp_path):\n        os.makedirs(tmp_path)\n\n    file_pattern = os.path.join(model_path, \"pytorch_model-*.bin\")\n    files = glob.glob(file_pattern)\n\n    part = 0\n    try:\n        for file_path in tqdm(files):\n            state_dict = torch.load(file_path)\n            new_state_dict = {}\n\n            current_size = 0\n            for name, param in state_dict.items():\n                param_size = param.numel() * param.element_size()\n\n                if current_size + param_size > split_size:\n                    new_file_name = f\"pytorch_model-{part}.bin\"\n                    new_file_path = os.path.join(tmp_path, new_file_name)\n                    torch.save(new_state_dict, new_file_path)\n                    current_size = 0\n                    new_state_dict = None\n                    gc.collect()\n                    new_state_dict = {}\n                    part += 1\n\n                new_state_dict[name] = param\n                current_size += param_size\n\n            new_file_name = f\"pytorch_model-{part}.bin\"\n            new_file_path = os.path.join(tmp_path, new_file_name)\n            torch.save(new_state_dict, new_file_path)\n            new_state_dict = None\n            gc.collect()\n            new_state_dict = {}\n            part += 1\n    except Exception as e:\n        print(f\"An error occurred during split_files: {e}\")\n        shutil.rmtree(tmp_path)\n        raise\n\n\ndef apply_delta_low_cpu_mem(base_model_path, target_model_path, delta_path):\n    delta_tokenizer = AutoTokenizer.from_pretrained(delta_path, use_fast=False)\n    delta_config = AutoConfig.from_pretrained(delta_path)\n\n    if os.path.exists(target_model_path):\n        shutil.rmtree(target_model_path)\n    os.makedirs(target_model_path)\n\n    split_size = 4 * GB\n\n    with tempfile.TemporaryDirectory() as tmp_base_path, tempfile.TemporaryDirectory() as tmp_delta_path:\n        print(f\"Split files for the base model to {tmp_base_path}\")\n        split_files(base_model_path, tmp_base_path, split_size)\n        print(f\"Split files for the delta weights to {tmp_delta_path}\")\n        split_files(delta_path, tmp_delta_path, split_size)\n\n        base_pattern = os.path.join(tmp_base_path, \"pytorch_model-*.bin\")\n        base_files = glob.glob(base_pattern)\n        delta_pattern = os.path.join(tmp_delta_path, \"pytorch_model-*.bin\")\n        delta_files = glob.glob(delta_pattern)\n        delta_state_dict = torch.load(delta_files[0])\n\n        print(\"Applying the delta\")\n        weight_map = {}\n        total_size = 0\n\n        for i, base_file in tqdm(enumerate(base_files)):\n            state_dict = torch.load(base_file)\n            file_name = f\"pytorch_model-{i}.bin\"\n            for name, param in state_dict.items():\n                if name not in delta_state_dict:\n                    for delta_file in delta_files:\n                        delta_state_dict = torch.load(delta_file)\n                        gc.collect()\n                        if name in delta_state_dict:\n                            break\n\n                state_dict[name] += delta_state_dict[name]\n                weight_map[name] = file_name\n                total_size += param.numel() * param.element_size()\n                gc.collect()\n            torch.save(state_dict, os.path.join(target_model_path, file_name))\n\n        with open(\n            os.path.join(target_model_path, \"pytorch_model.bin.index.json\"), \"w\"\n        ) as f:\n            json.dump(\n                {\"weight_map\": weight_map, \"metadata\": {\"total_size\": total_size}}, f\n            )\n\n    print(f\"Saving the target model to {target_model_path}\")\n    delta_tokenizer.save_pretrained(target_model_path)\n    delta_config.save_pretrained(target_model_path)\n\n\ndef apply_delta(base_model_path, target_model_path, delta_path):\n    print(f\"Loading the delta weights from {delta_path}\")\n    delta_tokenizer = AutoTokenizer.from_pretrained(delta_path, use_fast=False)\n    delta = AutoModelForCausalLM.from_pretrained(\n        delta_path, torch_dtype=torch.float16, low_cpu_mem_usage=True\n    )\n\n    print(f\"Loading the base model from {base_model_path}\")\n    base = AutoModelForCausalLM.from_pretrained(\n        base_model_path, torch_dtype=torch.float16, low_cpu_mem_usage=True\n    )\n\n    print(\"Applying the delta\")\n    for name, param in tqdm(base.state_dict().items(), desc=\"Applying delta\"):\n        assert name in delta.state_dict()\n        param.data += delta.state_dict()[name]\n\n    print(f\"Saving the target model to {target_model_path}\")\n    base.save_pretrained(target_model_path)\n    delta_tokenizer.save_pretrained(target_model_path)\n\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--base-model-path\", type=str, required=True)\n    parser.add_argument(\"--target-model-path\", type=str, required=True)\n    parser.add_argument(\"--delta-path\", type=str, required=True)\n    parser.add_argument(\n        \"--low-cpu-mem\",\n        action=\"store_true\",\n        help=\"Lower the cpu memory usage. This will split large files and use \"\n        \"disk as swap to reduce the memory usage below 10GB.\",\n    )\n    args = parser.parse_args()\n\n    if args.low_cpu_mem:\n        apply_delta_low_cpu_mem(\n            args.base_model_path, args.target_model_path, args.delta_path\n        )\n    else:\n        apply_delta(args.base_model_path, args.target_model_path, args.delta_path)","source_hash":"b33fef0c645fd573827306a76bf3f767a0fadbdbeb9f30cbd4e599fbc5a095c8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.apply_delta.split_files","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.apply_delta.split_files#L27-L69","kind":"function","name":"split_files","path":"gorilla/gorilla-main/inference/apply_delta.py","language":"python","start_line":27,"end_line":69,"context_start_line":7,"context_end_line":89,"code":"Thanks to LMSYS for the template of this code.\n\"\"\"\nimport argparse\nimport gc\nimport glob\nimport json\nimport os\nimport shutil\nimport tempfile\n\nfrom huggingface_hub import snapshot_download\nimport torch\nfrom torch import nn\nfrom tqdm import tqdm\nfrom transformers import AutoTokenizer, AutoModelForCausalLM, AutoConfig\n\n\nGB = 1 << 30\n\n\ndef split_files(model_path, tmp_path, split_size):\n    if not os.path.exists(model_path):\n        model_path = snapshot_download(repo_id=model_path)\n    if not os.path.exists(tmp_path):\n        os.makedirs(tmp_path)\n\n    file_pattern = os.path.join(model_path, \"pytorch_model-*.bin\")\n    files = glob.glob(file_pattern)\n\n    part = 0\n    try:\n        for file_path in tqdm(files):\n            state_dict = torch.load(file_path)\n            new_state_dict = {}\n\n            current_size = 0\n            for name, param in state_dict.items():\n                param_size = param.numel() * param.element_size()\n\n                if current_size + param_size > split_size:\n                    new_file_name = f\"pytorch_model-{part}.bin\"\n                    new_file_path = os.path.join(tmp_path, new_file_name)\n                    torch.save(new_state_dict, new_file_path)\n                    current_size = 0\n                    new_state_dict = None\n                    gc.collect()\n                    new_state_dict = {}\n                    part += 1\n\n                new_state_dict[name] = param\n                current_size += param_size\n\n            new_file_name = f\"pytorch_model-{part}.bin\"\n            new_file_path = os.path.join(tmp_path, new_file_name)\n            torch.save(new_state_dict, new_file_path)\n            new_state_dict = None\n            gc.collect()\n            new_state_dict = {}\n            part += 1\n    except Exception as e:\n        print(f\"An error occurred during split_files: {e}\")\n        shutil.rmtree(tmp_path)\n        raise\n\n\ndef apply_delta_low_cpu_mem(base_model_path, target_model_path, delta_path):\n    delta_tokenizer = AutoTokenizer.from_pretrained(delta_path, use_fast=False)\n    delta_config = AutoConfig.from_pretrained(delta_path)\n\n    if os.path.exists(target_model_path):\n        shutil.rmtree(target_model_path)\n    os.makedirs(target_model_path)\n\n    split_size = 4 * GB\n\n    with tempfile.TemporaryDirectory() as tmp_base_path, tempfile.TemporaryDirectory() as tmp_delta_path:\n        print(f\"Split files for the base model to {tmp_base_path}\")\n        split_files(base_model_path, tmp_base_path, split_size)\n        print(f\"Split files for the delta weights to {tmp_delta_path}\")\n        split_files(delta_path, tmp_delta_path, split_size)\n\n        base_pattern = os.path.join(tmp_base_path, \"pytorch_model-*.bin\")\n        base_files = glob.glob(base_pattern)","source_hash":"b33fef0c645fd573827306a76bf3f767a0fadbdbeb9f30cbd4e599fbc5a095c8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.apply_delta.apply_delta_low_cpu_mem","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.apply_delta.apply_delta_low_cpu_mem#L72-L124","kind":"function","name":"apply_delta_low_cpu_mem","path":"gorilla/gorilla-main/inference/apply_delta.py","language":"python","start_line":72,"end_line":124,"context_start_line":52,"context_end_line":144,"code":"                    gc.collect()\n                    new_state_dict = {}\n                    part += 1\n\n                new_state_dict[name] = param\n                current_size += param_size\n\n            new_file_name = f\"pytorch_model-{part}.bin\"\n            new_file_path = os.path.join(tmp_path, new_file_name)\n            torch.save(new_state_dict, new_file_path)\n            new_state_dict = None\n            gc.collect()\n            new_state_dict = {}\n            part += 1\n    except Exception as e:\n        print(f\"An error occurred during split_files: {e}\")\n        shutil.rmtree(tmp_path)\n        raise\n\n\ndef apply_delta_low_cpu_mem(base_model_path, target_model_path, delta_path):\n    delta_tokenizer = AutoTokenizer.from_pretrained(delta_path, use_fast=False)\n    delta_config = AutoConfig.from_pretrained(delta_path)\n\n    if os.path.exists(target_model_path):\n        shutil.rmtree(target_model_path)\n    os.makedirs(target_model_path)\n\n    split_size = 4 * GB\n\n    with tempfile.TemporaryDirectory() as tmp_base_path, tempfile.TemporaryDirectory() as tmp_delta_path:\n        print(f\"Split files for the base model to {tmp_base_path}\")\n        split_files(base_model_path, tmp_base_path, split_size)\n        print(f\"Split files for the delta weights to {tmp_delta_path}\")\n        split_files(delta_path, tmp_delta_path, split_size)\n\n        base_pattern = os.path.join(tmp_base_path, \"pytorch_model-*.bin\")\n        base_files = glob.glob(base_pattern)\n        delta_pattern = os.path.join(tmp_delta_path, \"pytorch_model-*.bin\")\n        delta_files = glob.glob(delta_pattern)\n        delta_state_dict = torch.load(delta_files[0])\n\n        print(\"Applying the delta\")\n        weight_map = {}\n        total_size = 0\n\n        for i, base_file in tqdm(enumerate(base_files)):\n            state_dict = torch.load(base_file)\n            file_name = f\"pytorch_model-{i}.bin\"\n            for name, param in state_dict.items():\n                if name not in delta_state_dict:\n                    for delta_file in delta_files:\n                        delta_state_dict = torch.load(delta_file)\n                        gc.collect()\n                        if name in delta_state_dict:\n                            break\n\n                state_dict[name] += delta_state_dict[name]\n                weight_map[name] = file_name\n                total_size += param.numel() * param.element_size()\n                gc.collect()\n            torch.save(state_dict, os.path.join(target_model_path, file_name))\n\n        with open(\n            os.path.join(target_model_path, \"pytorch_model.bin.index.json\"), \"w\"\n        ) as f:\n            json.dump(\n                {\"weight_map\": weight_map, \"metadata\": {\"total_size\": total_size}}, f\n            )\n\n    print(f\"Saving the target model to {target_model_path}\")\n    delta_tokenizer.save_pretrained(target_model_path)\n    delta_config.save_pretrained(target_model_path)\n\n\ndef apply_delta(base_model_path, target_model_path, delta_path):\n    print(f\"Loading the delta weights from {delta_path}\")\n    delta_tokenizer = AutoTokenizer.from_pretrained(delta_path, use_fast=False)\n    delta = AutoModelForCausalLM.from_pretrained(\n        delta_path, torch_dtype=torch.float16, low_cpu_mem_usage=True\n    )\n\n    print(f\"Loading the base model from {base_model_path}\")\n    base = AutoModelForCausalLM.from_pretrained(\n        base_model_path, torch_dtype=torch.float16, low_cpu_mem_usage=True\n    )\n\n    print(\"Applying the delta\")\n    for name, param in tqdm(base.state_dict().items(), desc=\"Applying delta\"):\n        assert name in delta.state_dict()\n        param.data += delta.state_dict()[name]\n\n    print(f\"Saving the target model to {target_model_path}\")","source_hash":"b33fef0c645fd573827306a76bf3f767a0fadbdbeb9f30cbd4e599fbc5a095c8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.apply_delta.apply_delta","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.apply_delta.apply_delta#L127-L146","kind":"function","name":"apply_delta","path":"gorilla/gorilla-main/inference/apply_delta.py","language":"python","start_line":127,"end_line":146,"context_start_line":107,"context_end_line":166,"code":"                            break\n\n                state_dict[name] += delta_state_dict[name]\n                weight_map[name] = file_name\n                total_size += param.numel() * param.element_size()\n                gc.collect()\n            torch.save(state_dict, os.path.join(target_model_path, file_name))\n\n        with open(\n            os.path.join(target_model_path, \"pytorch_model.bin.index.json\"), \"w\"\n        ) as f:\n            json.dump(\n                {\"weight_map\": weight_map, \"metadata\": {\"total_size\": total_size}}, f\n            )\n\n    print(f\"Saving the target model to {target_model_path}\")\n    delta_tokenizer.save_pretrained(target_model_path)\n    delta_config.save_pretrained(target_model_path)\n\n\ndef apply_delta(base_model_path, target_model_path, delta_path):\n    print(f\"Loading the delta weights from {delta_path}\")\n    delta_tokenizer = AutoTokenizer.from_pretrained(delta_path, use_fast=False)\n    delta = AutoModelForCausalLM.from_pretrained(\n        delta_path, torch_dtype=torch.float16, low_cpu_mem_usage=True\n    )\n\n    print(f\"Loading the base model from {base_model_path}\")\n    base = AutoModelForCausalLM.from_pretrained(\n        base_model_path, torch_dtype=torch.float16, low_cpu_mem_usage=True\n    )\n\n    print(\"Applying the delta\")\n    for name, param in tqdm(base.state_dict().items(), desc=\"Applying delta\"):\n        assert name in delta.state_dict()\n        param.data += delta.state_dict()[name]\n\n    print(f\"Saving the target model to {target_model_path}\")\n    base.save_pretrained(target_model_path)\n    delta_tokenizer.save_pretrained(target_model_path)\n\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--base-model-path\", type=str, required=True)\n    parser.add_argument(\"--target-model-path\", type=str, required=True)\n    parser.add_argument(\"--delta-path\", type=str, required=True)\n    parser.add_argument(\n        \"--low-cpu-mem\",\n        action=\"store_true\",\n        help=\"Lower the cpu memory usage. This will split large files and use \"\n        \"disk as swap to reduce the memory usage below 10GB.\",\n    )\n    args = parser.parse_args()\n\n    if args.low_cpu_mem:\n        apply_delta_low_cpu_mem(\n            args.base_model_path, args.target_model_path, args.delta_path\n        )\n    else:","source_hash":"b33fef0c645fd573827306a76bf3f767a0fadbdbeb9f30cbd4e599fbc5a095c8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.gorilla_eval","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.inference.gorilla_eval#L1-L176","kind":"module","name":"gorilla.gorilla-main.inference.gorilla_eval","path":"gorilla/gorilla-main/inference/gorilla_eval.py","language":"python","start_line":1,"end_line":176,"context_start_line":1,"context_end_line":176,"code":"import json\nimport argparse\nimport os\nfrom tqdm import tqdm\nimport torch\nfrom transformers import (\n    AutoConfig,\n    AutoModel,\n    AutoModelForCausalLM,\n    AutoModelForSeq2SeqLM,\n    AutoTokenizer,\n    LlamaTokenizer,\n    LlamaForCausalLM,\n    T5Tokenizer,\n)\n\n# Load Gorilla Model from HF\ndef load_model(\n        model_path: str,\n        device: str,\n        num_gpus: int,\n        max_gpu_memory: str = None,\n        load_8bit: bool = False,\n        cpu_offloading: bool = False,\n    ):\n \n    if device == \"cpu\":\n        kwargs = {\"torch_dtype\": torch.float32}\n    elif device == \"cuda\":\n        kwargs = {\"torch_dtype\": torch.float16}\n        if num_gpus != 1:\n            kwargs[\"device_map\"] = \"auto\"\n            if max_gpu_memory is None:\n                kwargs[\n                    \"device_map\"\n                ] = \"sequential\"  # This is important for not the same VRAM sizes\n                available_gpu_memory = get_gpu_memory(num_gpus)\n                kwargs[\"max_memory\"] = {\n                    i: str(int(available_gpu_memory[i] * 0.85)) + \"GiB\"\n                    for i in range(num_gpus)\n                }\n            else:\n                kwargs[\"max_memory\"] = {i: max_gpu_memory for i in range(num_gpus)}\n    else:\n        raise ValueError(f\"Invalid device: {device}\")\n\n    if cpu_offloading:\n        # raises an error on incompatible platforms\n        from transformers import BitsAndBytesConfig\n\n        if \"max_memory\" in kwargs:\n            kwargs[\"max_memory\"][\"cpu\"] = (\n                str(math.floor(psutil.virtual_memory().available / 2**20)) + \"Mib\"\n            )\n        kwargs[\"quantization_config\"] = BitsAndBytesConfig(\n            load_in_8bit_fp32_cpu_offload=cpu_offloading\n        )\n        kwargs[\"load_in_8bit\"] = load_8bit\n    elif load_8bit:\n        if num_gpus != 1:\n            warnings.warn(\n                \"8-bit quantization is not supported for multi-gpu inference.\"\n            )\n        else:\n            return load_compress_model(\n                model_path=model_path, device=device, torch_dtype=kwargs[\"torch_dtype\"]\n            )\n  \n    tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)\n    model = AutoModelForCausalLM.from_pretrained(\n        model_path,\n        low_cpu_mem_usage=True,\n        **kwargs,\n    )\n\n    return model, tokenizer\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\ndef run_eval(args, question_jsons):\n    # Evaluate the model for answers\n    model, tokenizer = load_model(\n        args.model_path, args.device, args.num_gpus, args.max_gpu_memory, args.load_8bit, args.cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n    # model = model.to(args.device)\n\n    ans_jsons = []\n    for i, line in enumerate(tqdm(question_jsons)):\n        ques_json = json.loads(line)\n        idx = ques_json[\"question_id\"]\n        prompt = ques_json[\"text\"]\n        prompt = \"###USER: \" + prompt + \"###ASSISTANT: \"\n        input_ids = tokenizer([prompt]).input_ids\n        output_ids = model.generate(\n            torch.as_tensor(input_ids).to(args.device),\n            do_sample=True,\n            temperature=0.7,\n            max_new_tokens=2048,\n        )\n        output_ids = output_ids[0][len(input_ids[0]) :]\n        outputs = tokenizer.decode(output_ids, skip_special_tokens=True).strip()\n        ans_jsons.append(\n            {\n                \"question_id\": idx,\n                \"questions\": prompt,\n                \"response\": outputs,\n            }\n        )\n\n    # Write output to file\n    with open(args.answer_file, \"w\") as ans_file:\n        for line in ans_jsons:\n            ans_file.write(json.dumps(line) + \"\\n\")\n\n    return ans_jsons\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\n        \"--model-path\", \n        type=str, \n        required=True)\n    parser.add_argument(\n        \"--question-file\", \n        type=str, \n        required=True)\n    parser.add_argument(\n        \"--device\", \n        type=str,\n        choices=[\"cpu\", \"cuda\", \"mps\"],\n        default=\"cuda\",\n        help=\"The device type\",\n    )\n    parser.add_argument(\n        \"--max-gpu-memory\",\n        type=str,\n        help=\"The maximum memory per gpu. A string like '13Gib'\",\n    )\n    parser.add_argument(\n        \"--load-8bit\", \n        action=\"store_true\", \n        help=\"Use 8-bit quantization\"\n    )\n    parser.add_argument(\n        \"--cpu-offloading\",\n        action=\"store_true\",\n        help=\"Only when using 8-bit quantization: Offload excess weights to the CPU that don't fit on the GPU\",\n    )\n    parser.add_argument(\n        \"--answer-file\", \n        type=str, \n        default=\"answer.jsonl\"\n    )\n    parser.add_argument(\n        \"--num-gpus\", \n        type=int, \n        default=1\n    )\n    args = parser.parse_args()\n\n    questions_json = get_questions(args.question_file)\n    run_eval(\n        args,\n        questions_json\n    )\n ","source_hash":"a5c70bd40bf9560b1db239ce916a8a3f7f4b2a798fed884fa623942ffcfca781","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.gorilla_eval.load_model","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.gorilla_eval.load_model#L18-L76","kind":"function","name":"load_model","path":"gorilla/gorilla-main/inference/gorilla_eval.py","language":"python","start_line":18,"end_line":76,"context_start_line":1,"context_end_line":96,"code":"import json\nimport argparse\nimport os\nfrom tqdm import tqdm\nimport torch\nfrom transformers import (\n    AutoConfig,\n    AutoModel,\n    AutoModelForCausalLM,\n    AutoModelForSeq2SeqLM,\n    AutoTokenizer,\n    LlamaTokenizer,\n    LlamaForCausalLM,\n    T5Tokenizer,\n)\n\n# Load Gorilla Model from HF\ndef load_model(\n        model_path: str,\n        device: str,\n        num_gpus: int,\n        max_gpu_memory: str = None,\n        load_8bit: bool = False,\n        cpu_offloading: bool = False,\n    ):\n \n    if device == \"cpu\":\n        kwargs = {\"torch_dtype\": torch.float32}\n    elif device == \"cuda\":\n        kwargs = {\"torch_dtype\": torch.float16}\n        if num_gpus != 1:\n            kwargs[\"device_map\"] = \"auto\"\n            if max_gpu_memory is None:\n                kwargs[\n                    \"device_map\"\n                ] = \"sequential\"  # This is important for not the same VRAM sizes\n                available_gpu_memory = get_gpu_memory(num_gpus)\n                kwargs[\"max_memory\"] = {\n                    i: str(int(available_gpu_memory[i] * 0.85)) + \"GiB\"\n                    for i in range(num_gpus)\n                }\n            else:\n                kwargs[\"max_memory\"] = {i: max_gpu_memory for i in range(num_gpus)}\n    else:\n        raise ValueError(f\"Invalid device: {device}\")\n\n    if cpu_offloading:\n        # raises an error on incompatible platforms\n        from transformers import BitsAndBytesConfig\n\n        if \"max_memory\" in kwargs:\n            kwargs[\"max_memory\"][\"cpu\"] = (\n                str(math.floor(psutil.virtual_memory().available / 2**20)) + \"Mib\"\n            )\n        kwargs[\"quantization_config\"] = BitsAndBytesConfig(\n            load_in_8bit_fp32_cpu_offload=cpu_offloading\n        )\n        kwargs[\"load_in_8bit\"] = load_8bit\n    elif load_8bit:\n        if num_gpus != 1:\n            warnings.warn(\n                \"8-bit quantization is not supported for multi-gpu inference.\"\n            )\n        else:\n            return load_compress_model(\n                model_path=model_path, device=device, torch_dtype=kwargs[\"torch_dtype\"]\n            )\n  \n    tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)\n    model = AutoModelForCausalLM.from_pretrained(\n        model_path,\n        low_cpu_mem_usage=True,\n        **kwargs,\n    )\n\n    return model, tokenizer\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\ndef run_eval(args, question_jsons):\n    # Evaluate the model for answers\n    model, tokenizer = load_model(\n        args.model_path, args.device, args.num_gpus, args.max_gpu_memory, args.load_8bit, args.cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n    # model = model.to(args.device)\n","source_hash":"a5c70bd40bf9560b1db239ce916a8a3f7f4b2a798fed884fa623942ffcfca781","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.gorilla_eval.get_questions","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.gorilla_eval.get_questions#L78-L86","kind":"function","name":"get_questions","path":"gorilla/gorilla-main/inference/gorilla_eval.py","language":"python","start_line":78,"end_line":86,"context_start_line":58,"context_end_line":106,"code":"        kwargs[\"load_in_8bit\"] = load_8bit\n    elif load_8bit:\n        if num_gpus != 1:\n            warnings.warn(\n                \"8-bit quantization is not supported for multi-gpu inference.\"\n            )\n        else:\n            return load_compress_model(\n                model_path=model_path, device=device, torch_dtype=kwargs[\"torch_dtype\"]\n            )\n  \n    tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)\n    model = AutoModelForCausalLM.from_pretrained(\n        model_path,\n        low_cpu_mem_usage=True,\n        **kwargs,\n    )\n\n    return model, tokenizer\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\ndef run_eval(args, question_jsons):\n    # Evaluate the model for answers\n    model, tokenizer = load_model(\n        args.model_path, args.device, args.num_gpus, args.max_gpu_memory, args.load_8bit, args.cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n    # model = model.to(args.device)\n\n    ans_jsons = []\n    for i, line in enumerate(tqdm(question_jsons)):\n        ques_json = json.loads(line)\n        idx = ques_json[\"question_id\"]\n        prompt = ques_json[\"text\"]\n        prompt = \"###USER: \" + prompt + \"###ASSISTANT: \"\n        input_ids = tokenizer([prompt]).input_ids\n        output_ids = model.generate(\n            torch.as_tensor(input_ids).to(args.device),\n            do_sample=True,","source_hash":"a5c70bd40bf9560b1db239ce916a8a3f7f4b2a798fed884fa623942ffcfca781","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.gorilla_eval.run_eval","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.gorilla_eval.run_eval#L88-L125","kind":"function","name":"run_eval","path":"gorilla/gorilla-main/inference/gorilla_eval.py","language":"python","start_line":88,"end_line":125,"context_start_line":68,"context_end_line":145,"code":"  \n    tokenizer = AutoTokenizer.from_pretrained(model_path, use_fast=False)\n    model = AutoModelForCausalLM.from_pretrained(\n        model_path,\n        low_cpu_mem_usage=True,\n        **kwargs,\n    )\n\n    return model, tokenizer\n\ndef get_questions(question_file):\n \n    # Load questions file\n    question_jsons = []\n    with open(question_file, \"r\") as ques_file:\n        for line in ques_file:\n            question_jsons.append(line)\n\n    return question_jsons\n\ndef run_eval(args, question_jsons):\n    # Evaluate the model for answers\n    model, tokenizer = load_model(\n        args.model_path, args.device, args.num_gpus, args.max_gpu_memory, args.load_8bit, args.cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n    # model = model.to(args.device)\n\n    ans_jsons = []\n    for i, line in enumerate(tqdm(question_jsons)):\n        ques_json = json.loads(line)\n        idx = ques_json[\"question_id\"]\n        prompt = ques_json[\"text\"]\n        prompt = \"###USER: \" + prompt + \"###ASSISTANT: \"\n        input_ids = tokenizer([prompt]).input_ids\n        output_ids = model.generate(\n            torch.as_tensor(input_ids).to(args.device),\n            do_sample=True,\n            temperature=0.7,\n            max_new_tokens=2048,\n        )\n        output_ids = output_ids[0][len(input_ids[0]) :]\n        outputs = tokenizer.decode(output_ids, skip_special_tokens=True).strip()\n        ans_jsons.append(\n            {\n                \"question_id\": idx,\n                \"questions\": prompt,\n                \"response\": outputs,\n            }\n        )\n\n    # Write output to file\n    with open(args.answer_file, \"w\") as ans_file:\n        for line in ans_jsons:\n            ans_file.write(json.dumps(line) + \"\\n\")\n\n    return ans_jsons\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\n        \"--model-path\", \n        type=str, \n        required=True)\n    parser.add_argument(\n        \"--question-file\", \n        type=str, \n        required=True)\n    parser.add_argument(\n        \"--device\", \n        type=str,\n        choices=[\"cpu\", \"cuda\", \"mps\"],\n        default=\"cuda\",\n        help=\"The device type\",\n    )\n    parser.add_argument(\n        \"--max-gpu-memory\",","source_hash":"a5c70bd40bf9560b1db239ce916a8a3f7f4b2a798fed884fa623942ffcfca781","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.inference.serve.conv_template#L1-L257","kind":"module","name":"gorilla.gorilla-main.inference.serve.conv_template","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":1,"end_line":257,"context_start_line":1,"context_end_line":257,"code":"\"\"\"\nConversation prompt templates.\n\nThanks to LMSYS for the template of this code.\n\"\"\"\n\nimport dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Any, Dict\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Separator styles.\"\"\"\n\n    ADD_COLON_SINGLE = auto()\n    ADD_COLON_TWO = auto()\n    ADD_COLON_SPACE_SINGLE = auto()\n    NO_COLON_SINGLE = auto()\n    ADD_NEW_LINE_SINGLE = auto()\n    DOLLY = auto()\n    RWKV = auto()\n    PHOENIX = auto()\n    NEW_LINE = auto()\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n\n    # The name of this template\n    name: str\n    # The system prompt\n    system: str\n    # Two roles\n    roles: List[str]\n    # All messages. Each item is (role, message).\n    messages: List[List[str]]\n    # The number of few shot examples\n    offset: int\n    # Separators\n    sep_style: SeparatorStyle\n    sep: str\n    sep2: str = None\n    # Stop criteria (the default one is EOS token)\n    stop_str: str = None\n    # Stops generation if meeting any token in this list\n    stop_token_ids: List[int] = None\n\n    def get_prompt(self) -> str:\n        \"\"\"Get the prompt for generation.\"\"\"\n        if self.sep_style == SeparatorStyle.ADD_COLON_SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.ADD_COLON_TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.ADD_COLON_SPACE_SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \": \"  # must be end with a space\n            return ret\n        elif self.sep_style == SeparatorStyle.NO_COLON_SINGLE:\n            ret = self.system\n            for role, message in self.messages:\n                if message:\n                    ret += role + message + self.sep\n                else:\n                    ret += role\n            return ret\n        elif self.sep_style == SeparatorStyle.ADD_NEW_LINE_SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \"\\n\" + message + self.sep\n                else:\n                    ret += role + \"\\n\"\n            return ret\n        elif self.sep_style == SeparatorStyle.DOLLY:\n            seps = [self.sep, self.sep2]\n            ret = self.system\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \":\\n\" + message + seps[i % 2]\n                    if i % 2 == 1:\n                        ret += \"\\n\\n\"\n                else:\n                    ret += role + \":\\n\"\n            return ret\n        elif self.sep_style == SeparatorStyle.RWKV:\n            ret = self.system\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += (\n                        role\n                        + \": \"\n                        + message.replace(\"\\r\\n\", \"\\n\").replace(\"\\n\\n\", \"\\n\")\n                    )\n                    ret += \"\\n\\n\"\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.PHOENIX:\n            ret = self.system\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + \"<s>\" + message + \"</s>\"\n                else:\n                    ret += role + \": \" + \"<s>\"\n            return ret\n        elif self.sep_style == SeparatorStyle.NEW_LINE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \"\\n\" + message + self.sep                                                    \n                else:\n                    ret += role + \"\\n\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role: str, message: str):\n        \"\"\"Append a new message.\"\"\"\n        self.messages.append([role, message])\n\n    def update_last_message(self, message: str):\n        \"\"\"Update the last output.\n\n        The last message is typically set to be None when constructing the prompt,\n        so we need to update it in-place after getting the response from a model.\n        \"\"\"\n        self.messages[-1][1] = message\n\n    def to_gradio_chatbot(self):\n        \"\"\"Convert the conversation to gradio chatbot format\"\"\"\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def to_openai_api_messages(self):\n        \"\"\"Convert the conversation to OpenAI chat completion format.\"\"\"\n        ret = [{\"role\": \"system\", \"content\": self.system}]\n\n        for i, (_, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append({\"role\": \"user\", \"content\": msg})\n            else:\n                if msg is not None:\n                    ret.append({\"role\": \"assistant\", \"content\": msg})\n        return ret\n\n    def copy(self):\n        return Conversation(\n            name=self.name,\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2,\n            stop_str=self.stop_str,\n            stop_token_ids=self.stop_token_ids,\n        )\n\n    def dict(self):\n        return {\n            \"name\": self.name,\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n        }\n\n\n# A global registry for all conversation templates\nconv_templates: Dict[str, Conversation] = {}\n\n\ndef register_conv_template(template: Conversation, override: bool = False):\n    \"\"\"Register a new conversation template.\"\"\"\n    if not override:\n        assert template.name not in conv_templates, f\"{name} has been registered.\"\n    conv_templates[template.name] = template\n\n\ndef get_conv_template(name: str) -> Conversation:\n    \"\"\"Get a conversation template.\"\"\"\n    return conv_templates[name].copy()\n\n# Gorilla v0 template\nregister_conv_template(\n    Conversation(\n        name=\"gorilla_v0\",\n        system=\"A chat between a curious user and an artificial intelligence assistant. \"\n        \"The assistant gives helpful, detailed, and polite answers to the user's questions.\",\n        roles=(\"USER\", \"ASSISTANT\"),\n        messages=(),\n        offset=0,\n        sep_style=SeparatorStyle.ADD_COLON_TWO,\n        sep=\"\\n\",\n        sep2=\"</s>\",\n    )\n)\n\n# Falcon Template\nregister_conv_template(\n    Conversation(\n        name=\"falcon\",\n        system=\"\",\n        # system=\"A chat between a curious user and an artificial intelligence assistant. \"        \n        # \"The assistant gives helpful, detailed, and polite answers to the user's questions.\",    \n        roles=(\"User\", \"Assistant\"),\n        messages=(),\n        offset=0,\n        sep_style=SeparatorStyle.ADD_COLON_TWO,\n        sep=\" \",\n        sep2=\"<|endoftext|>\",\n    )\n)\n\n# MPT Template\nregister_conv_template(\n    Conversation(\n        name=\"mpt\",\n        system=\"\"\"system\n- You are a helpful assistant chatbot trained by MosaicML.\n- You answer questions.\n- You are excited to be able to help the user, but will refuse to do anything that could be considered harmful to the user.\n- You are more than just an information source, you are also able to write poetry, short stories, and make jokes.\n\"\"\",\n        roles=(\"user\", \"assistant\"),\n        messages=(),\n        offset=0,\n        sep_style=SeparatorStyle.NEW_LINE,\n        sep=\" \",\n        sep2=\"<|endoftext|>\",\n        stop_token_ids=[50278, 0],\n    )\n)\n","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.SeparatorStyle","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.inference.serve.conv_template.SeparatorStyle#L12-L23","kind":"class","name":"SeparatorStyle","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":12,"end_line":23,"context_start_line":1,"context_end_line":43,"code":"\"\"\"\nConversation prompt templates.\n\nThanks to LMSYS for the template of this code.\n\"\"\"\n\nimport dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Any, Dict\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Separator styles.\"\"\"\n\n    ADD_COLON_SINGLE = auto()\n    ADD_COLON_TWO = auto()\n    ADD_COLON_SPACE_SINGLE = auto()\n    NO_COLON_SINGLE = auto()\n    ADD_NEW_LINE_SINGLE = auto()\n    DOLLY = auto()\n    RWKV = auto()\n    PHOENIX = auto()\n    NEW_LINE = auto()\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n\n    # The name of this template\n    name: str\n    # The system prompt\n    system: str\n    # Two roles\n    roles: List[str]\n    # All messages. Each item is (role, message).\n    messages: List[List[str]]\n    # The number of few shot examples\n    offset: int\n    # Separators\n    sep_style: SeparatorStyle\n    sep: str\n    sep2: str = None\n    # Stop criteria (the default one is EOS token)","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.Conversation","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.inference.serve.conv_template.Conversation#L26-L189","kind":"class","name":"Conversation","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":26,"end_line":189,"context_start_line":6,"context_end_line":209,"code":"\nimport dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Any, Dict\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Separator styles.\"\"\"\n\n    ADD_COLON_SINGLE = auto()\n    ADD_COLON_TWO = auto()\n    ADD_COLON_SPACE_SINGLE = auto()\n    NO_COLON_SINGLE = auto()\n    ADD_NEW_LINE_SINGLE = auto()\n    DOLLY = auto()\n    RWKV = auto()\n    PHOENIX = auto()\n    NEW_LINE = auto()\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n\n    # The name of this template\n    name: str\n    # The system prompt\n    system: str\n    # Two roles\n    roles: List[str]\n    # All messages. Each item is (role, message).\n    messages: List[List[str]]\n    # The number of few shot examples\n    offset: int\n    # Separators\n    sep_style: SeparatorStyle\n    sep: str\n    sep2: str = None\n    # Stop criteria (the default one is EOS token)\n    stop_str: str = None\n    # Stops generation if meeting any token in this list\n    stop_token_ids: List[int] = None\n\n    def get_prompt(self) -> str:\n        \"\"\"Get the prompt for generation.\"\"\"\n        if self.sep_style == SeparatorStyle.ADD_COLON_SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.ADD_COLON_TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.ADD_COLON_SPACE_SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \": \"  # must be end with a space\n            return ret\n        elif self.sep_style == SeparatorStyle.NO_COLON_SINGLE:\n            ret = self.system\n            for role, message in self.messages:\n                if message:\n                    ret += role + message + self.sep\n                else:\n                    ret += role\n            return ret\n        elif self.sep_style == SeparatorStyle.ADD_NEW_LINE_SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \"\\n\" + message + self.sep\n                else:\n                    ret += role + \"\\n\"\n            return ret\n        elif self.sep_style == SeparatorStyle.DOLLY:\n            seps = [self.sep, self.sep2]\n            ret = self.system\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \":\\n\" + message + seps[i % 2]\n                    if i % 2 == 1:\n                        ret += \"\\n\\n\"\n                else:\n                    ret += role + \":\\n\"\n            return ret\n        elif self.sep_style == SeparatorStyle.RWKV:\n            ret = self.system\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += (\n                        role\n                        + \": \"\n                        + message.replace(\"\\r\\n\", \"\\n\").replace(\"\\n\\n\", \"\\n\")\n                    )\n                    ret += \"\\n\\n\"\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.PHOENIX:\n            ret = self.system\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + \"<s>\" + message + \"</s>\"\n                else:\n                    ret += role + \": \" + \"<s>\"\n            return ret\n        elif self.sep_style == SeparatorStyle.NEW_LINE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \"\\n\" + message + self.sep                                                    \n                else:\n                    ret += role + \"\\n\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role: str, message: str):\n        \"\"\"Append a new message.\"\"\"\n        self.messages.append([role, message])\n\n    def update_last_message(self, message: str):\n        \"\"\"Update the last output.\n\n        The last message is typically set to be None when constructing the prompt,\n        so we need to update it in-place after getting the response from a model.\n        \"\"\"\n        self.messages[-1][1] = message\n\n    def to_gradio_chatbot(self):\n        \"\"\"Convert the conversation to gradio chatbot format\"\"\"\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def to_openai_api_messages(self):\n        \"\"\"Convert the conversation to OpenAI chat completion format.\"\"\"\n        ret = [{\"role\": \"system\", \"content\": self.system}]\n\n        for i, (_, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append({\"role\": \"user\", \"content\": msg})\n            else:\n                if msg is not None:\n                    ret.append({\"role\": \"assistant\", \"content\": msg})\n        return ret\n\n    def copy(self):\n        return Conversation(\n            name=self.name,\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2,\n            stop_str=self.stop_str,\n            stop_token_ids=self.stop_token_ids,\n        )\n\n    def dict(self):\n        return {\n            \"name\": self.name,\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n        }\n\n\n# A global registry for all conversation templates\nconv_templates: Dict[str, Conversation] = {}\n\n\ndef register_conv_template(template: Conversation, override: bool = False):\n    \"\"\"Register a new conversation template.\"\"\"\n    if not override:\n        assert template.name not in conv_templates, f\"{name} has been registered.\"\n    conv_templates[template.name] = template\n\n\ndef get_conv_template(name: str) -> Conversation:\n    \"\"\"Get a conversation template.\"\"\"\n    return conv_templates[name].copy()\n\n# Gorilla v0 template\nregister_conv_template(\n    Conversation(","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.register_conv_template","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.conv_template.register_conv_template#L196-L200","kind":"function","name":"register_conv_template","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":196,"end_line":200,"context_start_line":176,"context_end_line":220,"code":"            sep=self.sep,\n            sep2=self.sep2,\n            stop_str=self.stop_str,\n            stop_token_ids=self.stop_token_ids,\n        )\n\n    def dict(self):\n        return {\n            \"name\": self.name,\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n        }\n\n\n# A global registry for all conversation templates\nconv_templates: Dict[str, Conversation] = {}\n\n\ndef register_conv_template(template: Conversation, override: bool = False):\n    \"\"\"Register a new conversation template.\"\"\"\n    if not override:\n        assert template.name not in conv_templates, f\"{name} has been registered.\"\n    conv_templates[template.name] = template\n\n\ndef get_conv_template(name: str) -> Conversation:\n    \"\"\"Get a conversation template.\"\"\"\n    return conv_templates[name].copy()\n\n# Gorilla v0 template\nregister_conv_template(\n    Conversation(\n        name=\"gorilla_v0\",\n        system=\"A chat between a curious user and an artificial intelligence assistant. \"\n        \"The assistant gives helpful, detailed, and polite answers to the user's questions.\",\n        roles=(\"USER\", \"ASSISTANT\"),\n        messages=(),\n        offset=0,\n        sep_style=SeparatorStyle.ADD_COLON_TWO,\n        sep=\"\\n\",\n        sep2=\"</s>\",\n    )\n)","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.get_conv_template","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.conv_template.get_conv_template#L203-L205","kind":"function","name":"get_conv_template","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":203,"end_line":205,"context_start_line":183,"context_end_line":225,"code":"        return {\n            \"name\": self.name,\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n        }\n\n\n# A global registry for all conversation templates\nconv_templates: Dict[str, Conversation] = {}\n\n\ndef register_conv_template(template: Conversation, override: bool = False):\n    \"\"\"Register a new conversation template.\"\"\"\n    if not override:\n        assert template.name not in conv_templates, f\"{name} has been registered.\"\n    conv_templates[template.name] = template\n\n\ndef get_conv_template(name: str) -> Conversation:\n    \"\"\"Get a conversation template.\"\"\"\n    return conv_templates[name].copy()\n\n# Gorilla v0 template\nregister_conv_template(\n    Conversation(\n        name=\"gorilla_v0\",\n        system=\"A chat between a curious user and an artificial intelligence assistant. \"\n        \"The assistant gives helpful, detailed, and polite answers to the user's questions.\",\n        roles=(\"USER\", \"ASSISTANT\"),\n        messages=(),\n        offset=0,\n        sep_style=SeparatorStyle.ADD_COLON_TWO,\n        sep=\"\\n\",\n        sep2=\"</s>\",\n    )\n)\n\n# Falcon Template\nregister_conv_template(\n    Conversation(\n        name=\"falcon\",","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.get_prompt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.conv_template.get_prompt#L48-L132","kind":"function","name":"get_prompt","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":48,"end_line":132,"context_start_line":28,"context_end_line":152,"code":"\n    # The name of this template\n    name: str\n    # The system prompt\n    system: str\n    # Two roles\n    roles: List[str]\n    # All messages. Each item is (role, message).\n    messages: List[List[str]]\n    # The number of few shot examples\n    offset: int\n    # Separators\n    sep_style: SeparatorStyle\n    sep: str\n    sep2: str = None\n    # Stop criteria (the default one is EOS token)\n    stop_str: str = None\n    # Stops generation if meeting any token in this list\n    stop_token_ids: List[int] = None\n\n    def get_prompt(self) -> str:\n        \"\"\"Get the prompt for generation.\"\"\"\n        if self.sep_style == SeparatorStyle.ADD_COLON_SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.ADD_COLON_TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.ADD_COLON_SPACE_SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \": \"  # must be end with a space\n            return ret\n        elif self.sep_style == SeparatorStyle.NO_COLON_SINGLE:\n            ret = self.system\n            for role, message in self.messages:\n                if message:\n                    ret += role + message + self.sep\n                else:\n                    ret += role\n            return ret\n        elif self.sep_style == SeparatorStyle.ADD_NEW_LINE_SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \"\\n\" + message + self.sep\n                else:\n                    ret += role + \"\\n\"\n            return ret\n        elif self.sep_style == SeparatorStyle.DOLLY:\n            seps = [self.sep, self.sep2]\n            ret = self.system\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \":\\n\" + message + seps[i % 2]\n                    if i % 2 == 1:\n                        ret += \"\\n\\n\"\n                else:\n                    ret += role + \":\\n\"\n            return ret\n        elif self.sep_style == SeparatorStyle.RWKV:\n            ret = self.system\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += (\n                        role\n                        + \": \"\n                        + message.replace(\"\\r\\n\", \"\\n\").replace(\"\\n\\n\", \"\\n\")\n                    )\n                    ret += \"\\n\\n\"\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.PHOENIX:\n            ret = self.system\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + \"<s>\" + message + \"</s>\"\n                else:\n                    ret += role + \": \" + \"<s>\"\n            return ret\n        elif self.sep_style == SeparatorStyle.NEW_LINE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \"\\n\" + message + self.sep                                                    \n                else:\n                    ret += role + \"\\n\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role: str, message: str):\n        \"\"\"Append a new message.\"\"\"\n        self.messages.append([role, message])\n\n    def update_last_message(self, message: str):\n        \"\"\"Update the last output.\n\n        The last message is typically set to be None when constructing the prompt,\n        so we need to update it in-place after getting the response from a model.\n        \"\"\"\n        self.messages[-1][1] = message\n\n    def to_gradio_chatbot(self):\n        \"\"\"Convert the conversation to gradio chatbot format\"\"\"\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.append_message","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.conv_template.append_message#L134-L136","kind":"function","name":"append_message","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":134,"end_line":136,"context_start_line":114,"context_end_line":156,"code":"            return ret\n        elif self.sep_style == SeparatorStyle.PHOENIX:\n            ret = self.system\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + \"<s>\" + message + \"</s>\"\n                else:\n                    ret += role + \": \" + \"<s>\"\n            return ret\n        elif self.sep_style == SeparatorStyle.NEW_LINE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \"\\n\" + message + self.sep                                                    \n                else:\n                    ret += role + \"\\n\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role: str, message: str):\n        \"\"\"Append a new message.\"\"\"\n        self.messages.append([role, message])\n\n    def update_last_message(self, message: str):\n        \"\"\"Update the last output.\n\n        The last message is typically set to be None when constructing the prompt,\n        so we need to update it in-place after getting the response from a model.\n        \"\"\"\n        self.messages[-1][1] = message\n\n    def to_gradio_chatbot(self):\n        \"\"\"Convert the conversation to gradio chatbot format\"\"\"\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def to_openai_api_messages(self):","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.update_last_message","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.conv_template.update_last_message#L138-L144","kind":"function","name":"update_last_message","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":138,"end_line":144,"context_start_line":118,"context_end_line":164,"code":"                if message:\n                    ret += role + \": \" + \"<s>\" + message + \"</s>\"\n                else:\n                    ret += role + \": \" + \"<s>\"\n            return ret\n        elif self.sep_style == SeparatorStyle.NEW_LINE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \"\\n\" + message + self.sep                                                    \n                else:\n                    ret += role + \"\\n\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role: str, message: str):\n        \"\"\"Append a new message.\"\"\"\n        self.messages.append([role, message])\n\n    def update_last_message(self, message: str):\n        \"\"\"Update the last output.\n\n        The last message is typically set to be None when constructing the prompt,\n        so we need to update it in-place after getting the response from a model.\n        \"\"\"\n        self.messages[-1][1] = message\n\n    def to_gradio_chatbot(self):\n        \"\"\"Convert the conversation to gradio chatbot format\"\"\"\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def to_openai_api_messages(self):\n        \"\"\"Convert the conversation to OpenAI chat completion format.\"\"\"\n        ret = [{\"role\": \"system\", \"content\": self.system}]\n\n        for i, (_, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append({\"role\": \"user\", \"content\": msg})\n            else:\n                if msg is not None:","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.to_gradio_chatbot","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.conv_template.to_gradio_chatbot#L146-L154","kind":"function","name":"to_gradio_chatbot","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":146,"end_line":154,"context_start_line":126,"context_end_line":174,"code":"                if message:\n                    ret += role + \"\\n\" + message + self.sep                                                    \n                else:\n                    ret += role + \"\\n\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role: str, message: str):\n        \"\"\"Append a new message.\"\"\"\n        self.messages.append([role, message])\n\n    def update_last_message(self, message: str):\n        \"\"\"Update the last output.\n\n        The last message is typically set to be None when constructing the prompt,\n        so we need to update it in-place after getting the response from a model.\n        \"\"\"\n        self.messages[-1][1] = message\n\n    def to_gradio_chatbot(self):\n        \"\"\"Convert the conversation to gradio chatbot format\"\"\"\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def to_openai_api_messages(self):\n        \"\"\"Convert the conversation to OpenAI chat completion format.\"\"\"\n        ret = [{\"role\": \"system\", \"content\": self.system}]\n\n        for i, (_, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append({\"role\": \"user\", \"content\": msg})\n            else:\n                if msg is not None:\n                    ret.append({\"role\": \"assistant\", \"content\": msg})\n        return ret\n\n    def copy(self):\n        return Conversation(\n            name=self.name,\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.to_openai_api_messages","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.conv_template.to_openai_api_messages#L156-L166","kind":"function","name":"to_openai_api_messages","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":156,"end_line":166,"context_start_line":136,"context_end_line":186,"code":"        self.messages.append([role, message])\n\n    def update_last_message(self, message: str):\n        \"\"\"Update the last output.\n\n        The last message is typically set to be None when constructing the prompt,\n        so we need to update it in-place after getting the response from a model.\n        \"\"\"\n        self.messages[-1][1] = message\n\n    def to_gradio_chatbot(self):\n        \"\"\"Convert the conversation to gradio chatbot format\"\"\"\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def to_openai_api_messages(self):\n        \"\"\"Convert the conversation to OpenAI chat completion format.\"\"\"\n        ret = [{\"role\": \"system\", \"content\": self.system}]\n\n        for i, (_, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append({\"role\": \"user\", \"content\": msg})\n            else:\n                if msg is not None:\n                    ret.append({\"role\": \"assistant\", \"content\": msg})\n        return ret\n\n    def copy(self):\n        return Conversation(\n            name=self.name,\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2,\n            stop_str=self.stop_str,\n            stop_token_ids=self.stop_token_ids,\n        )\n\n    def dict(self):\n        return {\n            \"name\": self.name,\n            \"system\": self.system,\n            \"roles\": self.roles,","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.copy","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.conv_template.copy#L168-L180","kind":"function","name":"copy","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":168,"end_line":180,"context_start_line":148,"context_end_line":200,"code":"        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def to_openai_api_messages(self):\n        \"\"\"Convert the conversation to OpenAI chat completion format.\"\"\"\n        ret = [{\"role\": \"system\", \"content\": self.system}]\n\n        for i, (_, msg) in enumerate(self.messages[self.offset :]):\n            if i % 2 == 0:\n                ret.append({\"role\": \"user\", \"content\": msg})\n            else:\n                if msg is not None:\n                    ret.append({\"role\": \"assistant\", \"content\": msg})\n        return ret\n\n    def copy(self):\n        return Conversation(\n            name=self.name,\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2,\n            stop_str=self.stop_str,\n            stop_token_ids=self.stop_token_ids,\n        )\n\n    def dict(self):\n        return {\n            \"name\": self.name,\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n        }\n\n\n# A global registry for all conversation templates\nconv_templates: Dict[str, Conversation] = {}\n\n\ndef register_conv_template(template: Conversation, override: bool = False):\n    \"\"\"Register a new conversation template.\"\"\"\n    if not override:\n        assert template.name not in conv_templates, f\"{name} has been registered.\"\n    conv_templates[template.name] = template","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.conv_template.dict","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.conv_template.dict#L182-L189","kind":"function","name":"dict","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":182,"end_line":189,"context_start_line":162,"context_end_line":209,"code":"                ret.append({\"role\": \"user\", \"content\": msg})\n            else:\n                if msg is not None:\n                    ret.append({\"role\": \"assistant\", \"content\": msg})\n        return ret\n\n    def copy(self):\n        return Conversation(\n            name=self.name,\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2,\n            stop_str=self.stop_str,\n            stop_token_ids=self.stop_token_ids,\n        )\n\n    def dict(self):\n        return {\n            \"name\": self.name,\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n        }\n\n\n# A global registry for all conversation templates\nconv_templates: Dict[str, Conversation] = {}\n\n\ndef register_conv_template(template: Conversation, override: bool = False):\n    \"\"\"Register a new conversation template.\"\"\"\n    if not override:\n        assert template.name not in conv_templates, f\"{name} has been registered.\"\n    conv_templates[template.name] = template\n\n\ndef get_conv_template(name: str) -> Conversation:\n    \"\"\"Get a conversation template.\"\"\"\n    return conv_templates[name].copy()\n\n# Gorilla v0 template\nregister_conv_template(\n    Conversation(","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_falcon_cli","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.inference.serve.gorilla_falcon_cli#L1-L237","kind":"module","name":"gorilla.gorilla-main.inference.serve.gorilla_falcon_cli","path":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","language":"python","start_line":1,"end_line":237,"context_start_line":1,"context_end_line":237,"code":"\"\"\"\nChat with a model with command line interface.\n\nUsage:\npython3 -m gorilla_cli --model path/to/gorilla-7b-hf-v0\n\nThanks to LMSYS for the template of this code.\n\"\"\"\nimport argparse\nimport gc\nimport os\nimport re\nimport sys\nimport abc\nimport torch\nfrom transformers import (\n    AutoConfig,\n    AutoModel,\n    AutoModelForCausalLM,\n    AutoModelForSeq2SeqLM,\n    AutoTokenizer,\n    LlamaTokenizer,\n    LlamaForCausalLM,\n    T5Tokenizer,\n)\n\nfrom prompt_toolkit import PromptSession\nfrom prompt_toolkit.auto_suggest import AutoSuggestFromHistory\nfrom prompt_toolkit.completion import WordCompleter\nfrom prompt_toolkit.history import InMemoryHistory\nfrom conv_template import get_conv_template\n\nimport warnings\nwarnings.filterwarnings('ignore')\n\n# Load Gorilla Model from HF\ndef load_model(\n        model_path: str,\n        device: str,\n        num_gpus: int,\n        max_gpu_memory: str = None,\n        load_8bit: bool = False,\n        cpu_offloading: bool = False,\n    ):\n \n    if device == \"cpu\":\n        kwargs = {\"torch_dtype\": torch.float32}\n    elif device == \"cuda\":\n        kwargs = {\"torch_dtype\": torch.float16}\n        if num_gpus != 1:\n            kwargs[\"device_map\"] = \"auto\"\n            if max_gpu_memory is None:\n                kwargs[\n                    \"device_map\"\n                ] = \"sequential\"  # This is important for not the same VRAM sizes\n                available_gpu_memory = get_gpu_memory(num_gpus)\n                kwargs[\"max_memory\"] = {\n                    i: str(int(available_gpu_memory[i] * 0.85)) + \"GiB\"\n                    for i in range(num_gpus)\n                }\n            else:\n                kwargs[\"max_memory\"] = {i: max_gpu_memory for i in range(num_gpus)}\n    else:\n        raise ValueError(f\"Invalid device: {device}\")\n\n    if cpu_offloading:\n        # raises an error on incompatible platforms\n        from transformers import BitsAndBytesConfig\n\n        if \"max_memory\" in kwargs:\n            kwargs[\"max_memory\"][\"cpu\"] = (\n                str(math.floor(psutil.virtual_memory().available / 2**20)) + \"Mib\"\n            )\n        kwargs[\"quantization_config\"] = BitsAndBytesConfig(\n            load_in_8bit_fp32_cpu_offload=cpu_offloading\n        )\n        kwargs[\"load_in_8bit\"] = load_8bit\n    elif load_8bit:\n        if num_gpus != 1:\n            warnings.warn(\n                \"8-bit quantization is not supported for multi-gpu inference.\"\n            )\n        else:\n            return load_compress_model(\n                model_path=model_path, device=device, torch_dtype=kwargs[\"torch_dtype\"]\n            )\n  \n    tokenizer = AutoTokenizer.from_pretrained(model_path)\n    tokenizer.pad_token = tokenizer.eos_token\n    tokenizer.pad_token_id = 11\n    model = AutoModelForCausalLM.from_pretrained(\n        model_path,\n        trust_remote_code=True,\n        low_cpu_mem_usage=True,\n        **kwargs,\n    )\n\n    return model, tokenizer\n\n@torch.inference_mode()\ndef get_response(prompt, model, tokenizer, device):\n    input_ids = tokenizer([prompt]).input_ids\n    output_ids = model.generate(\n        torch.as_tensor(input_ids).to(device),\n        do_sample=True,\n        temperature=0.7,\n        max_new_tokens=1024,\n    )\n    output_ids = output_ids[0][len(input_ids[0]) :]\n    outputs = tokenizer.decode(output_ids, skip_special_tokens=True).strip()\n\n    yield {\"text\": outputs}\n\n    # clean\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,\n    max_gpu_memory: str,\n    load_8bit: bool,\n    cpu_offloading: bool,\n    chatio: abc.ABC,\n):\n    # Model\n    model, tokenizer = load_model(\n        model_path, device, num_gpus, max_gpu_memory, load_8bit, cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n\n    while True:\n        # Chat\n        if \"falcon\" in model_path:\n            conv = get_conv_template(\"falcon\")\n        elif \"mpt\" in model_path:\n            conv = get_conv_template(\"mpt\")\n        else:\n            conv = get_conv_template(\"gorilla_v0\")\n        \n        try:\n            inp = chatio.prompt_for_input(conv.roles[0])\n        except EOFError:\n            inp = \"\"\n        if not inp:\n            print(\"exit...\")\n            break\n\n        conv.append_message(conv.roles[0], inp)\n        conv.append_message(conv.roles[1], None)\n        prompt = conv.get_prompt()\n\n        chatio.prompt_for_output(conv.roles[1])\n        output_stream = get_response(prompt, model, tokenizer, device)\n        outputs = chatio.stream_output(output_stream)\n        conv.update_last_message(outputs.strip())\n\ndef main(args):\n    if args.gpus:\n        if len(args.gpus.split(\",\")) < args.num_gpus:\n            raise ValueError(\n                f\"Larger --num-gpus ({args.num_gpus}) than --gpus {args.gpus}!\"\n            )\n        os.environ[\"CUDA_VISIBLE_DEVICES\"] = args.gpus\n\n    chatio = SimpleChatIO()\n    \n    try:\n        chat_loop(\n            args.model_path,\n            args.device,\n            args.num_gpus,\n            args.max_gpu_memory,\n            args.load_8bit,\n            args.cpu_offloading,\n            chatio,\n        )\n    except KeyboardInterrupt:\n        print(\"exit...\")\n\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n\n    parser.add_argument(\n        \"--model-path\", type=str, default=None, \n        help=\"Model path to the pretrained model.\"\n    )\n    parser.add_argument(\n        \"--gpus\", type=str, default=None,\n        help=\"A single GPU like 1 or multiple GPUs like 0,2.\"\n    )\n    parser.add_argument(\n        \"--num-gpus\", \n        type=int, \n        default=1)\n    parser.add_argument(\n        \"--device\", type=str, default='cuda',\n        help=\"Which device to use.\"\n    )\n    parser.add_argument(\n        \"--max-gpu-memory\",\n        type=str,\n        help=\"The maximum memory per gpu. Use a string like '13Gib'\",\n    )\n    parser.add_argument(\n        \"--load-8bit\", action=\"store_true\", help=\"Use 8-bit quantization\"\n    )\n    parser.add_argument(\n        \"--cpu-offloading\",\n        action=\"store_true\",\n        help=\"Only when using 8-bit quantization: Offload excess weights to the CPU that don't fit on the GPU\",\n    )\n\n    args = parser.parse_args()\n    main(args)","source_hash":"3bdbae2592db9c84ea5e72251871a044816e35c09c9b7c59adee9337373a14ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.load_model","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.load_model#L37-L98","kind":"function","name":"load_model","path":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","language":"python","start_line":37,"end_line":98,"context_start_line":17,"context_end_line":118,"code":"    AutoConfig,\n    AutoModel,\n    AutoModelForCausalLM,\n    AutoModelForSeq2SeqLM,\n    AutoTokenizer,\n    LlamaTokenizer,\n    LlamaForCausalLM,\n    T5Tokenizer,\n)\n\nfrom prompt_toolkit import PromptSession\nfrom prompt_toolkit.auto_suggest import AutoSuggestFromHistory\nfrom prompt_toolkit.completion import WordCompleter\nfrom prompt_toolkit.history import InMemoryHistory\nfrom conv_template import get_conv_template\n\nimport warnings\nwarnings.filterwarnings('ignore')\n\n# Load Gorilla Model from HF\ndef load_model(\n        model_path: str,\n        device: str,\n        num_gpus: int,\n        max_gpu_memory: str = None,\n        load_8bit: bool = False,\n        cpu_offloading: bool = False,\n    ):\n \n    if device == \"cpu\":\n        kwargs = {\"torch_dtype\": torch.float32}\n    elif device == \"cuda\":\n        kwargs = {\"torch_dtype\": torch.float16}\n        if num_gpus != 1:\n            kwargs[\"device_map\"] = \"auto\"\n            if max_gpu_memory is None:\n                kwargs[\n                    \"device_map\"\n                ] = \"sequential\"  # This is important for not the same VRAM sizes\n                available_gpu_memory = get_gpu_memory(num_gpus)\n                kwargs[\"max_memory\"] = {\n                    i: str(int(available_gpu_memory[i] * 0.85)) + \"GiB\"\n                    for i in range(num_gpus)\n                }\n            else:\n                kwargs[\"max_memory\"] = {i: max_gpu_memory for i in range(num_gpus)}\n    else:\n        raise ValueError(f\"Invalid device: {device}\")\n\n    if cpu_offloading:\n        # raises an error on incompatible platforms\n        from transformers import BitsAndBytesConfig\n\n        if \"max_memory\" in kwargs:\n            kwargs[\"max_memory\"][\"cpu\"] = (\n                str(math.floor(psutil.virtual_memory().available / 2**20)) + \"Mib\"\n            )\n        kwargs[\"quantization_config\"] = BitsAndBytesConfig(\n            load_in_8bit_fp32_cpu_offload=cpu_offloading\n        )\n        kwargs[\"load_in_8bit\"] = load_8bit\n    elif load_8bit:\n        if num_gpus != 1:\n            warnings.warn(\n                \"8-bit quantization is not supported for multi-gpu inference.\"\n            )\n        else:\n            return load_compress_model(\n                model_path=model_path, device=device, torch_dtype=kwargs[\"torch_dtype\"]\n            )\n  \n    tokenizer = AutoTokenizer.from_pretrained(model_path)\n    tokenizer.pad_token = tokenizer.eos_token\n    tokenizer.pad_token_id = 11\n    model = AutoModelForCausalLM.from_pretrained(\n        model_path,\n        trust_remote_code=True,\n        low_cpu_mem_usage=True,\n        **kwargs,\n    )\n\n    return model, tokenizer\n\n@torch.inference_mode()\ndef get_response(prompt, model, tokenizer, device):\n    input_ids = tokenizer([prompt]).input_ids\n    output_ids = model.generate(\n        torch.as_tensor(input_ids).to(device),\n        do_sample=True,\n        temperature=0.7,\n        max_new_tokens=1024,\n    )\n    output_ids = output_ids[0][len(input_ids[0]) :]\n    outputs = tokenizer.decode(output_ids, skip_special_tokens=True).strip()\n\n    yield {\"text\": outputs}\n\n    # clean\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):","source_hash":"3bdbae2592db9c84ea5e72251871a044816e35c09c9b7c59adee9337373a14ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.get_response","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.get_response#L101-L116","kind":"function","name":"get_response","path":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","language":"python","start_line":101,"end_line":116,"context_start_line":81,"context_end_line":136,"code":"                \"8-bit quantization is not supported for multi-gpu inference.\"\n            )\n        else:\n            return load_compress_model(\n                model_path=model_path, device=device, torch_dtype=kwargs[\"torch_dtype\"]\n            )\n  \n    tokenizer = AutoTokenizer.from_pretrained(model_path)\n    tokenizer.pad_token = tokenizer.eos_token\n    tokenizer.pad_token_id = 11\n    model = AutoModelForCausalLM.from_pretrained(\n        model_path,\n        trust_remote_code=True,\n        low_cpu_mem_usage=True,\n        **kwargs,\n    )\n\n    return model, tokenizer\n\n@torch.inference_mode()\ndef get_response(prompt, model, tokenizer, device):\n    input_ids = tokenizer([prompt]).input_ids\n    output_ids = model.generate(\n        torch.as_tensor(input_ids).to(device),\n        do_sample=True,\n        temperature=0.7,\n        max_new_tokens=1024,\n    )\n    output_ids = output_ids[0][len(input_ids[0]) :]\n    outputs = tokenizer.decode(output_ids, skip_special_tokens=True).strip()\n\n    yield {\"text\": outputs}\n\n    # clean\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n","source_hash":"3bdbae2592db9c84ea5e72251871a044816e35c09c9b7c59adee9337373a14ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.SimpleChatIO","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.SimpleChatIO#L118-L135","kind":"class","name":"SimpleChatIO","path":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","language":"python","start_line":118,"end_line":135,"context_start_line":98,"context_end_line":155,"code":"    return model, tokenizer\n\n@torch.inference_mode()\ndef get_response(prompt, model, tokenizer, device):\n    input_ids = tokenizer([prompt]).input_ids\n    output_ids = model.generate(\n        torch.as_tensor(input_ids).to(device),\n        do_sample=True,\n        temperature=0.7,\n        max_new_tokens=1024,\n    )\n    output_ids = output_ids[0][len(input_ids[0]) :]\n    outputs = tokenizer.decode(output_ids, skip_special_tokens=True).strip()\n\n    yield {\"text\": outputs}\n\n    # clean\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,\n    max_gpu_memory: str,\n    load_8bit: bool,\n    cpu_offloading: bool,\n    chatio: abc.ABC,\n):\n    # Model\n    model, tokenizer = load_model(\n        model_path, device, num_gpus, max_gpu_memory, load_8bit, cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n\n    while True:\n        # Chat\n        if \"falcon\" in model_path:","source_hash":"3bdbae2592db9c84ea5e72251871a044816e35c09c9b7c59adee9337373a14ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.chat_loop","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.chat_loop#L137-L177","kind":"function","name":"chat_loop","path":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","language":"python","start_line":137,"end_line":177,"context_start_line":117,"context_end_line":197,"code":"\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,\n    max_gpu_memory: str,\n    load_8bit: bool,\n    cpu_offloading: bool,\n    chatio: abc.ABC,\n):\n    # Model\n    model, tokenizer = load_model(\n        model_path, device, num_gpus, max_gpu_memory, load_8bit, cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n\n    while True:\n        # Chat\n        if \"falcon\" in model_path:\n            conv = get_conv_template(\"falcon\")\n        elif \"mpt\" in model_path:\n            conv = get_conv_template(\"mpt\")\n        else:\n            conv = get_conv_template(\"gorilla_v0\")\n        \n        try:\n            inp = chatio.prompt_for_input(conv.roles[0])\n        except EOFError:\n            inp = \"\"\n        if not inp:\n            print(\"exit...\")\n            break\n\n        conv.append_message(conv.roles[0], inp)\n        conv.append_message(conv.roles[1], None)\n        prompt = conv.get_prompt()\n\n        chatio.prompt_for_output(conv.roles[1])\n        output_stream = get_response(prompt, model, tokenizer, device)\n        outputs = chatio.stream_output(output_stream)\n        conv.update_last_message(outputs.strip())\n\ndef main(args):\n    if args.gpus:\n        if len(args.gpus.split(\",\")) < args.num_gpus:\n            raise ValueError(\n                f\"Larger --num-gpus ({args.num_gpus}) than --gpus {args.gpus}!\"\n            )\n        os.environ[\"CUDA_VISIBLE_DEVICES\"] = args.gpus\n\n    chatio = SimpleChatIO()\n    \n    try:\n        chat_loop(\n            args.model_path,\n            args.device,\n            args.num_gpus,\n            args.max_gpu_memory,\n            args.load_8bit,\n            args.cpu_offloading,\n            chatio,","source_hash":"3bdbae2592db9c84ea5e72251871a044816e35c09c9b7c59adee9337373a14ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.main","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.main#L179-L200","kind":"function","name":"main","path":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","language":"python","start_line":179,"end_line":200,"context_start_line":159,"context_end_line":220,"code":"        else:\n            conv = get_conv_template(\"gorilla_v0\")\n        \n        try:\n            inp = chatio.prompt_for_input(conv.roles[0])\n        except EOFError:\n            inp = \"\"\n        if not inp:\n            print(\"exit...\")\n            break\n\n        conv.append_message(conv.roles[0], inp)\n        conv.append_message(conv.roles[1], None)\n        prompt = conv.get_prompt()\n\n        chatio.prompt_for_output(conv.roles[1])\n        output_stream = get_response(prompt, model, tokenizer, device)\n        outputs = chatio.stream_output(output_stream)\n        conv.update_last_message(outputs.strip())\n\ndef main(args):\n    if args.gpus:\n        if len(args.gpus.split(\",\")) < args.num_gpus:\n            raise ValueError(\n                f\"Larger --num-gpus ({args.num_gpus}) than --gpus {args.gpus}!\"\n            )\n        os.environ[\"CUDA_VISIBLE_DEVICES\"] = args.gpus\n\n    chatio = SimpleChatIO()\n    \n    try:\n        chat_loop(\n            args.model_path,\n            args.device,\n            args.num_gpus,\n            args.max_gpu_memory,\n            args.load_8bit,\n            args.cpu_offloading,\n            chatio,\n        )\n    except KeyboardInterrupt:\n        print(\"exit...\")\n\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n\n    parser.add_argument(\n        \"--model-path\", type=str, default=None, \n        help=\"Model path to the pretrained model.\"\n    )\n    parser.add_argument(\n        \"--gpus\", type=str, default=None,\n        help=\"A single GPU like 1 or multiple GPUs like 0,2.\"\n    )\n    parser.add_argument(\n        \"--num-gpus\", \n        type=int, \n        default=1)\n    parser.add_argument(\n        \"--device\", type=str, default='cuda',\n        help=\"Which device to use.\"","source_hash":"3bdbae2592db9c84ea5e72251871a044816e35c09c9b7c59adee9337373a14ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.prompt_for_input","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.prompt_for_input#L119-L120","kind":"function","name":"prompt_for_input","path":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","language":"python","start_line":119,"end_line":120,"context_start_line":99,"context_end_line":140,"code":"\n@torch.inference_mode()\ndef get_response(prompt, model, tokenizer, device):\n    input_ids = tokenizer([prompt]).input_ids\n    output_ids = model.generate(\n        torch.as_tensor(input_ids).to(device),\n        do_sample=True,\n        temperature=0.7,\n        max_new_tokens=1024,\n    )\n    output_ids = output_ids[0][len(input_ids[0]) :]\n    outputs = tokenizer.decode(output_ids, skip_special_tokens=True).strip()\n\n    yield {\"text\": outputs}\n\n    # clean\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,","source_hash":"3bdbae2592db9c84ea5e72251871a044816e35c09c9b7c59adee9337373a14ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.prompt_for_output","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.prompt_for_output#L122-L123","kind":"function","name":"prompt_for_output","path":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","language":"python","start_line":122,"end_line":123,"context_start_line":102,"context_end_line":143,"code":"    input_ids = tokenizer([prompt]).input_ids\n    output_ids = model.generate(\n        torch.as_tensor(input_ids).to(device),\n        do_sample=True,\n        temperature=0.7,\n        max_new_tokens=1024,\n    )\n    output_ids = output_ids[0][len(input_ids[0]) :]\n    outputs = tokenizer.decode(output_ids, skip_special_tokens=True).strip()\n\n    yield {\"text\": outputs}\n\n    # clean\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,\n    max_gpu_memory: str,\n    load_8bit: bool,\n    cpu_offloading: bool,","source_hash":"3bdbae2592db9c84ea5e72251871a044816e35c09c9b7c59adee9337373a14ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.stream_output","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_falcon_cli.stream_output#L125-L135","kind":"function","name":"stream_output","path":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","language":"python","start_line":125,"end_line":135,"context_start_line":105,"context_end_line":155,"code":"        do_sample=True,\n        temperature=0.7,\n        max_new_tokens=1024,\n    )\n    output_ids = output_ids[0][len(input_ids[0]) :]\n    outputs = tokenizer.decode(output_ids, skip_special_tokens=True).strip()\n\n    yield {\"text\": outputs}\n\n    # clean\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,\n    max_gpu_memory: str,\n    load_8bit: bool,\n    cpu_offloading: bool,\n    chatio: abc.ABC,\n):\n    # Model\n    model, tokenizer = load_model(\n        model_path, device, num_gpus, max_gpu_memory, load_8bit, cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n\n    while True:\n        # Chat\n        if \"falcon\" in model_path:","source_hash":"3bdbae2592db9c84ea5e72251871a044816e35c09c9b7c59adee9337373a14ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_cli","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.inference.serve.gorilla_cli#L1-L304","kind":"module","name":"gorilla.gorilla-main.inference.serve.gorilla_cli","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":1,"end_line":304,"context_start_line":1,"context_end_line":304,"code":"\"\"\"\nChat with a model with command line interface.\n\nUsage:\npython3 gorilla_cli.py --model-path path/to/gorilla-7b-hf-v0\n\nThanks to LMSYS for the template of this code.\n\"\"\"\nimport argparse\nimport gc\nimport os\nimport re\nimport sys\nimport abc\nimport torch\nfrom transformers import (\n    AutoConfig,\n    AutoModel,\n    AutoModelForCausalLM,\n    AutoModelForSeq2SeqLM,\n    AutoTokenizer,\n    LlamaTokenizer,\n    LlamaForCausalLM,\n    T5Tokenizer,\n)\nfrom transformers.generation.logits_process import (\n    LogitsProcessorList,\n    RepetitionPenaltyLogitsProcessor,\n    TemperatureLogitsWarper,\n    TopKLogitsWarper,\n    TopPLogitsWarper,\n)\n\nfrom prompt_toolkit import PromptSession\nfrom prompt_toolkit.auto_suggest import AutoSuggestFromHistory\nfrom prompt_toolkit.completion import WordCompleter\nfrom prompt_toolkit.history import InMemoryHistory\nfrom conv_template import get_conv_template\n\nimport warnings\nwarnings.filterwarnings('ignore')\n\n# Load Gorilla Model from HF\ndef load_model(\n        model_path: str,\n        device: str,\n        num_gpus: int,\n        max_gpu_memory: str = None,\n        load_8bit: bool = False,\n        cpu_offloading: bool = False,\n    ):\n \n    if device == \"cpu\":\n        kwargs = {\"torch_dtype\": torch.float32}\n    elif device == \"cuda\":\n        kwargs = {\"torch_dtype\": torch.float16}\n        if num_gpus != 1:\n            kwargs[\"device_map\"] = \"auto\"\n            if max_gpu_memory is None:\n                kwargs[\n                    \"device_map\"\n                ] = \"sequential\"  # This is important for not the same VRAM sizes\n                available_gpu_memory = get_gpu_memory(num_gpus)\n                kwargs[\"max_memory\"] = {\n                    i: str(int(available_gpu_memory[i] * 0.85)) + \"GiB\"\n                    for i in range(num_gpus)\n                }\n            else:\n                kwargs[\"max_memory\"] = {i: max_gpu_memory for i in range(num_gpus)}\n    else:\n        raise ValueError(f\"Invalid device: {device}\")\n\n    if cpu_offloading:\n        # raises an error on incompatible platforms\n        from transformers import BitsAndBytesConfig\n\n        if \"max_memory\" in kwargs:\n            kwargs[\"max_memory\"][\"cpu\"] = (\n                str(math.floor(psutil.virtual_memory().available / 2**20)) + \"Mib\"\n            )\n        kwargs[\"quantization_config\"] = BitsAndBytesConfig(\n            load_in_8bit_fp32_cpu_offload=cpu_offloading\n        )\n        kwargs[\"load_in_8bit\"] = load_8bit\n    elif load_8bit:\n        if num_gpus != 1:\n            warnings.warn(\n                \"8-bit quantization is not supported for multi-gpu inference.\"\n            )\n        else:\n            return load_compress_model(\n                model_path=model_path, device=device, torch_dtype=kwargs[\"torch_dtype\"]\n            )\n  \n    tokenizer = AutoTokenizer.from_pretrained(model_path)\n    model = AutoModelForCausalLM.from_pretrained(\n        model_path,\n        trust_remote_code=True,\n        low_cpu_mem_usage=True,\n        **kwargs,\n    )\n\n    return model, tokenizer\n\ndef prepare_logits_processor(\n    temperature: float, repetition_penalty: float, top_p: float, top_k: int\n):\n    processor_list = LogitsProcessorList()\n\n    if temperature >= 1e-5 and temperature != 1.0:\n        processor_list.append(TemperatureLogitsWarper(temperature))\n    if repetition_penalty > 1.0:\n        processor_list.append(RepetitionPenaltyLogitsProcessor(repetition_penalty))\n    if 1e-8 <= top_p < 1.0:\n        processor_list.append(TopPLogitsWarper(top_p))\n    if top_k > 0:\n        processor_list.append(TopKLogitsWarper(top_k))\n    return processor_list\n\n@torch.inference_mode()\ndef get_response(prompt, model, tokenizer, device):\n\n    logits_processor = prepare_logits_processor(\n        0.1, 0.0, 1.0, -1\n    )\n\n    context_len = 2048\n    max_new_tokens = 1024\n    stream_interval=2\n    input_ids = tokenizer(prompt).input_ids\n    input_echo_len = len(input_ids)\n    output_ids = list(input_ids)\n    max_src_len = context_len - max_new_tokens - 8\n    input_ids = input_ids[-max_src_len:]\n    stop_token_ids = [tokenizer.eos_token_id]\n    \n    past_key_values = out = None\n    for i in range(max_new_tokens):\n        if i == 0:\n            out = model(torch.as_tensor([input_ids], device=device),\n                use_cache=True)\n            logits = out.logits\n            past_key_values = out.past_key_values\n        else:\n            out = model(\n                input_ids=torch.as_tensor([[token]], device=device),\n                use_cache=True,\n                past_key_values=past_key_values,\n            )\n            logits = out.logits\n            past_key_values = out.past_key_values\n\n        tmp_output_ids = None\n        last_token_logits = logits_processor(tmp_output_ids, logits[:, -1, :])[0]\n        probs = torch.softmax(last_token_logits, dim=-1)\n        token = int(torch.multinomial(probs, num_samples=1))\n        output_ids.append(token)\n\n        if token in stop_token_ids:\n            stopped = True\n        else:\n            stopped = False\n\n        if i % stream_interval == 0 or i == max_new_tokens - 1 or stopped:\n            tmp_output_ids = output_ids[input_echo_len:]\n            rfind_start = 0\n\n            output = tokenizer.decode(\n                tmp_output_ids,\n                skip_special_tokens=True,\n                spaces_between_special_tokens=False,\n            )\n\n            yield {\n                \"text\": output\n            }\n\n        if stopped:\n            break\n\n    yield {\"text\": output}\n\n    # clean\n    del past_key_values, out\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,\n    max_gpu_memory: str,\n    load_8bit: bool,\n    cpu_offloading: bool,\n    chatio: abc.ABC,\n):\n    # Model\n    model, tokenizer = load_model(\n        model_path, device, num_gpus, max_gpu_memory, load_8bit, cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n\n    while True:\n        # Chat\n        if \"mpt\" in model_path:\n            conv = get_conv_template(\"mpt\")\n        elif \"gorilla\" in model_path:\n            conv = get_conv_template(\"gorilla_v0\")\n        try:\n            inp = chatio.prompt_for_input(conv.roles[0])\n        except EOFError:\n            inp = \"\"\n        if not inp:\n            print(\"exit...\")\n            break\n\n        conv.append_message(conv.roles[0], inp)\n        conv.append_message(conv.roles[1], None)\n        prompt = conv.get_prompt()\n\n        chatio.prompt_for_output(conv.roles[1])\n        output_stream = get_response(prompt, model, tokenizer, device)\n        outputs = chatio.stream_output(output_stream)\n        conv.update_last_message(outputs.strip())\n\ndef main(args):\n    if args.gpus:\n        if len(args.gpus.split(\",\")) < args.num_gpus:\n            raise ValueError(\n                f\"Larger --num-gpus ({args.num_gpus}) than --gpus {args.gpus}!\"\n            )\n        os.environ[\"CUDA_VISIBLE_DEVICES\"] = args.gpus\n\n    chatio = SimpleChatIO()\n    \n    try:\n        chat_loop(\n            args.model_path,\n            args.device,\n            args.num_gpus,\n            args.max_gpu_memory,\n            args.load_8bit,\n            args.cpu_offloading,\n            chatio,\n        )\n    except KeyboardInterrupt:\n        print(\"exit...\")\n\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n\n    parser.add_argument(\n        \"--model-path\", type=str, default=None, \n        help=\"Model path to the pretrained model.\"\n    )\n    parser.add_argument(\n        \"--gpus\", type=str, default=None,\n        help=\"A single GPU like 1 or multiple GPUs like 0,2.\"\n    )\n    parser.add_argument(\n        \"--num-gpus\", \n        type=int, \n        default=1)\n    parser.add_argument(\n        \"--device\", type=str, default='cuda',\n        help=\"Which device to use.\"\n    )\n    parser.add_argument(\n        \"--max-gpu-memory\",\n        type=str,\n        help=\"The maximum memory per gpu. Use a string like '13Gib'\",\n    )\n    parser.add_argument(\n        \"--load-8bit\", action=\"store_true\", help=\"Use 8-bit quantization\"\n    )\n    parser.add_argument(\n        \"--cpu-offloading\",\n        action=\"store_true\",\n        help=\"Only when using 8-bit quantization: Offload excess weights to the CPU that don't fit on the GPU\",\n    )\n\n    args = parser.parse_args()\n    main(args)","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_cli.load_model","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_cli.load_model#L44-L103","kind":"function","name":"load_model","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":44,"end_line":103,"context_start_line":24,"context_end_line":123,"code":"    T5Tokenizer,\n)\nfrom transformers.generation.logits_process import (\n    LogitsProcessorList,\n    RepetitionPenaltyLogitsProcessor,\n    TemperatureLogitsWarper,\n    TopKLogitsWarper,\n    TopPLogitsWarper,\n)\n\nfrom prompt_toolkit import PromptSession\nfrom prompt_toolkit.auto_suggest import AutoSuggestFromHistory\nfrom prompt_toolkit.completion import WordCompleter\nfrom prompt_toolkit.history import InMemoryHistory\nfrom conv_template import get_conv_template\n\nimport warnings\nwarnings.filterwarnings('ignore')\n\n# Load Gorilla Model from HF\ndef load_model(\n        model_path: str,\n        device: str,\n        num_gpus: int,\n        max_gpu_memory: str = None,\n        load_8bit: bool = False,\n        cpu_offloading: bool = False,\n    ):\n \n    if device == \"cpu\":\n        kwargs = {\"torch_dtype\": torch.float32}\n    elif device == \"cuda\":\n        kwargs = {\"torch_dtype\": torch.float16}\n        if num_gpus != 1:\n            kwargs[\"device_map\"] = \"auto\"\n            if max_gpu_memory is None:\n                kwargs[\n                    \"device_map\"\n                ] = \"sequential\"  # This is important for not the same VRAM sizes\n                available_gpu_memory = get_gpu_memory(num_gpus)\n                kwargs[\"max_memory\"] = {\n                    i: str(int(available_gpu_memory[i] * 0.85)) + \"GiB\"\n                    for i in range(num_gpus)\n                }\n            else:\n                kwargs[\"max_memory\"] = {i: max_gpu_memory for i in range(num_gpus)}\n    else:\n        raise ValueError(f\"Invalid device: {device}\")\n\n    if cpu_offloading:\n        # raises an error on incompatible platforms\n        from transformers import BitsAndBytesConfig\n\n        if \"max_memory\" in kwargs:\n            kwargs[\"max_memory\"][\"cpu\"] = (\n                str(math.floor(psutil.virtual_memory().available / 2**20)) + \"Mib\"\n            )\n        kwargs[\"quantization_config\"] = BitsAndBytesConfig(\n            load_in_8bit_fp32_cpu_offload=cpu_offloading\n        )\n        kwargs[\"load_in_8bit\"] = load_8bit\n    elif load_8bit:\n        if num_gpus != 1:\n            warnings.warn(\n                \"8-bit quantization is not supported for multi-gpu inference.\"\n            )\n        else:\n            return load_compress_model(\n                model_path=model_path, device=device, torch_dtype=kwargs[\"torch_dtype\"]\n            )\n  \n    tokenizer = AutoTokenizer.from_pretrained(model_path)\n    model = AutoModelForCausalLM.from_pretrained(\n        model_path,\n        trust_remote_code=True,\n        low_cpu_mem_usage=True,\n        **kwargs,\n    )\n\n    return model, tokenizer\n\ndef prepare_logits_processor(\n    temperature: float, repetition_penalty: float, top_p: float, top_k: int\n):\n    processor_list = LogitsProcessorList()\n\n    if temperature >= 1e-5 and temperature != 1.0:\n        processor_list.append(TemperatureLogitsWarper(temperature))\n    if repetition_penalty > 1.0:\n        processor_list.append(RepetitionPenaltyLogitsProcessor(repetition_penalty))\n    if 1e-8 <= top_p < 1.0:\n        processor_list.append(TopPLogitsWarper(top_p))\n    if top_k > 0:\n        processor_list.append(TopKLogitsWarper(top_k))\n    return processor_list\n\n@torch.inference_mode()\ndef get_response(prompt, model, tokenizer, device):\n\n    logits_processor = prepare_logits_processor(","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_cli.prepare_logits_processor","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_cli.prepare_logits_processor#L105-L118","kind":"function","name":"prepare_logits_processor","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":105,"end_line":118,"context_start_line":85,"context_end_line":138,"code":"    elif load_8bit:\n        if num_gpus != 1:\n            warnings.warn(\n                \"8-bit quantization is not supported for multi-gpu inference.\"\n            )\n        else:\n            return load_compress_model(\n                model_path=model_path, device=device, torch_dtype=kwargs[\"torch_dtype\"]\n            )\n  \n    tokenizer = AutoTokenizer.from_pretrained(model_path)\n    model = AutoModelForCausalLM.from_pretrained(\n        model_path,\n        trust_remote_code=True,\n        low_cpu_mem_usage=True,\n        **kwargs,\n    )\n\n    return model, tokenizer\n\ndef prepare_logits_processor(\n    temperature: float, repetition_penalty: float, top_p: float, top_k: int\n):\n    processor_list = LogitsProcessorList()\n\n    if temperature >= 1e-5 and temperature != 1.0:\n        processor_list.append(TemperatureLogitsWarper(temperature))\n    if repetition_penalty > 1.0:\n        processor_list.append(RepetitionPenaltyLogitsProcessor(repetition_penalty))\n    if 1e-8 <= top_p < 1.0:\n        processor_list.append(TopPLogitsWarper(top_p))\n    if top_k > 0:\n        processor_list.append(TopKLogitsWarper(top_k))\n    return processor_list\n\n@torch.inference_mode()\ndef get_response(prompt, model, tokenizer, device):\n\n    logits_processor = prepare_logits_processor(\n        0.1, 0.0, 1.0, -1\n    )\n\n    context_len = 2048\n    max_new_tokens = 1024\n    stream_interval=2\n    input_ids = tokenizer(prompt).input_ids\n    input_echo_len = len(input_ids)\n    output_ids = list(input_ids)\n    max_src_len = context_len - max_new_tokens - 8\n    input_ids = input_ids[-max_src_len:]\n    stop_token_ids = [tokenizer.eos_token_id]\n    \n    past_key_values = out = None\n    for i in range(max_new_tokens):","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_cli.get_response","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_cli.get_response#L121-L186","kind":"function","name":"get_response","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":121,"end_line":186,"context_start_line":101,"context_end_line":206,"code":"    )\n\n    return model, tokenizer\n\ndef prepare_logits_processor(\n    temperature: float, repetition_penalty: float, top_p: float, top_k: int\n):\n    processor_list = LogitsProcessorList()\n\n    if temperature >= 1e-5 and temperature != 1.0:\n        processor_list.append(TemperatureLogitsWarper(temperature))\n    if repetition_penalty > 1.0:\n        processor_list.append(RepetitionPenaltyLogitsProcessor(repetition_penalty))\n    if 1e-8 <= top_p < 1.0:\n        processor_list.append(TopPLogitsWarper(top_p))\n    if top_k > 0:\n        processor_list.append(TopKLogitsWarper(top_k))\n    return processor_list\n\n@torch.inference_mode()\ndef get_response(prompt, model, tokenizer, device):\n\n    logits_processor = prepare_logits_processor(\n        0.1, 0.0, 1.0, -1\n    )\n\n    context_len = 2048\n    max_new_tokens = 1024\n    stream_interval=2\n    input_ids = tokenizer(prompt).input_ids\n    input_echo_len = len(input_ids)\n    output_ids = list(input_ids)\n    max_src_len = context_len - max_new_tokens - 8\n    input_ids = input_ids[-max_src_len:]\n    stop_token_ids = [tokenizer.eos_token_id]\n    \n    past_key_values = out = None\n    for i in range(max_new_tokens):\n        if i == 0:\n            out = model(torch.as_tensor([input_ids], device=device),\n                use_cache=True)\n            logits = out.logits\n            past_key_values = out.past_key_values\n        else:\n            out = model(\n                input_ids=torch.as_tensor([[token]], device=device),\n                use_cache=True,\n                past_key_values=past_key_values,\n            )\n            logits = out.logits\n            past_key_values = out.past_key_values\n\n        tmp_output_ids = None\n        last_token_logits = logits_processor(tmp_output_ids, logits[:, -1, :])[0]\n        probs = torch.softmax(last_token_logits, dim=-1)\n        token = int(torch.multinomial(probs, num_samples=1))\n        output_ids.append(token)\n\n        if token in stop_token_ids:\n            stopped = True\n        else:\n            stopped = False\n\n        if i % stream_interval == 0 or i == max_new_tokens - 1 or stopped:\n            tmp_output_ids = output_ids[input_echo_len:]\n            rfind_start = 0\n\n            output = tokenizer.decode(\n                tmp_output_ids,\n                skip_special_tokens=True,\n                spaces_between_special_tokens=False,\n            )\n\n            yield {\n                \"text\": output\n            }\n\n        if stopped:\n            break\n\n    yield {\"text\": output}\n\n    # clean\n    del past_key_values, out\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_cli.SimpleChatIO","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.inference.serve.gorilla_cli.SimpleChatIO#L188-L205","kind":"class","name":"SimpleChatIO","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":188,"end_line":205,"context_start_line":168,"context_end_line":225,"code":"            output = tokenizer.decode(\n                tmp_output_ids,\n                skip_special_tokens=True,\n                spaces_between_special_tokens=False,\n            )\n\n            yield {\n                \"text\": output\n            }\n\n        if stopped:\n            break\n\n    yield {\"text\": output}\n\n    # clean\n    del past_key_values, out\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,\n    max_gpu_memory: str,\n    load_8bit: bool,\n    cpu_offloading: bool,\n    chatio: abc.ABC,\n):\n    # Model\n    model, tokenizer = load_model(\n        model_path, device, num_gpus, max_gpu_memory, load_8bit, cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n\n    while True:\n        # Chat\n        if \"mpt\" in model_path:","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_cli.chat_loop","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_cli.chat_loop#L207-L244","kind":"function","name":"chat_loop","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":207,"end_line":244,"context_start_line":187,"context_end_line":264,"code":"\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,\n    max_gpu_memory: str,\n    load_8bit: bool,\n    cpu_offloading: bool,\n    chatio: abc.ABC,\n):\n    # Model\n    model, tokenizer = load_model(\n        model_path, device, num_gpus, max_gpu_memory, load_8bit, cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n\n    while True:\n        # Chat\n        if \"mpt\" in model_path:\n            conv = get_conv_template(\"mpt\")\n        elif \"gorilla\" in model_path:\n            conv = get_conv_template(\"gorilla_v0\")\n        try:\n            inp = chatio.prompt_for_input(conv.roles[0])\n        except EOFError:\n            inp = \"\"\n        if not inp:\n            print(\"exit...\")\n            break\n\n        conv.append_message(conv.roles[0], inp)\n        conv.append_message(conv.roles[1], None)\n        prompt = conv.get_prompt()\n\n        chatio.prompt_for_output(conv.roles[1])\n        output_stream = get_response(prompt, model, tokenizer, device)\n        outputs = chatio.stream_output(output_stream)\n        conv.update_last_message(outputs.strip())\n\ndef main(args):\n    if args.gpus:\n        if len(args.gpus.split(\",\")) < args.num_gpus:\n            raise ValueError(\n                f\"Larger --num-gpus ({args.num_gpus}) than --gpus {args.gpus}!\"\n            )\n        os.environ[\"CUDA_VISIBLE_DEVICES\"] = args.gpus\n\n    chatio = SimpleChatIO()\n    \n    try:\n        chat_loop(\n            args.model_path,\n            args.device,\n            args.num_gpus,\n            args.max_gpu_memory,\n            args.load_8bit,\n            args.cpu_offloading,\n            chatio,","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_cli.main","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_cli.main#L246-L267","kind":"function","name":"main","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":246,"end_line":267,"context_start_line":226,"context_end_line":287,"code":"            conv = get_conv_template(\"mpt\")\n        elif \"gorilla\" in model_path:\n            conv = get_conv_template(\"gorilla_v0\")\n        try:\n            inp = chatio.prompt_for_input(conv.roles[0])\n        except EOFError:\n            inp = \"\"\n        if not inp:\n            print(\"exit...\")\n            break\n\n        conv.append_message(conv.roles[0], inp)\n        conv.append_message(conv.roles[1], None)\n        prompt = conv.get_prompt()\n\n        chatio.prompt_for_output(conv.roles[1])\n        output_stream = get_response(prompt, model, tokenizer, device)\n        outputs = chatio.stream_output(output_stream)\n        conv.update_last_message(outputs.strip())\n\ndef main(args):\n    if args.gpus:\n        if len(args.gpus.split(\",\")) < args.num_gpus:\n            raise ValueError(\n                f\"Larger --num-gpus ({args.num_gpus}) than --gpus {args.gpus}!\"\n            )\n        os.environ[\"CUDA_VISIBLE_DEVICES\"] = args.gpus\n\n    chatio = SimpleChatIO()\n    \n    try:\n        chat_loop(\n            args.model_path,\n            args.device,\n            args.num_gpus,\n            args.max_gpu_memory,\n            args.load_8bit,\n            args.cpu_offloading,\n            chatio,\n        )\n    except KeyboardInterrupt:\n        print(\"exit...\")\n\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n\n    parser.add_argument(\n        \"--model-path\", type=str, default=None, \n        help=\"Model path to the pretrained model.\"\n    )\n    parser.add_argument(\n        \"--gpus\", type=str, default=None,\n        help=\"A single GPU like 1 or multiple GPUs like 0,2.\"\n    )\n    parser.add_argument(\n        \"--num-gpus\", \n        type=int, \n        default=1)\n    parser.add_argument(\n        \"--device\", type=str, default='cuda',\n        help=\"Which device to use.\"","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_cli.prompt_for_input","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_cli.prompt_for_input#L189-L190","kind":"function","name":"prompt_for_input","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":189,"end_line":190,"context_start_line":169,"context_end_line":210,"code":"                tmp_output_ids,\n                skip_special_tokens=True,\n                spaces_between_special_tokens=False,\n            )\n\n            yield {\n                \"text\": output\n            }\n\n        if stopped:\n            break\n\n    yield {\"text\": output}\n\n    # clean\n    del past_key_values, out\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_cli.prompt_for_output","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_cli.prompt_for_output#L192-L193","kind":"function","name":"prompt_for_output","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":192,"end_line":193,"context_start_line":172,"context_end_line":213,"code":"            )\n\n            yield {\n                \"text\": output\n            }\n\n        if stopped:\n            break\n\n    yield {\"text\": output}\n\n    # clean\n    del past_key_values, out\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,\n    max_gpu_memory: str,\n    load_8bit: bool,\n    cpu_offloading: bool,","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.inference.serve.gorilla_cli.stream_output","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.inference.serve.gorilla_cli.stream_output#L195-L205","kind":"function","name":"stream_output","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":195,"end_line":205,"context_start_line":175,"context_end_line":225,"code":"                \"text\": output\n            }\n\n        if stopped:\n            break\n\n    yield {\"text\": output}\n\n    # clean\n    del past_key_values, out\n    gc.collect()\n    torch.cuda.empty_cache()\n\nclass SimpleChatIO(abc.ABC):\n    def prompt_for_input(self, role) -> str:\n        return input(f\"{role}: \")\n\n    def prompt_for_output(self, role: str):\n        print(f\"{role}: \", end=\"\", flush=True)\n\n    def stream_output(self, output_stream):\n        pre = 0\n        for outputs in output_stream:\n            output_text = outputs[\"text\"]\n            output_text = output_text.strip().split(\" \")\n            now = len(output_text) - 1\n            if now > pre:\n                print(\" \".join(output_text[pre:now]), end=\" \", flush=True)\n                pre = now\n        print(\" \".join(output_text[pre:]), flush=True)\n        return \" \".join(output_text)\n\ndef chat_loop(\n    model_path: str,\n    device: str,\n    num_gpus: int,\n    max_gpu_memory: str,\n    load_8bit: bool,\n    cpu_offloading: bool,\n    chatio: abc.ABC,\n):\n    # Model\n    model, tokenizer = load_model(\n        model_path, device, num_gpus, max_gpu_memory, load_8bit, cpu_offloading\n    )\n    if (args.device == \"cuda\" and args.num_gpus == 1 and not args.cpu_offloading) or args.device == \"mps\":\n        model.to(args.device)\n\n    while True:\n        # Chat\n        if \"mpt\" in model_path:","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.get_llm_responses","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.get_llm_responses#L1-L116","kind":"module","name":"gorilla.gorilla-main.eval.get_llm_responses","path":"gorilla/gorilla-main/eval/get_llm_responses.py","language":"python","start_line":1,"end_line":116,"context_start_line":1,"context_end_line":116,"code":"import argparse\nimport sys\nimport json\nimport openai\nimport anthropic\nimport multiprocessing as mp\nimport time\n\ndef encode_question(question, api_name):\n    \"\"\"Encode multiple prompt instructions into a single string.\"\"\"\n    \n    prompts = []\n    if api_name == \"torchhub\":\n        domains = \"1. $DOMAIN is inferred from the task description and should include one of {Classification, Semantic Segmentation, Object Detection, Audio Separation, Video Classification, Text-to-Speech}.\"\n    elif api_name == \"huggingface\":\n        domains = \"1. $DOMAIN should include one of {Multimodal Feature Extraction, Multimodal Text-to-Image, Multimodal Image-to-Text, Multimodal Text-to-Video, \\\n        Multimodal Visual Question Answering, Multimodal Document Question Answer, Multimodal Graph Machine Learning, Computer Vision Depth Estimation,\\\n        Computer Vision Image Classification, Computer Vision Object Detection, Computer Vision Image Segmentation, Computer Vision Image-to-Image, \\\n        Computer Vision Unconditional Image Generation, Computer Vision Video Classification, Computer Vision Zero-Shor Image Classification, \\\n        Natural Language Processing Text Classification, Natural Language Processing Token Classification, Natural Language Processing Table Question Answering, \\\n        Natural Language Processing Question Answering, Natural Language Processing Zero-Shot Classification, Natural Language Processing Translation, \\\n        Natural Language Processing Summarization, Natural Language Processing Conversational, Natural Language Processing Text Generation, Natural Language Processing Fill-Mask,\\\n        Natural Language Processing Text2Text Generation, Natural Language Processing Sentence Similarity, Audio Text-to-Speech, Audio Automatic Speech Recognition, \\\n        Audio Audio-to-Audio, Audio Audio Classification, Audio Voice Activity Detection, Tabular Tabular Classification, Tabular Tabular Regression, \\\n        Reinforcement Learning Reinforcement Learning, Reinforcement Learning Robotics }\"\n    elif api_name == \"tensorhub\":\n        domains = \"1. $DOMAIN is inferred from the task description and should include one of {text-sequence-alignment, text-embedding, text-language-model, text-preprocessing, text-classification, text-generation, text-question-answering, text-retrieval-question-answering, text-segmentation, text-to-mel, image-classification, image-feature-vector, image-object-detection, image-segmentation, image-generator, image-pose-detection, image-rnn-agent, image-augmentation, image-classifier, image-style-transfer, image-aesthetic-quality, image-depth-estimation, image-super-resolution, image-deblurring, image-extrapolation, image-text-recognition, image-dehazing, image-deraining, image-enhancemenmt, image-classification-logits, image-frame-interpolation, image-text-detection, image-denoising, image-others, video-classification, video-feature-extraction, video-generation, video-audio-text, video-text, audio-embedding, audio-event-classification, audio-command-detection, audio-paralinguists-classification, audio-speech-to-text, audio-speech-synthesis, audio-synthesis, audio-pitch-extraction}\"\n    else:\n        print(\"Error: API name is not supported.\")\n\n    prompt = question + \"\\nWrite a python program in 1 to 2 lines to call API in \" + api_name + \".\\n\\nThe answer should follow the format: <<<domain>>> $DOMAIN, <<<api_call>>>: $API_CALL, <<<api_provider>>>: $API_PROVIDER, <<<explanation>>>: $EXPLANATION, <<<code>>>: $CODE}. Here are the requirements:\\n\" + domains + \"\\n2. The $API_CALL should have only 1 line of code that calls api.\\n3. The $API_PROVIDER should be the programming framework used.\\n4. $EXPLANATION should be a step-by-step explanation.\\n5. The $CODE is the python code.\\n6. Do not repeat the format in your answer.\"\n    prompts.append({\"role\": \"system\", \"content\": \"You are a helpful API writer who can write APIs based on requirements.\"})\n    prompts.append({\"role\": \"user\", \"content\": prompt})\n    return prompts\n\ndef get_response(get_response_input, api_key):\n    question, question_id, api_name, model = get_response_input\n    question = encode_question(question, api_name)\n    \n    try:\n        if \"gpt\" in model:\n            openai.api_key = api_key\n            responses = openai.ChatCompletion.create(\n                model=model,\n                messages=question,\n                n=1,\n                temperature=0,\n            )\n            response = responses['choices'][0]['message']['content']\n        elif \"claude\" in model:\n            client = anthropic.Client(api_key)\n            responses = client.completion(\n                prompt=f\"{anthropic.HUMAN_PROMPT} {question[0]['content']}{question[1]['content']}{anthropic.AI_PROMPT}\",\n                stop_sequences=[anthropic.HUMAN_PROMPT],\n                model=\"claude-v1\",\n                max_tokens_to_sample=2048,\n            )\n            response = responses[\"completion\"].strip()\n        else:\n            print(\"Error: Model is not supported.\")\n    except Exception as e:\n        print(\"Error:\", e)\n        return None\n        \n    print(\"=>\",)\n    return {'text': response, \"question_id\": question_id, \"answer_id\": \"None\", \"model_id\": model, \"metadata\": {}}\n\ndef process_entry(entry, api_key):\n    question, question_id, api_name, model = entry\n    result = get_response((question, question_id, api_name, model), api_key)\n    return result\n\ndef write_result_to_file(result, output_file):\n    global file_write_lock\n    with file_write_lock:\n        with open(output_file, \"a\") as outfile:\n            json.dump(result, outfile)\n            outfile.write(\"\\n\")\n\ndef callback_with_lock(result, output_file):\n    global file_write_lock\n    write_result_to_file(result, output_file, file_write_lock)\n\nif __name__ == '__main__':\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--model\", type=str, default=None, help=\"which model you want to use for eval, only support ['gpt*', 'claude*'] now\")\n    parser.add_argument(\"--api_key\", type=str, default=None, help=\"the api key provided for calling\")\n    parser.add_argument(\"--output_file\", type=str, default=None, help=\"the output file this script writes to\")\n    parser.add_argument(\"--question_data\", type=str, default=None, help=\"path to the questions data file\")\n    parser.add_argument(\"--api_name\", type=str, default=None, help=\"this will be the api dataset name you are testing, only support ['torchhub', 'tensorhun', 'huggingface'] now\")\n    args = parser.parse_args()\n\n    start_time = time.time()\n    # Read the question file\n    questions = []\n    question_ids = []\n    with open(args.question_data, 'r') as f:\n        for idx, line in enumerate(f):\n            questions.append(json.loads(line)[\"text\"])\n            question_ids.append(json.loads(line)[\"question_id\"])\n\n    file_write_lock = mp.Lock()\n    with mp.Pool(1) as pool:\n        results = []\n        for idx, (question, question_id) in enumerate(zip(questions, question_ids)):\n            result = pool.apply_async(\n                process_entry,\n                args=((question, question_id, args.api_name, args.model), args.api_key),\n                callback=lambda result: write_result_to_file(result, args.output_file),\n            )\n            results.append(result)\n        pool.close()\n        pool.join()\n\n    end_time = time.time()\n    print(\"Total time used: \", end_time - start_time)","source_hash":"a1f2396797292461fb9aa31463a8180134e82de030722fa8f65f8ae18b9f957c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.get_llm_responses.encode_question","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.get_llm_responses.encode_question#L9-L34","kind":"function","name":"encode_question","path":"gorilla/gorilla-main/eval/get_llm_responses.py","language":"python","start_line":9,"end_line":34,"context_start_line":1,"context_end_line":54,"code":"import argparse\nimport sys\nimport json\nimport openai\nimport anthropic\nimport multiprocessing as mp\nimport time\n\ndef encode_question(question, api_name):\n    \"\"\"Encode multiple prompt instructions into a single string.\"\"\"\n    \n    prompts = []\n    if api_name == \"torchhub\":\n        domains = \"1. $DOMAIN is inferred from the task description and should include one of {Classification, Semantic Segmentation, Object Detection, Audio Separation, Video Classification, Text-to-Speech}.\"\n    elif api_name == \"huggingface\":\n        domains = \"1. $DOMAIN should include one of {Multimodal Feature Extraction, Multimodal Text-to-Image, Multimodal Image-to-Text, Multimodal Text-to-Video, \\\n        Multimodal Visual Question Answering, Multimodal Document Question Answer, Multimodal Graph Machine Learning, Computer Vision Depth Estimation,\\\n        Computer Vision Image Classification, Computer Vision Object Detection, Computer Vision Image Segmentation, Computer Vision Image-to-Image, \\\n        Computer Vision Unconditional Image Generation, Computer Vision Video Classification, Computer Vision Zero-Shor Image Classification, \\\n        Natural Language Processing Text Classification, Natural Language Processing Token Classification, Natural Language Processing Table Question Answering, \\\n        Natural Language Processing Question Answering, Natural Language Processing Zero-Shot Classification, Natural Language Processing Translation, \\\n        Natural Language Processing Summarization, Natural Language Processing Conversational, Natural Language Processing Text Generation, Natural Language Processing Fill-Mask,\\\n        Natural Language Processing Text2Text Generation, Natural Language Processing Sentence Similarity, Audio Text-to-Speech, Audio Automatic Speech Recognition, \\\n        Audio Audio-to-Audio, Audio Audio Classification, Audio Voice Activity Detection, Tabular Tabular Classification, Tabular Tabular Regression, \\\n        Reinforcement Learning Reinforcement Learning, Reinforcement Learning Robotics }\"\n    elif api_name == \"tensorhub\":\n        domains = \"1. $DOMAIN is inferred from the task description and should include one of {text-sequence-alignment, text-embedding, text-language-model, text-preprocessing, text-classification, text-generation, text-question-answering, text-retrieval-question-answering, text-segmentation, text-to-mel, image-classification, image-feature-vector, image-object-detection, image-segmentation, image-generator, image-pose-detection, image-rnn-agent, image-augmentation, image-classifier, image-style-transfer, image-aesthetic-quality, image-depth-estimation, image-super-resolution, image-deblurring, image-extrapolation, image-text-recognition, image-dehazing, image-deraining, image-enhancemenmt, image-classification-logits, image-frame-interpolation, image-text-detection, image-denoising, image-others, video-classification, video-feature-extraction, video-generation, video-audio-text, video-text, audio-embedding, audio-event-classification, audio-command-detection, audio-paralinguists-classification, audio-speech-to-text, audio-speech-synthesis, audio-synthesis, audio-pitch-extraction}\"\n    else:\n        print(\"Error: API name is not supported.\")\n\n    prompt = question + \"\\nWrite a python program in 1 to 2 lines to call API in \" + api_name + \".\\n\\nThe answer should follow the format: <<<domain>>> $DOMAIN, <<<api_call>>>: $API_CALL, <<<api_provider>>>: $API_PROVIDER, <<<explanation>>>: $EXPLANATION, <<<code>>>: $CODE}. Here are the requirements:\\n\" + domains + \"\\n2. The $API_CALL should have only 1 line of code that calls api.\\n3. The $API_PROVIDER should be the programming framework used.\\n4. $EXPLANATION should be a step-by-step explanation.\\n5. The $CODE is the python code.\\n6. Do not repeat the format in your answer.\"\n    prompts.append({\"role\": \"system\", \"content\": \"You are a helpful API writer who can write APIs based on requirements.\"})\n    prompts.append({\"role\": \"user\", \"content\": prompt})\n    return prompts\n\ndef get_response(get_response_input, api_key):\n    question, question_id, api_name, model = get_response_input\n    question = encode_question(question, api_name)\n    \n    try:\n        if \"gpt\" in model:\n            openai.api_key = api_key\n            responses = openai.ChatCompletion.create(\n                model=model,\n                messages=question,\n                n=1,\n                temperature=0,\n            )\n            response = responses['choices'][0]['message']['content']\n        elif \"claude\" in model:\n            client = anthropic.Client(api_key)\n            responses = client.completion(\n                prompt=f\"{anthropic.HUMAN_PROMPT} {question[0]['content']}{question[1]['content']}{anthropic.AI_PROMPT}\",\n                stop_sequences=[anthropic.HUMAN_PROMPT],","source_hash":"a1f2396797292461fb9aa31463a8180134e82de030722fa8f65f8ae18b9f957c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.get_llm_responses.get_response","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.get_llm_responses.get_response#L36-L66","kind":"function","name":"get_response","path":"gorilla/gorilla-main/eval/get_llm_responses.py","language":"python","start_line":36,"end_line":66,"context_start_line":16,"context_end_line":86,"code":"        domains = \"1. $DOMAIN should include one of {Multimodal Feature Extraction, Multimodal Text-to-Image, Multimodal Image-to-Text, Multimodal Text-to-Video, \\\n        Multimodal Visual Question Answering, Multimodal Document Question Answer, Multimodal Graph Machine Learning, Computer Vision Depth Estimation,\\\n        Computer Vision Image Classification, Computer Vision Object Detection, Computer Vision Image Segmentation, Computer Vision Image-to-Image, \\\n        Computer Vision Unconditional Image Generation, Computer Vision Video Classification, Computer Vision Zero-Shor Image Classification, \\\n        Natural Language Processing Text Classification, Natural Language Processing Token Classification, Natural Language Processing Table Question Answering, \\\n        Natural Language Processing Question Answering, Natural Language Processing Zero-Shot Classification, Natural Language Processing Translation, \\\n        Natural Language Processing Summarization, Natural Language Processing Conversational, Natural Language Processing Text Generation, Natural Language Processing Fill-Mask,\\\n        Natural Language Processing Text2Text Generation, Natural Language Processing Sentence Similarity, Audio Text-to-Speech, Audio Automatic Speech Recognition, \\\n        Audio Audio-to-Audio, Audio Audio Classification, Audio Voice Activity Detection, Tabular Tabular Classification, Tabular Tabular Regression, \\\n        Reinforcement Learning Reinforcement Learning, Reinforcement Learning Robotics }\"\n    elif api_name == \"tensorhub\":\n        domains = \"1. $DOMAIN is inferred from the task description and should include one of {text-sequence-alignment, text-embedding, text-language-model, text-preprocessing, text-classification, text-generation, text-question-answering, text-retrieval-question-answering, text-segmentation, text-to-mel, image-classification, image-feature-vector, image-object-detection, image-segmentation, image-generator, image-pose-detection, image-rnn-agent, image-augmentation, image-classifier, image-style-transfer, image-aesthetic-quality, image-depth-estimation, image-super-resolution, image-deblurring, image-extrapolation, image-text-recognition, image-dehazing, image-deraining, image-enhancemenmt, image-classification-logits, image-frame-interpolation, image-text-detection, image-denoising, image-others, video-classification, video-feature-extraction, video-generation, video-audio-text, video-text, audio-embedding, audio-event-classification, audio-command-detection, audio-paralinguists-classification, audio-speech-to-text, audio-speech-synthesis, audio-synthesis, audio-pitch-extraction}\"\n    else:\n        print(\"Error: API name is not supported.\")\n\n    prompt = question + \"\\nWrite a python program in 1 to 2 lines to call API in \" + api_name + \".\\n\\nThe answer should follow the format: <<<domain>>> $DOMAIN, <<<api_call>>>: $API_CALL, <<<api_provider>>>: $API_PROVIDER, <<<explanation>>>: $EXPLANATION, <<<code>>>: $CODE}. Here are the requirements:\\n\" + domains + \"\\n2. The $API_CALL should have only 1 line of code that calls api.\\n3. The $API_PROVIDER should be the programming framework used.\\n4. $EXPLANATION should be a step-by-step explanation.\\n5. The $CODE is the python code.\\n6. Do not repeat the format in your answer.\"\n    prompts.append({\"role\": \"system\", \"content\": \"You are a helpful API writer who can write APIs based on requirements.\"})\n    prompts.append({\"role\": \"user\", \"content\": prompt})\n    return prompts\n\ndef get_response(get_response_input, api_key):\n    question, question_id, api_name, model = get_response_input\n    question = encode_question(question, api_name)\n    \n    try:\n        if \"gpt\" in model:\n            openai.api_key = api_key\n            responses = openai.ChatCompletion.create(\n                model=model,\n                messages=question,\n                n=1,\n                temperature=0,\n            )\n            response = responses['choices'][0]['message']['content']\n        elif \"claude\" in model:\n            client = anthropic.Client(api_key)\n            responses = client.completion(\n                prompt=f\"{anthropic.HUMAN_PROMPT} {question[0]['content']}{question[1]['content']}{anthropic.AI_PROMPT}\",\n                stop_sequences=[anthropic.HUMAN_PROMPT],\n                model=\"claude-v1\",\n                max_tokens_to_sample=2048,\n            )\n            response = responses[\"completion\"].strip()\n        else:\n            print(\"Error: Model is not supported.\")\n    except Exception as e:\n        print(\"Error:\", e)\n        return None\n        \n    print(\"=>\",)\n    return {'text': response, \"question_id\": question_id, \"answer_id\": \"None\", \"model_id\": model, \"metadata\": {}}\n\ndef process_entry(entry, api_key):\n    question, question_id, api_name, model = entry\n    result = get_response((question, question_id, api_name, model), api_key)\n    return result\n\ndef write_result_to_file(result, output_file):\n    global file_write_lock\n    with file_write_lock:\n        with open(output_file, \"a\") as outfile:\n            json.dump(result, outfile)\n            outfile.write(\"\\n\")\n\ndef callback_with_lock(result, output_file):\n    global file_write_lock\n    write_result_to_file(result, output_file, file_write_lock)\n\nif __name__ == '__main__':\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--model\", type=str, default=None, help=\"which model you want to use for eval, only support ['gpt*', 'claude*'] now\")","source_hash":"a1f2396797292461fb9aa31463a8180134e82de030722fa8f65f8ae18b9f957c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.get_llm_responses.process_entry","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.get_llm_responses.process_entry#L68-L71","kind":"function","name":"process_entry","path":"gorilla/gorilla-main/eval/get_llm_responses.py","language":"python","start_line":68,"end_line":71,"context_start_line":48,"context_end_line":91,"code":"            )\n            response = responses['choices'][0]['message']['content']\n        elif \"claude\" in model:\n            client = anthropic.Client(api_key)\n            responses = client.completion(\n                prompt=f\"{anthropic.HUMAN_PROMPT} {question[0]['content']}{question[1]['content']}{anthropic.AI_PROMPT}\",\n                stop_sequences=[anthropic.HUMAN_PROMPT],\n                model=\"claude-v1\",\n                max_tokens_to_sample=2048,\n            )\n            response = responses[\"completion\"].strip()\n        else:\n            print(\"Error: Model is not supported.\")\n    except Exception as e:\n        print(\"Error:\", e)\n        return None\n        \n    print(\"=>\",)\n    return {'text': response, \"question_id\": question_id, \"answer_id\": \"None\", \"model_id\": model, \"metadata\": {}}\n\ndef process_entry(entry, api_key):\n    question, question_id, api_name, model = entry\n    result = get_response((question, question_id, api_name, model), api_key)\n    return result\n\ndef write_result_to_file(result, output_file):\n    global file_write_lock\n    with file_write_lock:\n        with open(output_file, \"a\") as outfile:\n            json.dump(result, outfile)\n            outfile.write(\"\\n\")\n\ndef callback_with_lock(result, output_file):\n    global file_write_lock\n    write_result_to_file(result, output_file, file_write_lock)\n\nif __name__ == '__main__':\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--model\", type=str, default=None, help=\"which model you want to use for eval, only support ['gpt*', 'claude*'] now\")\n    parser.add_argument(\"--api_key\", type=str, default=None, help=\"the api key provided for calling\")\n    parser.add_argument(\"--output_file\", type=str, default=None, help=\"the output file this script writes to\")\n    parser.add_argument(\"--question_data\", type=str, default=None, help=\"path to the questions data file\")\n    parser.add_argument(\"--api_name\", type=str, default=None, help=\"this will be the api dataset name you are testing, only support ['torchhub', 'tensorhun', 'huggingface'] now\")\n    args = parser.parse_args()","source_hash":"a1f2396797292461fb9aa31463a8180134e82de030722fa8f65f8ae18b9f957c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.get_llm_responses.write_result_to_file","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.get_llm_responses.write_result_to_file#L73-L78","kind":"function","name":"write_result_to_file","path":"gorilla/gorilla-main/eval/get_llm_responses.py","language":"python","start_line":73,"end_line":78,"context_start_line":53,"context_end_line":98,"code":"                prompt=f\"{anthropic.HUMAN_PROMPT} {question[0]['content']}{question[1]['content']}{anthropic.AI_PROMPT}\",\n                stop_sequences=[anthropic.HUMAN_PROMPT],\n                model=\"claude-v1\",\n                max_tokens_to_sample=2048,\n            )\n            response = responses[\"completion\"].strip()\n        else:\n            print(\"Error: Model is not supported.\")\n    except Exception as e:\n        print(\"Error:\", e)\n        return None\n        \n    print(\"=>\",)\n    return {'text': response, \"question_id\": question_id, \"answer_id\": \"None\", \"model_id\": model, \"metadata\": {}}\n\ndef process_entry(entry, api_key):\n    question, question_id, api_name, model = entry\n    result = get_response((question, question_id, api_name, model), api_key)\n    return result\n\ndef write_result_to_file(result, output_file):\n    global file_write_lock\n    with file_write_lock:\n        with open(output_file, \"a\") as outfile:\n            json.dump(result, outfile)\n            outfile.write(\"\\n\")\n\ndef callback_with_lock(result, output_file):\n    global file_write_lock\n    write_result_to_file(result, output_file, file_write_lock)\n\nif __name__ == '__main__':\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--model\", type=str, default=None, help=\"which model you want to use for eval, only support ['gpt*', 'claude*'] now\")\n    parser.add_argument(\"--api_key\", type=str, default=None, help=\"the api key provided for calling\")\n    parser.add_argument(\"--output_file\", type=str, default=None, help=\"the output file this script writes to\")\n    parser.add_argument(\"--question_data\", type=str, default=None, help=\"path to the questions data file\")\n    parser.add_argument(\"--api_name\", type=str, default=None, help=\"this will be the api dataset name you are testing, only support ['torchhub', 'tensorhun', 'huggingface'] now\")\n    args = parser.parse_args()\n\n    start_time = time.time()\n    # Read the question file\n    questions = []\n    question_ids = []\n    with open(args.question_data, 'r') as f:\n        for idx, line in enumerate(f):","source_hash":"a1f2396797292461fb9aa31463a8180134e82de030722fa8f65f8ae18b9f957c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.get_llm_responses.callback_with_lock","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.get_llm_responses.callback_with_lock#L80-L82","kind":"function","name":"callback_with_lock","path":"gorilla/gorilla-main/eval/get_llm_responses.py","language":"python","start_line":80,"end_line":82,"context_start_line":60,"context_end_line":102,"code":"            print(\"Error: Model is not supported.\")\n    except Exception as e:\n        print(\"Error:\", e)\n        return None\n        \n    print(\"=>\",)\n    return {'text': response, \"question_id\": question_id, \"answer_id\": \"None\", \"model_id\": model, \"metadata\": {}}\n\ndef process_entry(entry, api_key):\n    question, question_id, api_name, model = entry\n    result = get_response((question, question_id, api_name, model), api_key)\n    return result\n\ndef write_result_to_file(result, output_file):\n    global file_write_lock\n    with file_write_lock:\n        with open(output_file, \"a\") as outfile:\n            json.dump(result, outfile)\n            outfile.write(\"\\n\")\n\ndef callback_with_lock(result, output_file):\n    global file_write_lock\n    write_result_to_file(result, output_file, file_write_lock)\n\nif __name__ == '__main__':\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--model\", type=str, default=None, help=\"which model you want to use for eval, only support ['gpt*', 'claude*'] now\")\n    parser.add_argument(\"--api_key\", type=str, default=None, help=\"the api key provided for calling\")\n    parser.add_argument(\"--output_file\", type=str, default=None, help=\"the output file this script writes to\")\n    parser.add_argument(\"--question_data\", type=str, default=None, help=\"path to the questions data file\")\n    parser.add_argument(\"--api_name\", type=str, default=None, help=\"this will be the api dataset name you are testing, only support ['torchhub', 'tensorhun', 'huggingface'] now\")\n    args = parser.parse_args()\n\n    start_time = time.time()\n    # Read the question file\n    questions = []\n    question_ids = []\n    with open(args.question_data, 'r') as f:\n        for idx, line in enumerate(f):\n            questions.append(json.loads(line)[\"text\"])\n            question_ids.append(json.loads(line)[\"question_id\"])\n\n    file_write_lock = mp.Lock()","source_hash":"a1f2396797292461fb9aa31463a8180134e82de030722fa8f65f8ae18b9f957c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_th","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.ast_eval_th#L1-L213","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.ast_eval_th","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","language":"python","start_line":1,"end_line":213,"context_start_line":1,"context_end_line":213,"code":"import argparse\nimport json\nfrom codebleu.parser import (\n    DFG_python,\n    DFG_java,\n    DFG_ruby,\n    DFG_go,\n    DFG_php,\n    DFG_javascript,\n    DFG_csharp,\n)\nfrom codebleu.parser import (\n    remove_comments_and_docstrings,\n    tree_to_token_index,\n    index_to_code_token,\n    tree_to_variable_index,\n)\nfrom tree_sitter import Language, Parser\nimport concurrent.futures\n\ndfg_function = {\n    \"python\": DFG_python,\n    \"java\": DFG_java,\n    \"ruby\": DFG_ruby,\n    \"go\": DFG_go,\n    \"php\": DFG_php,\n    \"javascript\": DFG_javascript,\n    \"c_sharp\": DFG_csharp,\n}\n\n# Get all the subtrees given a root_node\ndef get_all_sub_trees(root_node):\n    node_stack = []\n    sub_tree_sexp_list = []\n    depth = 1\n    text = root_node.text\n    node_stack.append([root_node, depth])\n    while len(node_stack) != 0:\n        cur_node, cur_depth = node_stack.pop()\n        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append(\n                [cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text]\n            )\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang=\"python\"):\n    LANGUAGE = Language(\"codebleu/parser/my-languages.so\", lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n\n    candidate_tree = parser.parse(bytes(candidate, \"utf8\")).root_node\n    return candidate_tree\n\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if \"repo_or_dir\" in child.text.decode() or \"model\" in child.text.decode():\n            args_list.append(child.children[2].text)\n    return args_list\n\n\n# Check if there is an api match\ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break\n        if ast_match:\n            return idx\n    return -1\n\n\n# Parse the dataset\ndef parse_dataset(args):\n    # Read the api dataset\n    api_database = []\n    with open(args.api_dataset, \"r\") as f:\n        for line in f:\n            api_database.append(json.loads(line))\n\n    # Read the question answer pair dataset\n    qa_pairs = []\n    with open(args.apibench, \"r\") as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n\n    # Read the language model response dataset\n    llm_responses = []\n    with open(args.llm_responses, \"r\") as f:\n        for line in f:\n            llm_responses.append(json.loads(line))\n\n    # Parse all APIs to AST trees\n    ast_database = []\n    with concurrent.futures.ThreadPoolExecutor() as executor:\n        ast_trees = executor.map(ast_parse, (data[\"api_call\"] for data in api_database))\n        for ast_tree in ast_trees:\n            ast_database.append(ast_tree)\n\n    return api_database, qa_pairs, llm_responses, ast_database\n\n\ndef process_response(response, api_database, qa_pairs, ast_database):\n    # Read the line from JSON file\n    try:\n        output = response[\"text\"]\n    except:\n        print(\"Error: cannot parse line \", response[\"index\"])\n        return False, False\n\n    # Index the \"api_call\" domain\n    output = output.split(\"api_call\")\n    if len(output) == 1:\n        return False, False\n    else:\n        output = output[1].split(\"api_provider\")[0]\n        if \":\" not in output:\n            start = 0\n        else:\n            start = output.index(\":\")\n        if \")\" not in output:\n            end = -2\n        else:\n            end = output.rindex(\")\")\n        api_call = output[start + 2 : end + 1]\n\n    # Parse the api_call into AST tree\n    ast_tree = ast_parse(api_call)\n    # Search for a subtree\n    ast_subtree_list = get_all_sub_trees(ast_tree)\n    # Check which ast tree is matching\n    database_index = ast_check(ast_subtree_list, ast_database)\n    # We cannot index this ast in our database\n    if database_index == -1:\n        return False, True\n    # We index our reference api_call\n    ref_api_call = api_database[database_index]\n    # Check for functionality\n    if ref_api_call[\"domain\"] == qa_pairs[response[\"question_id\"] - 1][\"domain\"]:\n        return True, False\n    else:\n        return False, False\n\n\ndef main(args):\n    # Read datasets\n    api_database, qa_pairs, llm_responses, ast_database = parse_dataset(args)\n\n    # Check correctness\n    total_correct = 0\n    total_hallucination = 0\n    num_responses = len(llm_responses)\n\n    with concurrent.futures.ThreadPoolExecutor() as executor:\n        results = [\n            executor.submit(\n                process_response,\n                response,\n                api_database,\n                qa_pairs,\n                ast_database,\n            )\n            for response in llm_responses\n        ]\n\n        for result in concurrent.futures.as_completed(results):\n            correct, hallucination = result.result()\n            if correct:\n                total_correct += 1\n            if hallucination:\n                total_hallucination += 1\n\n    print(\"Final Functionality accuracy:\", total_correct / num_responses)\n    print(\"Final hallucination:\", total_hallucination / num_responses)\n\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--api_dataset\", type=str, default=None, help=\"path to your api dataset\")\n    parser.add_argument(\n        \"--apibench\",\n        type=str,\n        default=None,\n        help=\"path to your apibench dataset including the question and answer pairs\",\n    )\n    parser.add_argument(\"--llm_responses\", type=str, default=None, help=\"path to the language model responses\")\n    args = parser.parse_args()\n    main(args)","source_hash":"2ef51deb15484e775819fcaadf0d4372048d1371641d7f1025bed961f7a664e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.get_all_sub_trees","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.get_all_sub_trees#L32-L50","kind":"function","name":"get_all_sub_trees","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","language":"python","start_line":32,"end_line":50,"context_start_line":12,"context_end_line":70,"code":"from codebleu.parser import (\n    remove_comments_and_docstrings,\n    tree_to_token_index,\n    index_to_code_token,\n    tree_to_variable_index,\n)\nfrom tree_sitter import Language, Parser\nimport concurrent.futures\n\ndfg_function = {\n    \"python\": DFG_python,\n    \"java\": DFG_java,\n    \"ruby\": DFG_ruby,\n    \"go\": DFG_go,\n    \"php\": DFG_php,\n    \"javascript\": DFG_javascript,\n    \"c_sharp\": DFG_csharp,\n}\n\n# Get all the subtrees given a root_node\ndef get_all_sub_trees(root_node):\n    node_stack = []\n    sub_tree_sexp_list = []\n    depth = 1\n    text = root_node.text\n    node_stack.append([root_node, depth])\n    while len(node_stack) != 0:\n        cur_node, cur_depth = node_stack.pop()\n        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append(\n                [cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text]\n            )\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang=\"python\"):\n    LANGUAGE = Language(\"codebleu/parser/my-languages.so\", lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n\n    candidate_tree = parser.parse(bytes(candidate, \"utf8\")).root_node\n    return candidate_tree\n\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if \"repo_or_dir\" in child.text.decode() or \"model\" in child.text.decode():\n            args_list.append(child.children[2].text)","source_hash":"2ef51deb15484e775819fcaadf0d4372048d1371641d7f1025bed961f7a664e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.ast_parse","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.ast_parse#L54-L60","kind":"function","name":"ast_parse","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","language":"python","start_line":54,"end_line":60,"context_start_line":34,"context_end_line":80,"code":"    sub_tree_sexp_list = []\n    depth = 1\n    text = root_node.text\n    node_stack.append([root_node, depth])\n    while len(node_stack) != 0:\n        cur_node, cur_depth = node_stack.pop()\n        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append(\n                [cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text]\n            )\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang=\"python\"):\n    LANGUAGE = Language(\"codebleu/parser/my-languages.so\", lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n\n    candidate_tree = parser.parse(bytes(candidate, \"utf8\")).root_node\n    return candidate_tree\n\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if \"repo_or_dir\" in child.text.decode() or \"model\" in child.text.decode():\n            args_list.append(child.children[2].text)\n    return args_list\n\n\n# Check if there is an api match\ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:","source_hash":"2ef51deb15484e775819fcaadf0d4372048d1371641d7f1025bed961f7a664e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.get_args","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.get_args#L64-L71","kind":"function","name":"get_args","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","language":"python","start_line":64,"end_line":71,"context_start_line":44,"context_end_line":91,"code":"        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang=\"python\"):\n    LANGUAGE = Language(\"codebleu/parser/my-languages.so\", lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n\n    candidate_tree = parser.parse(bytes(candidate, \"utf8\")).root_node\n    return candidate_tree\n\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if \"repo_or_dir\" in child.text.decode() or \"model\" in child.text.decode():\n            args_list.append(child.children[2].text)\n    return args_list\n\n\n# Check if there is an api match\ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False","source_hash":"2ef51deb15484e775819fcaadf0d4372048d1371641d7f1025bed961f7a664e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.ast_check","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.ast_check#L75-L95","kind":"function","name":"ast_check","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","language":"python","start_line":75,"end_line":95,"context_start_line":55,"context_end_line":115,"code":"    LANGUAGE = Language(\"codebleu/parser/my-languages.so\", lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n\n    candidate_tree = parser.parse(bytes(candidate, \"utf8\")).root_node\n    return candidate_tree\n\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if \"repo_or_dir\" in child.text.decode() or \"model\" in child.text.decode():\n            args_list.append(child.children[2].text)\n    return args_list\n\n\n# Check if there is an api match\ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break\n        if ast_match:\n            return idx\n    return -1\n\n\n# Parse the dataset\ndef parse_dataset(args):\n    # Read the api dataset\n    api_database = []\n    with open(args.api_dataset, \"r\") as f:\n        for line in f:\n            api_database.append(json.loads(line))\n\n    # Read the question answer pair dataset\n    qa_pairs = []\n    with open(args.apibench, \"r\") as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n\n    # Read the language model response dataset\n    llm_responses = []\n    with open(args.llm_responses, \"r\") as f:\n        for line in f:","source_hash":"2ef51deb15484e775819fcaadf0d4372048d1371641d7f1025bed961f7a664e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.parse_dataset","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.parse_dataset#L99-L125","kind":"function","name":"parse_dataset","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","language":"python","start_line":99,"end_line":125,"context_start_line":79,"context_end_line":145,"code":"        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break\n        if ast_match:\n            return idx\n    return -1\n\n\n# Parse the dataset\ndef parse_dataset(args):\n    # Read the api dataset\n    api_database = []\n    with open(args.api_dataset, \"r\") as f:\n        for line in f:\n            api_database.append(json.loads(line))\n\n    # Read the question answer pair dataset\n    qa_pairs = []\n    with open(args.apibench, \"r\") as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n\n    # Read the language model response dataset\n    llm_responses = []\n    with open(args.llm_responses, \"r\") as f:\n        for line in f:\n            llm_responses.append(json.loads(line))\n\n    # Parse all APIs to AST trees\n    ast_database = []\n    with concurrent.futures.ThreadPoolExecutor() as executor:\n        ast_trees = executor.map(ast_parse, (data[\"api_call\"] for data in api_database))\n        for ast_tree in ast_trees:\n            ast_database.append(ast_tree)\n\n    return api_database, qa_pairs, llm_responses, ast_database\n\n\ndef process_response(response, api_database, qa_pairs, ast_database):\n    # Read the line from JSON file\n    try:\n        output = response[\"text\"]\n    except:\n        print(\"Error: cannot parse line \", response[\"index\"])\n        return False, False\n\n    # Index the \"api_call\" domain\n    output = output.split(\"api_call\")\n    if len(output) == 1:\n        return False, False\n    else:\n        output = output[1].split(\"api_provider\")[0]\n        if \":\" not in output:\n            start = 0\n        else:\n            start = output.index(\":\")","source_hash":"2ef51deb15484e775819fcaadf0d4372048d1371641d7f1025bed961f7a664e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.process_response","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.process_response#L128-L167","kind":"function","name":"process_response","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","language":"python","start_line":128,"end_line":167,"context_start_line":108,"context_end_line":187,"code":"    with open(args.apibench, \"r\") as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n\n    # Read the language model response dataset\n    llm_responses = []\n    with open(args.llm_responses, \"r\") as f:\n        for line in f:\n            llm_responses.append(json.loads(line))\n\n    # Parse all APIs to AST trees\n    ast_database = []\n    with concurrent.futures.ThreadPoolExecutor() as executor:\n        ast_trees = executor.map(ast_parse, (data[\"api_call\"] for data in api_database))\n        for ast_tree in ast_trees:\n            ast_database.append(ast_tree)\n\n    return api_database, qa_pairs, llm_responses, ast_database\n\n\ndef process_response(response, api_database, qa_pairs, ast_database):\n    # Read the line from JSON file\n    try:\n        output = response[\"text\"]\n    except:\n        print(\"Error: cannot parse line \", response[\"index\"])\n        return False, False\n\n    # Index the \"api_call\" domain\n    output = output.split(\"api_call\")\n    if len(output) == 1:\n        return False, False\n    else:\n        output = output[1].split(\"api_provider\")[0]\n        if \":\" not in output:\n            start = 0\n        else:\n            start = output.index(\":\")\n        if \")\" not in output:\n            end = -2\n        else:\n            end = output.rindex(\")\")\n        api_call = output[start + 2 : end + 1]\n\n    # Parse the api_call into AST tree\n    ast_tree = ast_parse(api_call)\n    # Search for a subtree\n    ast_subtree_list = get_all_sub_trees(ast_tree)\n    # Check which ast tree is matching\n    database_index = ast_check(ast_subtree_list, ast_database)\n    # We cannot index this ast in our database\n    if database_index == -1:\n        return False, True\n    # We index our reference api_call\n    ref_api_call = api_database[database_index]\n    # Check for functionality\n    if ref_api_call[\"domain\"] == qa_pairs[response[\"question_id\"] - 1][\"domain\"]:\n        return True, False\n    else:\n        return False, False\n\n\ndef main(args):\n    # Read datasets\n    api_database, qa_pairs, llm_responses, ast_database = parse_dataset(args)\n\n    # Check correctness\n    total_correct = 0\n    total_hallucination = 0\n    num_responses = len(llm_responses)\n\n    with concurrent.futures.ThreadPoolExecutor() as executor:\n        results = [\n            executor.submit(\n                process_response,\n                response,\n                api_database,\n                qa_pairs,\n                ast_database,\n            )","source_hash":"2ef51deb15484e775819fcaadf0d4372048d1371641d7f1025bed961f7a664e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.main","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_th.main#L170-L199","kind":"function","name":"main","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","language":"python","start_line":170,"end_line":199,"context_start_line":150,"context_end_line":213,"code":"        api_call = output[start + 2 : end + 1]\n\n    # Parse the api_call into AST tree\n    ast_tree = ast_parse(api_call)\n    # Search for a subtree\n    ast_subtree_list = get_all_sub_trees(ast_tree)\n    # Check which ast tree is matching\n    database_index = ast_check(ast_subtree_list, ast_database)\n    # We cannot index this ast in our database\n    if database_index == -1:\n        return False, True\n    # We index our reference api_call\n    ref_api_call = api_database[database_index]\n    # Check for functionality\n    if ref_api_call[\"domain\"] == qa_pairs[response[\"question_id\"] - 1][\"domain\"]:\n        return True, False\n    else:\n        return False, False\n\n\ndef main(args):\n    # Read datasets\n    api_database, qa_pairs, llm_responses, ast_database = parse_dataset(args)\n\n    # Check correctness\n    total_correct = 0\n    total_hallucination = 0\n    num_responses = len(llm_responses)\n\n    with concurrent.futures.ThreadPoolExecutor() as executor:\n        results = [\n            executor.submit(\n                process_response,\n                response,\n                api_database,\n                qa_pairs,\n                ast_database,\n            )\n            for response in llm_responses\n        ]\n\n        for result in concurrent.futures.as_completed(results):\n            correct, hallucination = result.result()\n            if correct:\n                total_correct += 1\n            if hallucination:\n                total_hallucination += 1\n\n    print(\"Final Functionality accuracy:\", total_correct / num_responses)\n    print(\"Final hallucination:\", total_hallucination / num_responses)\n\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--api_dataset\", type=str, default=None, help=\"path to your api dataset\")\n    parser.add_argument(\n        \"--apibench\",\n        type=str,\n        default=None,\n        help=\"path to your apibench dataset including the question and answer pairs\",\n    )\n    parser.add_argument(\"--llm_responses\", type=str, default=None, help=\"path to the language model responses\")\n    args = parser.parse_args()\n    main(args)","source_hash":"2ef51deb15484e775819fcaadf0d4372048d1371641d7f1025bed961f7a664e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf#L1-L172","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","language":"python","start_line":1,"end_line":172,"context_start_line":1,"context_end_line":172,"code":"import argparse\nimport json \nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\n# Get all the subtrees given a root_node\ndef get_all_sub_trees(root_node):\n    node_stack = []\n    sub_tree_sexp_list = []\n    depth = 1\n    text = root_node.text\n    node_stack.append([root_node, depth])\n    while len(node_stack) != 0:\n        cur_node, cur_depth = node_stack.pop()\n        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text])\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang='python'):\n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    \n    candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n    return candidate_tree\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if \"=\" in child.text.decode():\n            args_list.append(child.children[2].text)\n        elif child.text.decode() != \"(\" and child.text.decode() != \")\" and child.text.decode() != \",\":\n            args_list.append(child.text)\n    return args_list\n\n# Check if there is an api match \ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break\n        if ast_match:\n            return idx\n    return -1\n\n# Parse the dataset\ndef parse_dataset(args):\n    # Read the api datasest\n    api_database = []\n    with open(args.api_dataset, 'r') as f:\n        for line in f:\n            api_database.append(json.loads(line))\n\n    # Read the question answer pair datasest\n    qa_pairs = []\n    with open(args.apibench, 'r') as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n    \n    # Read the language model response datasest\n    llm_responses = []\n    with open(args.llm_responses, 'r') as f:\n        for line in f:\n            llm_responses.append(json.loads(line))\n\n    # Parse all apis to ast trees\n    ast_database = []\n    for data in api_database:\n        ast_tree = ast_parse(data['api_call'])\n        ast_database.append(ast_tree)\n\n    return api_database, qa_pairs, llm_responses, ast_database\n\ndef main(args):\n    # Read datsets\n    api_database, qa_pairs, llm_responses, ast_database = parse_dataset(args)\n\n    # Check correctness\n    total_correct = 0\n    total_hallucination = 0\n    for idx, response in enumerate(llm_responses):\n        try:\n            output = response['text']\n        except:\n            print('Error: cannot parse line ', idx)\n            continue\n\n        # Index the \"api_call\" domain\n        output = output.split(\"api_call\")\n        if len(output) == 1:\n            # print('Error: line ', idx, ' is not the right format')\n            # continue\n            api_call = output[0]\n        else:\n            # Parse the output\n            output = output[1].split(\"api_provider\")[0]\n            if \":\" not in output:\n                start = 0\n            else:\n                start = output.index(\":\")\n            if \")\" not in output:\n                end = -2\n            else:\n                end = output.rindex(\")\")\n            api_call = output[start+2:end+1]\n\n\n        # Parse the api_call into AST tree\n        ast_tree = ast_parse(api_call)\n        # Search for a subtree\n        ast_subtree_list = get_all_sub_trees(ast_tree)\n        # Check which ast tree is matching\n        database_index = ast_check(ast_subtree_list, ast_database)\n        # We cannot index this ast in our database\n        if database_index == -1: \n            total_hallucination += 1\n            continue\n        # We index our reference api_call\n        ref_api_call = api_database[database_index]\n        # Check for functionality\n        if ref_api_call['domain'] == qa_pairs[response['question_id'] - 1]['domain']:\n            total_correct += 1\n        else:\n            pass\n\n    print('Final Functionality accuracy: ', total_correct / len(llm_responses))\n    print('Final hallucination: ', total_hallucination/len(llm_responses))\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--api_dataset\", type=str, default=None, help=\"path to your api dataset\")\n    parser.add_argument(\"--apibench\", type=str, default=None, help=\"path to your apibench dataset including the question and answer pairs\")\n    parser.add_argument(\"--llm_responses\", type=str, default=None, help=\"path to the language model responses\")\n    args = parser.parse_args()\n    main(args)","source_hash":"b8f8c1fd383b1a114370357504ffacbb0af6dfec89b4435f7bea9fcdee0775ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.get_all_sub_trees","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.get_all_sub_trees#L21-L37","kind":"function","name":"get_all_sub_trees","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","language":"python","start_line":21,"end_line":37,"context_start_line":1,"context_end_line":57,"code":"import argparse\nimport json \nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\n# Get all the subtrees given a root_node\ndef get_all_sub_trees(root_node):\n    node_stack = []\n    sub_tree_sexp_list = []\n    depth = 1\n    text = root_node.text\n    node_stack.append([root_node, depth])\n    while len(node_stack) != 0:\n        cur_node, cur_depth = node_stack.pop()\n        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text])\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang='python'):\n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    \n    candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n    return candidate_tree\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if \"=\" in child.text.decode():\n            args_list.append(child.children[2].text)\n        elif child.text.decode() != \"(\" and child.text.decode() != \")\" and child.text.decode() != \",\":\n            args_list.append(child.text)","source_hash":"b8f8c1fd383b1a114370357504ffacbb0af6dfec89b4435f7bea9fcdee0775ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.ast_parse","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.ast_parse#L40-L46","kind":"function","name":"ast_parse","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","language":"python","start_line":40,"end_line":46,"context_start_line":20,"context_end_line":66,"code":"# Get all the subtrees given a root_node\ndef get_all_sub_trees(root_node):\n    node_stack = []\n    sub_tree_sexp_list = []\n    depth = 1\n    text = root_node.text\n    node_stack.append([root_node, depth])\n    while len(node_stack) != 0:\n        cur_node, cur_depth = node_stack.pop()\n        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text])\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang='python'):\n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    \n    candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n    return candidate_tree\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if \"=\" in child.text.decode():\n            args_list.append(child.children[2].text)\n        elif child.text.decode() != \"(\" and child.text.decode() != \")\" and child.text.decode() != \",\":\n            args_list.append(child.text)\n    return args_list\n\n# Check if there is an api match \ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:","source_hash":"b8f8c1fd383b1a114370357504ffacbb0af6dfec89b4435f7bea9fcdee0775ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.get_args","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.get_args#L49-L58","kind":"function","name":"get_args","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","language":"python","start_line":49,"end_line":58,"context_start_line":29,"context_end_line":78,"code":"        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text])\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang='python'):\n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    \n    candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n    return candidate_tree\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if \"=\" in child.text.decode():\n            args_list.append(child.children[2].text)\n        elif child.text.decode() != \"(\" and child.text.decode() != \")\" and child.text.decode() != \",\":\n            args_list.append(child.text)\n    return args_list\n\n# Check if there is an api match \ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break","source_hash":"b8f8c1fd383b1a114370357504ffacbb0af6dfec89b4435f7bea9fcdee0775ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.ast_check","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.ast_check#L61-L81","kind":"function","name":"ast_check","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","language":"python","start_line":61,"end_line":81,"context_start_line":41,"context_end_line":101,"code":"    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    \n    candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n    return candidate_tree\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if \"=\" in child.text.decode():\n            args_list.append(child.children[2].text)\n        elif child.text.decode() != \"(\" and child.text.decode() != \")\" and child.text.decode() != \",\":\n            args_list.append(child.text)\n    return args_list\n\n# Check if there is an api match \ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break\n        if ast_match:\n            return idx\n    return -1\n\n# Parse the dataset\ndef parse_dataset(args):\n    # Read the api datasest\n    api_database = []\n    with open(args.api_dataset, 'r') as f:\n        for line in f:\n            api_database.append(json.loads(line))\n\n    # Read the question answer pair datasest\n    qa_pairs = []\n    with open(args.apibench, 'r') as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n    \n    # Read the language model response datasest\n    llm_responses = []\n    with open(args.llm_responses, 'r') as f:\n        for line in f:\n            llm_responses.append(json.loads(line))","source_hash":"b8f8c1fd383b1a114370357504ffacbb0af6dfec89b4435f7bea9fcdee0775ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.parse_dataset","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.parse_dataset#L84-L109","kind":"function","name":"parse_dataset","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","language":"python","start_line":84,"end_line":109,"context_start_line":64,"context_end_line":129,"code":"            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break\n        if ast_match:\n            return idx\n    return -1\n\n# Parse the dataset\ndef parse_dataset(args):\n    # Read the api datasest\n    api_database = []\n    with open(args.api_dataset, 'r') as f:\n        for line in f:\n            api_database.append(json.loads(line))\n\n    # Read the question answer pair datasest\n    qa_pairs = []\n    with open(args.apibench, 'r') as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n    \n    # Read the language model response datasest\n    llm_responses = []\n    with open(args.llm_responses, 'r') as f:\n        for line in f:\n            llm_responses.append(json.loads(line))\n\n    # Parse all apis to ast trees\n    ast_database = []\n    for data in api_database:\n        ast_tree = ast_parse(data['api_call'])\n        ast_database.append(ast_tree)\n\n    return api_database, qa_pairs, llm_responses, ast_database\n\ndef main(args):\n    # Read datsets\n    api_database, qa_pairs, llm_responses, ast_database = parse_dataset(args)\n\n    # Check correctness\n    total_correct = 0\n    total_hallucination = 0\n    for idx, response in enumerate(llm_responses):\n        try:\n            output = response['text']\n        except:\n            print('Error: cannot parse line ', idx)\n            continue\n\n        # Index the \"api_call\" domain\n        output = output.split(\"api_call\")\n        if len(output) == 1:\n            # print('Error: line ', idx, ' is not the right format')\n            # continue","source_hash":"b8f8c1fd383b1a114370357504ffacbb0af6dfec89b4435f7bea9fcdee0775ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.main","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_hf.main#L111-L164","kind":"function","name":"main","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","language":"python","start_line":111,"end_line":164,"context_start_line":91,"context_end_line":172,"code":"    # Read the question answer pair datasest\n    qa_pairs = []\n    with open(args.apibench, 'r') as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n    \n    # Read the language model response datasest\n    llm_responses = []\n    with open(args.llm_responses, 'r') as f:\n        for line in f:\n            llm_responses.append(json.loads(line))\n\n    # Parse all apis to ast trees\n    ast_database = []\n    for data in api_database:\n        ast_tree = ast_parse(data['api_call'])\n        ast_database.append(ast_tree)\n\n    return api_database, qa_pairs, llm_responses, ast_database\n\ndef main(args):\n    # Read datsets\n    api_database, qa_pairs, llm_responses, ast_database = parse_dataset(args)\n\n    # Check correctness\n    total_correct = 0\n    total_hallucination = 0\n    for idx, response in enumerate(llm_responses):\n        try:\n            output = response['text']\n        except:\n            print('Error: cannot parse line ', idx)\n            continue\n\n        # Index the \"api_call\" domain\n        output = output.split(\"api_call\")\n        if len(output) == 1:\n            # print('Error: line ', idx, ' is not the right format')\n            # continue\n            api_call = output[0]\n        else:\n            # Parse the output\n            output = output[1].split(\"api_provider\")[0]\n            if \":\" not in output:\n                start = 0\n            else:\n                start = output.index(\":\")\n            if \")\" not in output:\n                end = -2\n            else:\n                end = output.rindex(\")\")\n            api_call = output[start+2:end+1]\n\n\n        # Parse the api_call into AST tree\n        ast_tree = ast_parse(api_call)\n        # Search for a subtree\n        ast_subtree_list = get_all_sub_trees(ast_tree)\n        # Check which ast tree is matching\n        database_index = ast_check(ast_subtree_list, ast_database)\n        # We cannot index this ast in our database\n        if database_index == -1: \n            total_hallucination += 1\n            continue\n        # We index our reference api_call\n        ref_api_call = api_database[database_index]\n        # Check for functionality\n        if ref_api_call['domain'] == qa_pairs[response['question_id'] - 1]['domain']:\n            total_correct += 1\n        else:\n            pass\n\n    print('Final Functionality accuracy: ', total_correct / len(llm_responses))\n    print('Final hallucination: ', total_hallucination/len(llm_responses))\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--api_dataset\", type=str, default=None, help=\"path to your api dataset\")\n    parser.add_argument(\"--apibench\", type=str, default=None, help=\"path to your apibench dataset including the question and answer pairs\")\n    parser.add_argument(\"--llm_responses\", type=str, default=None, help=\"path to the language model responses\")\n    args = parser.parse_args()\n    main(args)","source_hash":"b8f8c1fd383b1a114370357504ffacbb0af6dfec89b4435f7bea9fcdee0775ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf#L1-L172","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","language":"python","start_line":1,"end_line":172,"context_start_line":1,"context_end_line":172,"code":"import argparse\nimport json \nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\n# Get all the subtrees given a root_node\ndef get_all_sub_trees(root_node):\n    node_stack = []\n    sub_tree_sexp_list = []\n    depth = 1\n    text = root_node.text\n    node_stack.append([root_node, depth])\n    while len(node_stack) != 0:\n        cur_node, cur_depth = node_stack.pop()\n        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text])\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang='python'):\n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    \n    candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n    return candidate_tree\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if 'model=' in child.text.decode() or 'model =' in child.text.decode():\n            args_list.append(child.children[2].text)\n        elif child.text.decode() != \"(\" and child.text.decode() != \")\" and child.text.decode() != \",\":\n            args_list.append(child.text)\n    return args_list\n\n# Check if there is an api match \ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break\n        if ast_match:\n            return idx\n    return -1\n\n# Parse the dataset\ndef parse_dataset(args):\n    # Read the api datasest\n    api_database = []\n    with open(args.api_dataset, 'r') as f:\n        for line in f:\n            api_database.append(json.loads(line))\n\n    # Read the question answer pair datasest\n    qa_pairs = []\n    with open(args.apibench, 'r') as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n    \n    # Read the language model response datasest\n    llm_responses = []\n    with open(args.llm_responses, 'r') as f:\n        for line in f:\n            llm_responses.append(json.loads(line))\n\n    # Parse all apis to ast trees\n    ast_database = []\n    for data in api_database:\n        ast_tree = ast_parse(data['api_call'])\n        ast_database.append(ast_tree)\n\n    return api_database, qa_pairs, llm_responses, ast_database\n\ndef main(args):\n    # Read datsets\n    api_database, qa_pairs, llm_responses, ast_database = parse_dataset(args)\n\n    # Check correctness\n    total_correct = 0\n    total_hallucination = 0\n    for idx, response in enumerate(llm_responses):\n        try:\n            output = response['text']\n        except:\n            print('Error: cannot parse line ', idx)\n            continue\n\n        # Index the \"api_call\" domain\n        output = output.split(\"api_call\")\n        if len(output) == 1:\n            # print('Error: line ', idx, ' is not the right format')\n            # continue\n            api_call = output[0]\n        else:\n            # Parse the output\n            output = output[1].split(\"api_provider\")[0]\n            if \":\" not in output:\n                start = 0\n            else:\n                start = output.index(\":\")\n            if \")\" not in output:\n                end = -2\n            else:\n                end = output.rindex(\")\")\n            api_call = output[start+2:end+1]\n\n\n        # Parse the api_call into AST tree\n        ast_tree = ast_parse(api_call)\n        # Search for a subtree\n        ast_subtree_list = get_all_sub_trees(ast_tree)\n        # Check which ast tree is matching\n        database_index = ast_check(ast_subtree_list, ast_database)\n        # We cannot index this ast in our database\n        if database_index == -1: \n            total_hallucination += 1\n            continue\n        # We index our reference api_call\n        ref_api_call = api_database[database_index]\n        # Check for functionality\n        if ref_api_call['domain'] == qa_pairs[response['question_id'] - 1]['domain']:\n            total_correct += 1\n        else:\n            pass\n\n    print('Final Functionality accuracy: ', total_correct / len(llm_responses))\n    print('Final hallucination: ', total_hallucination/len(llm_responses))\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--api_dataset\", type=str, default=None, help=\"path to your api dataset\")\n    parser.add_argument(\"--apibench\", type=str, default=None, help=\"path to your apibench dataset including the question and answer pairs\")\n    parser.add_argument(\"--llm_responses\", type=str, default=None, help=\"path to the language model responses\")\n    args = parser.parse_args()\n    main(args)","source_hash":"fc0147c13cab79eb5d58dd85c2f4bdc7ab72c92c4613339a42993caaf32523ed","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.get_all_sub_trees","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.get_all_sub_trees#L21-L37","kind":"function","name":"get_all_sub_trees","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","language":"python","start_line":21,"end_line":37,"context_start_line":1,"context_end_line":57,"code":"import argparse\nimport json \nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\n# Get all the subtrees given a root_node\ndef get_all_sub_trees(root_node):\n    node_stack = []\n    sub_tree_sexp_list = []\n    depth = 1\n    text = root_node.text\n    node_stack.append([root_node, depth])\n    while len(node_stack) != 0:\n        cur_node, cur_depth = node_stack.pop()\n        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text])\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang='python'):\n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    \n    candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n    return candidate_tree\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if 'model=' in child.text.decode() or 'model =' in child.text.decode():\n            args_list.append(child.children[2].text)\n        elif child.text.decode() != \"(\" and child.text.decode() != \")\" and child.text.decode() != \",\":\n            args_list.append(child.text)","source_hash":"fc0147c13cab79eb5d58dd85c2f4bdc7ab72c92c4613339a42993caaf32523ed","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.ast_parse","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.ast_parse#L40-L46","kind":"function","name":"ast_parse","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","language":"python","start_line":40,"end_line":46,"context_start_line":20,"context_end_line":66,"code":"# Get all the subtrees given a root_node\ndef get_all_sub_trees(root_node):\n    node_stack = []\n    sub_tree_sexp_list = []\n    depth = 1\n    text = root_node.text\n    node_stack.append([root_node, depth])\n    while len(node_stack) != 0:\n        cur_node, cur_depth = node_stack.pop()\n        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text])\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang='python'):\n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    \n    candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n    return candidate_tree\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if 'model=' in child.text.decode() or 'model =' in child.text.decode():\n            args_list.append(child.children[2].text)\n        elif child.text.decode() != \"(\" and child.text.decode() != \")\" and child.text.decode() != \",\":\n            args_list.append(child.text)\n    return args_list\n\n# Check if there is an api match \ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:","source_hash":"fc0147c13cab79eb5d58dd85c2f4bdc7ab72c92c4613339a42993caaf32523ed","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.get_args","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.get_args#L49-L58","kind":"function","name":"get_args","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","language":"python","start_line":49,"end_line":58,"context_start_line":29,"context_end_line":78,"code":"        if cur_node.child_count > 0:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, cur_node.children[0].text])\n        else:\n            sub_tree_sexp_list.append([cur_node.sexp(), cur_depth, cur_node, None])\n        for child_node in cur_node.children:\n            if len(child_node.children) != 0:\n                depth = cur_depth + 1\n                node_stack.append([child_node, depth])\n    return sub_tree_sexp_list\n\n# Parse the program into AST trees\ndef ast_parse(candidate, lang='python'):\n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    \n    candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n    return candidate_tree\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if 'model=' in child.text.decode() or 'model =' in child.text.decode():\n            args_list.append(child.children[2].text)\n        elif child.text.decode() != \"(\" and child.text.decode() != \")\" and child.text.decode() != \",\":\n            args_list.append(child.text)\n    return args_list\n\n# Check if there is an api match \ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break","source_hash":"fc0147c13cab79eb5d58dd85c2f4bdc7ab72c92c4613339a42993caaf32523ed","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.ast_check","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.ast_check#L61-L81","kind":"function","name":"ast_check","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","language":"python","start_line":61,"end_line":81,"context_start_line":41,"context_end_line":101,"code":"    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    \n    candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n    return candidate_tree\n\n# Get all the arguments in the ast tree\ndef get_args(node):\n    if node.child_count == 0:\n        return []\n    args_list = []\n    for child in node.children[0].children[0].children[1].children:\n        if 'model=' in child.text.decode() or 'model =' in child.text.decode():\n            args_list.append(child.children[2].text)\n        elif child.text.decode() != \"(\" and child.text.decode() != \")\" and child.text.decode() != \",\":\n            args_list.append(child.text)\n    return args_list\n\n# Check if there is an api match \ndef ast_check(candidate_subtree_list, base_tree_list):\n    for idx, base_tree in enumerate(base_tree_list):\n        if base_tree.children[0].children[0].child_count == 0:\n            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break\n        if ast_match:\n            return idx\n    return -1\n\n# Parse the dataset\ndef parse_dataset(args):\n    # Read the api datasest\n    api_database = []\n    with open(args.api_dataset, 'r') as f:\n        for line in f:\n            api_database.append(json.loads(line))\n\n    # Read the question answer pair datasest\n    qa_pairs = []\n    with open(args.apibench, 'r') as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n    \n    # Read the language model response datasest\n    llm_responses = []\n    with open(args.llm_responses, 'r') as f:\n        for line in f:\n            llm_responses.append(json.loads(line))","source_hash":"fc0147c13cab79eb5d58dd85c2f4bdc7ab72c92c4613339a42993caaf32523ed","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.parse_dataset","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.parse_dataset#L84-L109","kind":"function","name":"parse_dataset","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","language":"python","start_line":84,"end_line":109,"context_start_line":64,"context_end_line":129,"code":"            continue\n        api_name = base_tree.children[0].children[0].children[0].text\n        for candidate_tree in candidate_subtree_list:\n            if candidate_tree[3] == api_name:\n                break\n        # Now we have a sub-tree\n        candidate_tree = candidate_tree[2]\n        args_list = get_args(base_tree)\n        if len(args_list) == 0:\n            continue\n        ast_match = True\n        for arg in args_list:\n            if arg.decode().lstrip(\"'\").rstrip(\"'\") not in candidate_tree.text.decode():\n                ast_match = False\n                break\n        if ast_match:\n            return idx\n    return -1\n\n# Parse the dataset\ndef parse_dataset(args):\n    # Read the api datasest\n    api_database = []\n    with open(args.api_dataset, 'r') as f:\n        for line in f:\n            api_database.append(json.loads(line))\n\n    # Read the question answer pair datasest\n    qa_pairs = []\n    with open(args.apibench, 'r') as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n    \n    # Read the language model response datasest\n    llm_responses = []\n    with open(args.llm_responses, 'r') as f:\n        for line in f:\n            llm_responses.append(json.loads(line))\n\n    # Parse all apis to ast trees\n    ast_database = []\n    for data in api_database:\n        ast_tree = ast_parse(data['api_call'])\n        ast_database.append(ast_tree)\n\n    return api_database, qa_pairs, llm_responses, ast_database\n\ndef main(args):\n    # Read datsets\n    api_database, qa_pairs, llm_responses, ast_database = parse_dataset(args)\n\n    # Check correctness\n    total_correct = 0\n    total_hallucination = 0\n    for idx, response in enumerate(llm_responses):\n        try:\n            output = response['text']\n        except:\n            print('Error: cannot parse line ', idx)\n            continue\n\n        # Index the \"api_call\" domain\n        output = output.split(\"api_call\")\n        if len(output) == 1:\n            # print('Error: line ', idx, ' is not the right format')\n            # continue","source_hash":"fc0147c13cab79eb5d58dd85c2f4bdc7ab72c92c4613339a42993caaf32523ed","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.main","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.ast_eval_tf.main#L111-L164","kind":"function","name":"main","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","language":"python","start_line":111,"end_line":164,"context_start_line":91,"context_end_line":172,"code":"    # Read the question answer pair datasest\n    qa_pairs = []\n    with open(args.apibench, 'r') as f:\n        for line in f:\n            qa_pairs.append(json.loads(line)[\"api_data\"])\n    \n    # Read the language model response datasest\n    llm_responses = []\n    with open(args.llm_responses, 'r') as f:\n        for line in f:\n            llm_responses.append(json.loads(line))\n\n    # Parse all apis to ast trees\n    ast_database = []\n    for data in api_database:\n        ast_tree = ast_parse(data['api_call'])\n        ast_database.append(ast_tree)\n\n    return api_database, qa_pairs, llm_responses, ast_database\n\ndef main(args):\n    # Read datsets\n    api_database, qa_pairs, llm_responses, ast_database = parse_dataset(args)\n\n    # Check correctness\n    total_correct = 0\n    total_hallucination = 0\n    for idx, response in enumerate(llm_responses):\n        try:\n            output = response['text']\n        except:\n            print('Error: cannot parse line ', idx)\n            continue\n\n        # Index the \"api_call\" domain\n        output = output.split(\"api_call\")\n        if len(output) == 1:\n            # print('Error: line ', idx, ' is not the right format')\n            # continue\n            api_call = output[0]\n        else:\n            # Parse the output\n            output = output[1].split(\"api_provider\")[0]\n            if \":\" not in output:\n                start = 0\n            else:\n                start = output.index(\":\")\n            if \")\" not in output:\n                end = -2\n            else:\n                end = output.rindex(\")\")\n            api_call = output[start+2:end+1]\n\n\n        # Parse the api_call into AST tree\n        ast_tree = ast_parse(api_call)\n        # Search for a subtree\n        ast_subtree_list = get_all_sub_trees(ast_tree)\n        # Check which ast tree is matching\n        database_index = ast_check(ast_subtree_list, ast_database)\n        # We cannot index this ast in our database\n        if database_index == -1: \n            total_hallucination += 1\n            continue\n        # We index our reference api_call\n        ref_api_call = api_database[database_index]\n        # Check for functionality\n        if ref_api_call['domain'] == qa_pairs[response['question_id'] - 1]['domain']:\n            total_correct += 1\n        else:\n            pass\n\n    print('Final Functionality accuracy: ', total_correct / len(llm_responses))\n    print('Final hallucination: ', total_hallucination/len(llm_responses))\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--api_dataset\", type=str, default=None, help=\"path to your api dataset\")\n    parser.add_argument(\"--apibench\", type=str, default=None, help=\"path to your apibench dataset including the question and answer pairs\")\n    parser.add_argument(\"--llm_responses\", type=str, default=None, help=\"path to the language model responses\")\n    args = parser.parse_args()\n    main(args)","source_hash":"fc0147c13cab79eb5d58dd85c2f4bdc7ab72c92c4613339a42993caaf32523ed","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_check","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_check#L1-L76","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_check","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_check.py","language":"python","start_line":1,"end_line":76,"context_start_line":1,"context_end_line":76,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\ndef calc_syntax_match(references, candidate, lang):\n    return corpus_syntax_match([references], [candidate], lang)\n\ndef corpus_syntax_check(references, candidates, lang):   \n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    match_count = 0\n    total_count = 0\n    scores = []\n\n    for i in range(len(candidates)):\n        references_sample = references[i]\n        candidate = candidates[i] \n        for reference in references_sample:\n            try:\n                candidate=remove_comments_and_docstrings(candidate,'java')\n            except:\n                pass    \n            try:\n                reference=remove_comments_and_docstrings(reference,'java')\n            except:\n                pass  \n\n            candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n\n            reference_tree = parser.parse(bytes(reference,'utf8')).root_node\n\n            def get_all_sub_trees(root_node):\n                node_stack = []\n                sub_tree_sexp_list = []\n                depth = 1\n                node_stack.append([root_node, depth])\n                while len(node_stack) != 0:\n                    cur_node, cur_depth = node_stack.pop()\n                    sub_tree_sexp_list.append([cur_node.sexp(), cur_depth])\n                    for child_node in cur_node.children:\n                        if len(child_node.children) != 0:\n                            depth = cur_depth + 1\n                            node_stack.append([child_node, depth])\n                return sub_tree_sexp_list\n            cand_sexps = [x[0] for x in get_all_sub_trees(candidate_tree)]\n            # print(len(cand_sexps))\n            # for i, sub_tree in enumerate(cand_sexps):\n            #     print(i, sub_tree)\n            # exit()\n            ref_sexps = get_all_sub_trees(reference_tree)\n\n            score = 0 - str(candidate_tree.sexp()).count(\"ERROR\")\n            # for sub_tree, depth in ref_sexps:\n            #     if sub_tree in cand_sexps:\n            #          match_count += 1\n            # total_count += len(ref_sexps)          \n       \n    # score = match_count / total_count\n    return score","source_hash":"411e527290a836e47f9a7956d71dab64588928d88b05466da1cb880e8e917fea","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_check.calc_syntax_match","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_check.calc_syntax_match#L21-L22","kind":"function","name":"calc_syntax_match","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_check.py","language":"python","start_line":21,"end_line":22,"context_start_line":1,"context_end_line":42,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\ndef calc_syntax_match(references, candidate, lang):\n    return corpus_syntax_match([references], [candidate], lang)\n\ndef corpus_syntax_check(references, candidates, lang):   \n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    match_count = 0\n    total_count = 0\n    scores = []\n\n    for i in range(len(candidates)):\n        references_sample = references[i]\n        candidate = candidates[i] \n        for reference in references_sample:\n            try:\n                candidate=remove_comments_and_docstrings(candidate,'java')\n            except:\n                pass    \n            try:\n                reference=remove_comments_and_docstrings(reference,'java')\n            except:","source_hash":"411e527290a836e47f9a7956d71dab64588928d88b05466da1cb880e8e917fea","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_check.corpus_syntax_check","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_check.corpus_syntax_check#L24-L76","kind":"function","name":"corpus_syntax_check","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_check.py","language":"python","start_line":24,"end_line":76,"context_start_line":4,"context_end_line":76,"code":"from codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\ndef calc_syntax_match(references, candidate, lang):\n    return corpus_syntax_match([references], [candidate], lang)\n\ndef corpus_syntax_check(references, candidates, lang):   \n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    match_count = 0\n    total_count = 0\n    scores = []\n\n    for i in range(len(candidates)):\n        references_sample = references[i]\n        candidate = candidates[i] \n        for reference in references_sample:\n            try:\n                candidate=remove_comments_and_docstrings(candidate,'java')\n            except:\n                pass    \n            try:\n                reference=remove_comments_and_docstrings(reference,'java')\n            except:\n                pass  \n\n            candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n\n            reference_tree = parser.parse(bytes(reference,'utf8')).root_node\n\n            def get_all_sub_trees(root_node):\n                node_stack = []\n                sub_tree_sexp_list = []\n                depth = 1\n                node_stack.append([root_node, depth])\n                while len(node_stack) != 0:\n                    cur_node, cur_depth = node_stack.pop()\n                    sub_tree_sexp_list.append([cur_node.sexp(), cur_depth])\n                    for child_node in cur_node.children:\n                        if len(child_node.children) != 0:\n                            depth = cur_depth + 1\n                            node_stack.append([child_node, depth])\n                return sub_tree_sexp_list\n            cand_sexps = [x[0] for x in get_all_sub_trees(candidate_tree)]\n            # print(len(cand_sexps))\n            # for i, sub_tree in enumerate(cand_sexps):\n            #     print(i, sub_tree)\n            # exit()\n            ref_sexps = get_all_sub_trees(reference_tree)\n\n            score = 0 - str(candidate_tree.sexp()).count(\"ERROR\")\n            # for sub_tree, depth in ref_sexps:\n            #     if sub_tree in cand_sexps:\n            #          match_count += 1\n            # total_count += len(ref_sexps)          \n       \n    # score = match_count / total_count\n    return score","source_hash":"411e527290a836e47f9a7956d71dab64588928d88b05466da1cb880e8e917fea","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_check.get_all_sub_trees","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_check.get_all_sub_trees#L49-L61","kind":"function","name":"get_all_sub_trees","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_check.py","language":"python","start_line":49,"end_line":61,"context_start_line":29,"context_end_line":76,"code":"    total_count = 0\n    scores = []\n\n    for i in range(len(candidates)):\n        references_sample = references[i]\n        candidate = candidates[i] \n        for reference in references_sample:\n            try:\n                candidate=remove_comments_and_docstrings(candidate,'java')\n            except:\n                pass    \n            try:\n                reference=remove_comments_and_docstrings(reference,'java')\n            except:\n                pass  \n\n            candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n\n            reference_tree = parser.parse(bytes(reference,'utf8')).root_node\n\n            def get_all_sub_trees(root_node):\n                node_stack = []\n                sub_tree_sexp_list = []\n                depth = 1\n                node_stack.append([root_node, depth])\n                while len(node_stack) != 0:\n                    cur_node, cur_depth = node_stack.pop()\n                    sub_tree_sexp_list.append([cur_node.sexp(), cur_depth])\n                    for child_node in cur_node.children:\n                        if len(child_node.children) != 0:\n                            depth = cur_depth + 1\n                            node_stack.append([child_node, depth])\n                return sub_tree_sexp_list\n            cand_sexps = [x[0] for x in get_all_sub_trees(candidate_tree)]\n            # print(len(cand_sexps))\n            # for i, sub_tree in enumerate(cand_sexps):\n            #     print(i, sub_tree)\n            # exit()\n            ref_sexps = get_all_sub_trees(reference_tree)\n\n            score = 0 - str(candidate_tree.sexp()).count(\"ERROR\")\n            # for sub_tree, depth in ref_sexps:\n            #     if sub_tree in cand_sexps:\n            #          match_count += 1\n            # total_count += len(ref_sexps)          \n       \n    # score = match_count / total_count\n    return score","source_hash":"411e527290a836e47f9a7956d71dab64588928d88b05466da1cb880e8e917fea","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match#L1-L558","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":1,"end_line":558,"context_start_line":1,"context_end_line":558,"code":"# -*- coding: utf-8 -*-\n# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\n# Natural Language Toolkit: BLEU Score\n#\n# Copyright (C) 2001-2020 NLTK Project\n# Authors: Chin Yee Lee, Hengfeng Li, Ruxin Hou, Calvin Tanujaya Lim\n# Contributors: Björn Mattsson, Dmitrijs Milajevs, Liling Tan\n# URL: <http://nltk.org/>\n# For license information, see LICENSE.TXT\n\n\"\"\"BLEU score implementation.\"\"\"\n\nimport math\nimport sys\nfrom fractions import Fraction\nimport warnings\nfrom collections import Counter\n\nfrom codebleu.utils import ngrams\nimport pdb\n\n\ndef sentence_bleu(\n    references,\n    hypothesis,\n    weights=(0.25, 0.25, 0.25, 0.25),\n    smoothing_function=None,\n    auto_reweigh=False,\n):\n    \"\"\"\n    Calculate BLEU score (Bilingual Evaluation Understudy) from\n    Papineni, Kishore, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002.\n    \"BLEU: a method for automatic evaluation of machine translation.\"\n    In Proceedings of ACL. http://www.aclweb.org/anthology/P02-1040.pdf\n    >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n    ...               'ensures', 'that', 'the', 'military', 'always',\n    ...               'obeys', 'the', 'commands', 'of', 'the', 'party']\n    >>> hypothesis2 = ['It', 'is', 'to', 'insure', 'the', 'troops',\n    ...               'forever', 'hearing', 'the', 'activity', 'guidebook',\n    ...               'that', 'party', 'direct']\n    >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n    ...               'ensures', 'that', 'the', 'military', 'will', 'forever',\n    ...               'heed', 'Party', 'commands']\n    >>> reference2 = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n    ...               'guarantees', 'the', 'military', 'forces', 'always',\n    ...               'being', 'under', 'the', 'command', 'of', 'the',\n    ...               'Party']\n    >>> reference3 = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n    ...               'army', 'always', 'to', 'heed', 'the', 'directions',\n    ...               'of', 'the', 'party']\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis1) # doctest: +ELLIPSIS\n    0.5045...\n    If there is no ngrams overlap for any order of n-grams, BLEU returns the\n    value 0. This is because the precision for the order of n-grams without\n    overlap is 0, and the geometric mean in the final BLEU score computation\n    multiplies the 0 with the precision of other n-grams. This results in 0\n    (independently of the precision of the othe n-gram orders). The following\n    example has zero 3-gram and 4-gram overlaps:\n    >>> round(sentence_bleu([reference1, reference2, reference3], hypothesis2),4) # doctest: +ELLIPSIS\n    0.0\n    To avoid this harsh behaviour when no ngram overlaps are found a smoothing\n    function can be used.\n    >>> chencherry = SmoothingFunction()\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis2,\n    ...     smoothing_function=chencherry.method1) # doctest: +ELLIPSIS\n    0.0370...\n    The default BLEU calculates a score for up to 4-grams using uniform\n    weights (this is called BLEU-4). To evaluate your translations with\n    higher/lower order ngrams, use customized weights. E.g. when accounting\n    for up to 5-grams with uniform weights (this is called BLEU-5) use:\n    >>> weights = (1./5., 1./5., 1./5., 1./5., 1./5.)\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis1, weights) # doctest: +ELLIPSIS\n    0.3920...\n    :param references: reference sentences\n    :type references: list(list(str))\n    :param hypothesis: a hypothesis sentence\n    :type hypothesis: list(str)\n    :param weights: weights for unigrams, bigrams, trigrams and so on\n    :type weights: list(float)\n    :param smoothing_function:\n    :type smoothing_function: SmoothingFunction\n    :param auto_reweigh: Option to re-normalize the weights uniformly.\n    :type auto_reweigh: bool\n    :return: The sentence-level BLEU score.\n    :rtype: float\n    \"\"\"\n    return corpus_bleu(\n        [references], [hypothesis], weights, smoothing_function, auto_reweigh\n    )\n\n\ndef corpus_bleu(\n    list_of_references,\n    hypotheses,\n    weights=(0.25, 0.25, 0.25, 0.25),\n    smoothing_function=None,\n    auto_reweigh=False,\n):\n    \"\"\"\n    Calculate a single corpus-level BLEU score (aka. system-level BLEU) for all\n    the hypotheses and their respective references.\n    Instead of averaging the sentence level BLEU scores (i.e. marco-average\n    precision), the original BLEU metric (Papineni et al. 2002) accounts for\n    the micro-average precision (i.e. summing the numerators and denominators\n    for each hypothesis-reference(s) pairs before the division).\n    >>> hyp1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n    ...         'ensures', 'that', 'the', 'military', 'always',\n    ...         'obeys', 'the', 'commands', 'of', 'the', 'party']\n    >>> ref1a = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n    ...          'ensures', 'that', 'the', 'military', 'will', 'forever',\n    ...          'heed', 'Party', 'commands']\n    >>> ref1b = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n    ...          'guarantees', 'the', 'military', 'forces', 'always',\n    ...          'being', 'under', 'the', 'command', 'of', 'the', 'Party']\n    >>> ref1c = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n    ...          'army', 'always', 'to', 'heed', 'the', 'directions',\n    ...          'of', 'the', 'party']\n    >>> hyp2 = ['he', 'read', 'the', 'book', 'because', 'he', 'was',\n    ...         'interested', 'in', 'world', 'history']\n    >>> ref2a = ['he', 'was', 'interested', 'in', 'world', 'history',\n    ...          'because', 'he', 'read', 'the', 'book']\n    >>> list_of_references = [[ref1a, ref1b, ref1c], [ref2a]]\n    >>> hypotheses = [hyp1, hyp2]\n    >>> corpus_bleu(list_of_references, hypotheses) # doctest: +ELLIPSIS\n    0.5920...\n    The example below show that corpus_bleu() is different from averaging\n    sentence_bleu() for hypotheses\n    >>> score1 = sentence_bleu([ref1a, ref1b, ref1c], hyp1)\n    >>> score2 = sentence_bleu([ref2a], hyp2)\n    >>> (score1 + score2) / 2 # doctest: +ELLIPSIS\n    0.6223...\n    :param list_of_references: a corpus of lists of reference sentences, w.r.t. hypotheses\n    :type list_of_references: list(list(list(str)))\n    :param hypotheses: a list of hypothesis sentences\n    :type hypotheses: list(list(str))\n    :param weights: weights for unigrams, bigrams, trigrams and so on\n    :type weights: list(float)\n    :param smoothing_function:\n    :type smoothing_function: SmoothingFunction\n    :param auto_reweigh: Option to re-normalize the weights uniformly.\n    :type auto_reweigh: bool\n    :return: The corpus-level BLEU score.\n    :rtype: float\n    \"\"\"\n    # Before proceeding to compute BLEU, perform sanity checks.\n\n    p_numerators = Counter()  # Key = ngram order, and value = no. of ngram matches.\n    p_denominators = Counter()  # Key = ngram order, and value = no. of ngram in ref.\n    hyp_lengths, ref_lengths = 0, 0\n\n    assert len(list_of_references) == len(hypotheses), (\n        \"The number of hypotheses and their reference(s) should be the \" \"same \"\n    )\n\n    # Iterate through each hypothesis and their corresponding references.\n    for references, hypothesis in zip(list_of_references, hypotheses):\n        # For each order of ngram, calculate the numerator and\n        # denominator for the corpus-level modified precision.\n        for i, _ in enumerate(weights, start=1):\n            p_i_numeraotr, p_i_denominator = modified_recall(references, hypothesis, i)\n            p_numerators[i] += p_i_numeraotr\n            p_denominators[i] += p_i_denominator\n\n        # Calculate the hypothesis length and the closest reference length.\n        # Adds them to the corpus-level hypothesis and reference counts.\n        hyp_len = len(hypothesis)\n        hyp_lengths += hyp_len\n        ref_lengths += closest_ref_length(references, hyp_len)\n\n    # Calculate corpus-level brevity penalty.\n    bp = brevity_penalty(ref_lengths, hyp_lengths)\n\n    # Uniformly re-weighting based on maximum hypothesis lengths if largest\n    # order of n-grams < 4 and weights is set at default.\n    if auto_reweigh:\n        if hyp_lengths < 4 and weights == (0.25, 0.25, 0.25, 0.25):\n            weights = (1 / hyp_lengths,) * hyp_lengths\n\n    # Collects the various recall values for the different ngram orders.\n    p_n = [\n        (p_numerators[i], p_denominators[i])\n        for i, _ in enumerate(weights, start=1)\n    ]\n\n    # Returns 0 if there's no matching n-grams\n    # We only need to check for p_numerators[1] == 0, since if there's\n    # no unigrams, there won't be any higher order ngrams.\n    if p_numerators[1] == 0:\n        return 0\n\n    # If there's no smoothing, set use method0 from SmoothinFunction class.\n    if not smoothing_function:\n        smoothing_function = SmoothingFunction().method1\n    # Smoothen the modified precision.\n    # Note: smoothing_function() may convert values into floats;\n    #       it tries to retain the Fraction object as much as the\n    #       smoothing method allows.\n    p_n = smoothing_function(\n        p_n, references=references, hypothesis=hypothesis, hyp_len=hyp_lengths\n    )\n    # pdb.set_trace()\n    s = (w_i * math.log(p_i[0]/p_i[1]) for w_i, p_i in zip(weights, p_n))\n    s = bp * math.exp(math.fsum(s))\n    return s\n\n\ndef modified_recall(references, hypothesis, n):\n    \"\"\"\n    Calculate modified ngram recall.\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hypothesis: A hypothesis translation.\n    :type hypothesis: list(str)\n    :param n: The ngram order.\n    :type n: int\n    :return: BLEU's modified precision for the nth order ngram.\n    :rtype: Fraction\n    \"\"\"\n    # Extracts all ngrams in hypothesis\n    # Set an empty Counter if hypothesis is empty.\n    # pdb.set_trace()\n    numerator = 0\n    denominator = 0\n\n    counts = Counter(ngrams(hypothesis, n)) if len(hypothesis) >= n else Counter()\n    # Extract a union of references' counts.\n    # max_counts = reduce(or_, [Counter(ngrams(ref, n)) for ref in references])\n    max_counts = {}\n    for reference_and_weights in references:\n        reference = reference_and_weights[0]\n        weights = reference_and_weights[1]\n        reference_counts = (\n            Counter(ngrams(reference, n)) if len(reference) >= n else Counter()\n        )\n        # for ngram in reference_counts:\n        #     max_counts[ngram] = max(max_counts.get(ngram, 0), counts[ngram])\n        clipped_counts = {\n            ngram: min(count, counts[ngram]) for ngram, count in reference_counts.items()\n        }\n        # reweight\n        if n == 1 and len(weights) == len(reference_counts):\n            def weighted_sum(weights, counts):\n                sum_counts = 0\n                for ngram, count in counts.items():\n                    sum_counts += count * (weights[ngram[0]] if ngram[0] in weights else 1)\n                return sum_counts\n\n            numerator += weighted_sum(weights, clipped_counts)\n            denominator += max(1, weighted_sum(weights, reference_counts))\n\n        else:\n            numerator += sum(clipped_counts.values())\n            denominator += max(1, sum(reference_counts.values()))\n\n        # # Assigns the intersection between hypothesis and references' counts.\n        # clipped_counts = {\n        #     ngram: min(count, max_counts[ngram]) for ngram, count in counts.items()\n        # }\n\n        # numerator += sum(clipped_counts.values())\n        # # Ensures that denominator is minimum 1 to avoid ZeroDivisionError.\n        # # Usually this happens when the ngram order is > len(reference).\n        # denominator += max(1, sum(counts.values()))\n\n    #return Fraction(numerator, denominator, _normalize=False)\n    return numerator, denominator\n\n\ndef closest_ref_length(references, hyp_len):\n    \"\"\"\n    This function finds the reference that is the closest length to the\n    hypothesis. The closest reference length is referred to as *r* variable\n    from the brevity penalty formula in Papineni et. al. (2002)\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hyp_len: The length of the hypothesis.\n    :type hyp_len: int\n    :return: The length of the reference that's closest to the hypothesis.\n    :rtype: int\n    \"\"\"\n    ref_lens = (len(reference) for reference in references)\n    closest_ref_len = min(\n        ref_lens, key=lambda ref_len: (abs(ref_len - hyp_len), ref_len)\n    )\n    return closest_ref_len\n\n\ndef brevity_penalty(closest_ref_len, hyp_len):\n    \"\"\"\n    Calculate brevity penalty.\n    As the modified n-gram precision still has the problem from the short\n    length sentence, brevity penalty is used to modify the overall BLEU\n    score according to length.\n    An example from the paper. There are three references with length 12, 15\n    and 17. And a concise hypothesis of the length 12. The brevity penalty is 1.\n        >>> reference1 = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> reference2 = list('aaaaaaaaaaaaaaa')   # i.e. ['a'] * 15\n        >>> reference3 = list('aaaaaaaaaaaaaaaaa') # i.e. ['a'] * 17\n        >>> hypothesis = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> references = [reference1, reference2, reference3]\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    In case a hypothesis translation is shorter than the references, penalty is\n    applied.\n        >>> references = [['a'] * 28, ['a'] * 28]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        0.2635971381157267\n    The length of the closest reference is used to compute the penalty. If the\n    length of a hypothesis is 12, and the reference lengths are 13 and 2, the\n    penalty is applied because the hypothesis length (12) is less then the\n    closest reference length (13).\n        >>> references = [['a'] * 13, ['a'] * 2]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len) # doctest: +ELLIPSIS\n        0.9200...\n    The brevity penalty doesn't depend on reference order. More importantly,\n    when two reference sentences are at the same distance, the shortest\n    reference sentence length is used.\n        >>> references = [['a'] * 13, ['a'] * 11]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> bp1 = brevity_penalty(closest_ref_len, hyp_len)\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(reversed(references), hyp_len)\n        >>> bp2 = brevity_penalty(closest_ref_len, hyp_len)\n        >>> bp1 == bp2 == 1\n        True\n    A test example from mteval-v13a.pl (starting from the line 705):\n        >>> references = [['a'] * 11, ['a'] * 8]\n        >>> hypothesis = ['a'] * 7\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len) # doctest: +ELLIPSIS\n        0.8668...\n        >>> references = [['a'] * 11, ['a'] * 8, ['a'] * 6, ['a'] * 7]\n        >>> hypothesis = ['a'] * 7\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    :param hyp_len: The length of the hypothesis for a single sentence OR the\n    sum of all the hypotheses' lengths for a corpus\n    :type hyp_len: int\n    :param closest_ref_len: The length of the closest reference for a single\n    hypothesis OR the sum of all the closest references for every hypotheses.\n    :type closest_ref_len: int\n    :return: BLEU's brevity penalty.\n    :rtype: float\n    \"\"\"\n    if hyp_len > closest_ref_len:\n        return 1\n    # If hypothesis is empty, brevity penalty = 0 should result in BLEU = 0.0\n    elif hyp_len == 0:\n        return 0\n    else:\n        return math.exp(1 - closest_ref_len / hyp_len)\n\n\nclass SmoothingFunction:\n    \"\"\"\n    This is an implementation of the smoothing techniques\n    for segment-level BLEU scores that was presented in\n    Boxing Chen and Collin Cherry (2014) A Systematic Comparison of\n    Smoothing Techniques for Sentence-Level BLEU. In WMT14.\n    http://acl2014.org/acl2014/W14-33/pdf/W14-3346.pdf\n    \"\"\"\n\n    def __init__(self, epsilon=0.1, alpha=5, k=5):\n        \"\"\"\n        This will initialize the parameters required for the various smoothing\n        techniques, the default values are set to the numbers used in the\n        experiments from Chen and Cherry (2014).\n        >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which', 'ensures',\n        ...                 'that', 'the', 'military', 'always', 'obeys', 'the',\n        ...                 'commands', 'of', 'the', 'party']\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that', 'ensures',\n        ...               'that', 'the', 'military', 'will', 'forever', 'heed',\n        ...               'Party', 'commands']\n        >>> chencherry = SmoothingFunction()\n        >>> print(sentence_bleu([reference1], hypothesis1)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method0)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method1)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method2)) # doctest: +ELLIPSIS\n        0.4489...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method3)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method4)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method5)) # doctest: +ELLIPSIS\n        0.4905...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method6)) # doctest: +ELLIPSIS\n        0.4135...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method7)) # doctest: +ELLIPSIS\n        0.4905...\n        :param epsilon: the epsilon value use in method 1\n        :type epsilon: float\n        :param alpha: the alpha value use in method 6\n        :type alpha: int\n        :param k: the k value use in method 4\n        :type k: int\n        \"\"\"\n        self.epsilon = epsilon\n        self.alpha = alpha\n        self.k = k\n\n    def method0(self, p_n, *args, **kwargs):\n        \"\"\"\n        No smoothing.\n        \"\"\"\n        p_n_new = []\n        for i, p_i in enumerate(p_n):\n            if p_i[0] != 0:\n                p_n_new.append(p_i)\n            else:\n                _msg = str(\n                    \"\\nThe hypothesis contains 0 counts of {}-gram overlaps.\\n\"\n                    \"Therefore the BLEU score evaluates to 0, independently of\\n\"\n                    \"how many N-gram overlaps of lower order it contains.\\n\"\n                    \"Consider using lower n-gram order or use \"\n                    \"SmoothingFunction()\"\n                ).format(i + 1)\n                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.\n                # Due to BLEU geometric mean computation in logarithm space,\n                # we we need to take the return sys.float_info.min such that\n                # math.log(sys.float_i\n# ... truncated ...","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":true}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.sentence_bleu","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.sentence_bleu#L25-L91","kind":"function","name":"sentence_bleu","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":25,"end_line":91,"context_start_line":5,"context_end_line":111,"code":"# Natural Language Toolkit: BLEU Score\n#\n# Copyright (C) 2001-2020 NLTK Project\n# Authors: Chin Yee Lee, Hengfeng Li, Ruxin Hou, Calvin Tanujaya Lim\n# Contributors: Björn Mattsson, Dmitrijs Milajevs, Liling Tan\n# URL: <http://nltk.org/>\n# For license information, see LICENSE.TXT\n\n\"\"\"BLEU score implementation.\"\"\"\n\nimport math\nimport sys\nfrom fractions import Fraction\nimport warnings\nfrom collections import Counter\n\nfrom codebleu.utils import ngrams\nimport pdb\n\n\ndef sentence_bleu(\n    references,\n    hypothesis,\n    weights=(0.25, 0.25, 0.25, 0.25),\n    smoothing_function=None,\n    auto_reweigh=False,\n):\n    \"\"\"\n    Calculate BLEU score (Bilingual Evaluation Understudy) from\n    Papineni, Kishore, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002.\n    \"BLEU: a method for automatic evaluation of machine translation.\"\n    In Proceedings of ACL. http://www.aclweb.org/anthology/P02-1040.pdf\n    >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n    ...               'ensures', 'that', 'the', 'military', 'always',\n    ...               'obeys', 'the', 'commands', 'of', 'the', 'party']\n    >>> hypothesis2 = ['It', 'is', 'to', 'insure', 'the', 'troops',\n    ...               'forever', 'hearing', 'the', 'activity', 'guidebook',\n    ...               'that', 'party', 'direct']\n    >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n    ...               'ensures', 'that', 'the', 'military', 'will', 'forever',\n    ...               'heed', 'Party', 'commands']\n    >>> reference2 = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n    ...               'guarantees', 'the', 'military', 'forces', 'always',\n    ...               'being', 'under', 'the', 'command', 'of', 'the',\n    ...               'Party']\n    >>> reference3 = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n    ...               'army', 'always', 'to', 'heed', 'the', 'directions',\n    ...               'of', 'the', 'party']\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis1) # doctest: +ELLIPSIS\n    0.5045...\n    If there is no ngrams overlap for any order of n-grams, BLEU returns the\n    value 0. This is because the precision for the order of n-grams without\n    overlap is 0, and the geometric mean in the final BLEU score computation\n    multiplies the 0 with the precision of other n-grams. This results in 0\n    (independently of the precision of the othe n-gram orders). The following\n    example has zero 3-gram and 4-gram overlaps:\n    >>> round(sentence_bleu([reference1, reference2, reference3], hypothesis2),4) # doctest: +ELLIPSIS\n    0.0\n    To avoid this harsh behaviour when no ngram overlaps are found a smoothing\n    function can be used.\n    >>> chencherry = SmoothingFunction()\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis2,\n    ...     smoothing_function=chencherry.method1) # doctest: +ELLIPSIS\n    0.0370...\n    The default BLEU calculates a score for up to 4-grams using uniform\n    weights (this is called BLEU-4). To evaluate your translations with\n    higher/lower order ngrams, use customized weights. E.g. when accounting\n    for up to 5-grams with uniform weights (this is called BLEU-5) use:\n    >>> weights = (1./5., 1./5., 1./5., 1./5., 1./5.)\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis1, weights) # doctest: +ELLIPSIS\n    0.3920...\n    :param references: reference sentences\n    :type references: list(list(str))\n    :param hypothesis: a hypothesis sentence\n    :type hypothesis: list(str)\n    :param weights: weights for unigrams, bigrams, trigrams and so on\n    :type weights: list(float)\n    :param smoothing_function:\n    :type smoothing_function: SmoothingFunction\n    :param auto_reweigh: Option to re-normalize the weights uniformly.\n    :type auto_reweigh: bool\n    :return: The sentence-level BLEU score.\n    :rtype: float\n    \"\"\"\n    return corpus_bleu(\n        [references], [hypothesis], weights, smoothing_function, auto_reweigh\n    )\n\n\ndef corpus_bleu(\n    list_of_references,\n    hypotheses,\n    weights=(0.25, 0.25, 0.25, 0.25),\n    smoothing_function=None,\n    auto_reweigh=False,\n):\n    \"\"\"\n    Calculate a single corpus-level BLEU score (aka. system-level BLEU) for all\n    the hypotheses and their respective references.\n    Instead of averaging the sentence level BLEU scores (i.e. marco-average\n    precision), the original BLEU metric (Papineni et al. 2002) accounts for\n    the micro-average precision (i.e. summing the numerators and denominators\n    for each hypothesis-reference(s) pairs before the division).\n    >>> hyp1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n    ...         'ensures', 'that', 'the', 'military', 'always',\n    ...         'obeys', 'the', 'commands', 'of', 'the', 'party']\n    >>> ref1a = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.corpus_bleu","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.corpus_bleu#L94-L206","kind":"function","name":"corpus_bleu","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":94,"end_line":206,"context_start_line":74,"context_end_line":226,"code":"    >>> sentence_bleu([reference1, reference2, reference3], hypothesis1, weights) # doctest: +ELLIPSIS\n    0.3920...\n    :param references: reference sentences\n    :type references: list(list(str))\n    :param hypothesis: a hypothesis sentence\n    :type hypothesis: list(str)\n    :param weights: weights for unigrams, bigrams, trigrams and so on\n    :type weights: list(float)\n    :param smoothing_function:\n    :type smoothing_function: SmoothingFunction\n    :param auto_reweigh: Option to re-normalize the weights uniformly.\n    :type auto_reweigh: bool\n    :return: The sentence-level BLEU score.\n    :rtype: float\n    \"\"\"\n    return corpus_bleu(\n        [references], [hypothesis], weights, smoothing_function, auto_reweigh\n    )\n\n\ndef corpus_bleu(\n    list_of_references,\n    hypotheses,\n    weights=(0.25, 0.25, 0.25, 0.25),\n    smoothing_function=None,\n    auto_reweigh=False,\n):\n    \"\"\"\n    Calculate a single corpus-level BLEU score (aka. system-level BLEU) for all\n    the hypotheses and their respective references.\n    Instead of averaging the sentence level BLEU scores (i.e. marco-average\n    precision), the original BLEU metric (Papineni et al. 2002) accounts for\n    the micro-average precision (i.e. summing the numerators and denominators\n    for each hypothesis-reference(s) pairs before the division).\n    >>> hyp1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n    ...         'ensures', 'that', 'the', 'military', 'always',\n    ...         'obeys', 'the', 'commands', 'of', 'the', 'party']\n    >>> ref1a = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n    ...          'ensures', 'that', 'the', 'military', 'will', 'forever',\n    ...          'heed', 'Party', 'commands']\n    >>> ref1b = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n    ...          'guarantees', 'the', 'military', 'forces', 'always',\n    ...          'being', 'under', 'the', 'command', 'of', 'the', 'Party']\n    >>> ref1c = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n    ...          'army', 'always', 'to', 'heed', 'the', 'directions',\n    ...          'of', 'the', 'party']\n    >>> hyp2 = ['he', 'read', 'the', 'book', 'because', 'he', 'was',\n    ...         'interested', 'in', 'world', 'history']\n    >>> ref2a = ['he', 'was', 'interested', 'in', 'world', 'history',\n    ...          'because', 'he', 'read', 'the', 'book']\n    >>> list_of_references = [[ref1a, ref1b, ref1c], [ref2a]]\n    >>> hypotheses = [hyp1, hyp2]\n    >>> corpus_bleu(list_of_references, hypotheses) # doctest: +ELLIPSIS\n    0.5920...\n    The example below show that corpus_bleu() is different from averaging\n    sentence_bleu() for hypotheses\n    >>> score1 = sentence_bleu([ref1a, ref1b, ref1c], hyp1)\n    >>> score2 = sentence_bleu([ref2a], hyp2)\n    >>> (score1 + score2) / 2 # doctest: +ELLIPSIS\n    0.6223...\n    :param list_of_references: a corpus of lists of reference sentences, w.r.t. hypotheses\n    :type list_of_references: list(list(list(str)))\n    :param hypotheses: a list of hypothesis sentences\n    :type hypotheses: list(list(str))\n    :param weights: weights for unigrams, bigrams, trigrams and so on\n    :type weights: list(float)\n    :param smoothing_function:\n    :type smoothing_function: SmoothingFunction\n    :param auto_reweigh: Option to re-normalize the weights uniformly.\n    :type auto_reweigh: bool\n    :return: The corpus-level BLEU score.\n    :rtype: float\n    \"\"\"\n    # Before proceeding to compute BLEU, perform sanity checks.\n\n    p_numerators = Counter()  # Key = ngram order, and value = no. of ngram matches.\n    p_denominators = Counter()  # Key = ngram order, and value = no. of ngram in ref.\n    hyp_lengths, ref_lengths = 0, 0\n\n    assert len(list_of_references) == len(hypotheses), (\n        \"The number of hypotheses and their reference(s) should be the \" \"same \"\n    )\n\n    # Iterate through each hypothesis and their corresponding references.\n    for references, hypothesis in zip(list_of_references, hypotheses):\n        # For each order of ngram, calculate the numerator and\n        # denominator for the corpus-level modified precision.\n        for i, _ in enumerate(weights, start=1):\n            p_i_numeraotr, p_i_denominator = modified_recall(references, hypothesis, i)\n            p_numerators[i] += p_i_numeraotr\n            p_denominators[i] += p_i_denominator\n\n        # Calculate the hypothesis length and the closest reference length.\n        # Adds them to the corpus-level hypothesis and reference counts.\n        hyp_len = len(hypothesis)\n        hyp_lengths += hyp_len\n        ref_lengths += closest_ref_length(references, hyp_len)\n\n    # Calculate corpus-level brevity penalty.\n    bp = brevity_penalty(ref_lengths, hyp_lengths)\n\n    # Uniformly re-weighting based on maximum hypothesis lengths if largest\n    # order of n-grams < 4 and weights is set at default.\n    if auto_reweigh:\n        if hyp_lengths < 4 and weights == (0.25, 0.25, 0.25, 0.25):\n            weights = (1 / hyp_lengths,) * hyp_lengths\n\n    # Collects the various recall values for the different ngram orders.\n    p_n = [\n        (p_numerators[i], p_denominators[i])\n        for i, _ in enumerate(weights, start=1)\n    ]\n\n    # Returns 0 if there's no matching n-grams\n    # We only need to check for p_numerators[1] == 0, since if there's\n    # no unigrams, there won't be any higher order ngrams.\n    if p_numerators[1] == 0:\n        return 0\n\n    # If there's no smoothing, set use method0 from SmoothinFunction class.\n    if not smoothing_function:\n        smoothing_function = SmoothingFunction().method1\n    # Smoothen the modified precision.\n    # Note: smoothing_function() may convert values into floats;\n    #       it tries to retain the Fraction object as much as the\n    #       smoothing method allows.\n    p_n = smoothing_function(\n        p_n, references=references, hypothesis=hypothesis, hyp_len=hyp_lengths\n    )\n    # pdb.set_trace()\n    s = (w_i * math.log(p_i[0]/p_i[1]) for w_i, p_i in zip(weights, p_n))\n    s = bp * math.exp(math.fsum(s))\n    return s\n\n\ndef modified_recall(references, hypothesis, n):\n    \"\"\"\n    Calculate modified ngram recall.\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hypothesis: A hypothesis translation.\n    :type hypothesis: list(str)\n    :param n: The ngram order.\n    :type n: int\n    :return: BLEU's modified precision for the nth order ngram.\n    :rtype: Fraction\n    \"\"\"\n    # Extracts all ngrams in hypothesis\n    # Set an empty Counter if hypothesis is empty.\n    # pdb.set_trace()\n    numerator = 0\n    denominator = 0\n","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.modified_recall","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.modified_recall#L209-L268","kind":"function","name":"modified_recall","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":209,"end_line":268,"context_start_line":189,"context_end_line":288,"code":"    # no unigrams, there won't be any higher order ngrams.\n    if p_numerators[1] == 0:\n        return 0\n\n    # If there's no smoothing, set use method0 from SmoothinFunction class.\n    if not smoothing_function:\n        smoothing_function = SmoothingFunction().method1\n    # Smoothen the modified precision.\n    # Note: smoothing_function() may convert values into floats;\n    #       it tries to retain the Fraction object as much as the\n    #       smoothing method allows.\n    p_n = smoothing_function(\n        p_n, references=references, hypothesis=hypothesis, hyp_len=hyp_lengths\n    )\n    # pdb.set_trace()\n    s = (w_i * math.log(p_i[0]/p_i[1]) for w_i, p_i in zip(weights, p_n))\n    s = bp * math.exp(math.fsum(s))\n    return s\n\n\ndef modified_recall(references, hypothesis, n):\n    \"\"\"\n    Calculate modified ngram recall.\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hypothesis: A hypothesis translation.\n    :type hypothesis: list(str)\n    :param n: The ngram order.\n    :type n: int\n    :return: BLEU's modified precision for the nth order ngram.\n    :rtype: Fraction\n    \"\"\"\n    # Extracts all ngrams in hypothesis\n    # Set an empty Counter if hypothesis is empty.\n    # pdb.set_trace()\n    numerator = 0\n    denominator = 0\n\n    counts = Counter(ngrams(hypothesis, n)) if len(hypothesis) >= n else Counter()\n    # Extract a union of references' counts.\n    # max_counts = reduce(or_, [Counter(ngrams(ref, n)) for ref in references])\n    max_counts = {}\n    for reference_and_weights in references:\n        reference = reference_and_weights[0]\n        weights = reference_and_weights[1]\n        reference_counts = (\n            Counter(ngrams(reference, n)) if len(reference) >= n else Counter()\n        )\n        # for ngram in reference_counts:\n        #     max_counts[ngram] = max(max_counts.get(ngram, 0), counts[ngram])\n        clipped_counts = {\n            ngram: min(count, counts[ngram]) for ngram, count in reference_counts.items()\n        }\n        # reweight\n        if n == 1 and len(weights) == len(reference_counts):\n            def weighted_sum(weights, counts):\n                sum_counts = 0\n                for ngram, count in counts.items():\n                    sum_counts += count * (weights[ngram[0]] if ngram[0] in weights else 1)\n                return sum_counts\n\n            numerator += weighted_sum(weights, clipped_counts)\n            denominator += max(1, weighted_sum(weights, reference_counts))\n\n        else:\n            numerator += sum(clipped_counts.values())\n            denominator += max(1, sum(reference_counts.values()))\n\n        # # Assigns the intersection between hypothesis and references' counts.\n        # clipped_counts = {\n        #     ngram: min(count, max_counts[ngram]) for ngram, count in counts.items()\n        # }\n\n        # numerator += sum(clipped_counts.values())\n        # # Ensures that denominator is minimum 1 to avoid ZeroDivisionError.\n        # # Usually this happens when the ngram order is > len(reference).\n        # denominator += max(1, sum(counts.values()))\n\n    #return Fraction(numerator, denominator, _normalize=False)\n    return numerator, denominator\n\n\ndef closest_ref_length(references, hyp_len):\n    \"\"\"\n    This function finds the reference that is the closest length to the\n    hypothesis. The closest reference length is referred to as *r* variable\n    from the brevity penalty formula in Papineni et. al. (2002)\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hyp_len: The length of the hypothesis.\n    :type hyp_len: int\n    :return: The length of the reference that's closest to the hypothesis.\n    :rtype: int\n    \"\"\"\n    ref_lens = (len(reference) for reference in references)\n    closest_ref_len = min(\n        ref_lens, key=lambda ref_len: (abs(ref_len - hyp_len), ref_len)\n    )\n    return closest_ref_len\n","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.closest_ref_length","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.closest_ref_length#L271-L287","kind":"function","name":"closest_ref_length","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":271,"end_line":287,"context_start_line":251,"context_end_line":307,"code":"            denominator += max(1, weighted_sum(weights, reference_counts))\n\n        else:\n            numerator += sum(clipped_counts.values())\n            denominator += max(1, sum(reference_counts.values()))\n\n        # # Assigns the intersection between hypothesis and references' counts.\n        # clipped_counts = {\n        #     ngram: min(count, max_counts[ngram]) for ngram, count in counts.items()\n        # }\n\n        # numerator += sum(clipped_counts.values())\n        # # Ensures that denominator is minimum 1 to avoid ZeroDivisionError.\n        # # Usually this happens when the ngram order is > len(reference).\n        # denominator += max(1, sum(counts.values()))\n\n    #return Fraction(numerator, denominator, _normalize=False)\n    return numerator, denominator\n\n\ndef closest_ref_length(references, hyp_len):\n    \"\"\"\n    This function finds the reference that is the closest length to the\n    hypothesis. The closest reference length is referred to as *r* variable\n    from the brevity penalty formula in Papineni et. al. (2002)\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hyp_len: The length of the hypothesis.\n    :type hyp_len: int\n    :return: The length of the reference that's closest to the hypothesis.\n    :rtype: int\n    \"\"\"\n    ref_lens = (len(reference) for reference in references)\n    closest_ref_len = min(\n        ref_lens, key=lambda ref_len: (abs(ref_len - hyp_len), ref_len)\n    )\n    return closest_ref_len\n\n\ndef brevity_penalty(closest_ref_len, hyp_len):\n    \"\"\"\n    Calculate brevity penalty.\n    As the modified n-gram precision still has the problem from the short\n    length sentence, brevity penalty is used to modify the overall BLEU\n    score according to length.\n    An example from the paper. There are three references with length 12, 15\n    and 17. And a concise hypothesis of the length 12. The brevity penalty is 1.\n        >>> reference1 = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> reference2 = list('aaaaaaaaaaaaaaa')   # i.e. ['a'] * 15\n        >>> reference3 = list('aaaaaaaaaaaaaaaaa') # i.e. ['a'] * 17\n        >>> hypothesis = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> references = [reference1, reference2, reference3]\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    In case a hypothesis translation is shorter than the references, penalty is","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.brevity_penalty","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.brevity_penalty#L290-L366","kind":"function","name":"brevity_penalty","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":290,"end_line":366,"context_start_line":270,"context_end_line":386,"code":"\ndef closest_ref_length(references, hyp_len):\n    \"\"\"\n    This function finds the reference that is the closest length to the\n    hypothesis. The closest reference length is referred to as *r* variable\n    from the brevity penalty formula in Papineni et. al. (2002)\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hyp_len: The length of the hypothesis.\n    :type hyp_len: int\n    :return: The length of the reference that's closest to the hypothesis.\n    :rtype: int\n    \"\"\"\n    ref_lens = (len(reference) for reference in references)\n    closest_ref_len = min(\n        ref_lens, key=lambda ref_len: (abs(ref_len - hyp_len), ref_len)\n    )\n    return closest_ref_len\n\n\ndef brevity_penalty(closest_ref_len, hyp_len):\n    \"\"\"\n    Calculate brevity penalty.\n    As the modified n-gram precision still has the problem from the short\n    length sentence, brevity penalty is used to modify the overall BLEU\n    score according to length.\n    An example from the paper. There are three references with length 12, 15\n    and 17. And a concise hypothesis of the length 12. The brevity penalty is 1.\n        >>> reference1 = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> reference2 = list('aaaaaaaaaaaaaaa')   # i.e. ['a'] * 15\n        >>> reference3 = list('aaaaaaaaaaaaaaaaa') # i.e. ['a'] * 17\n        >>> hypothesis = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> references = [reference1, reference2, reference3]\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    In case a hypothesis translation is shorter than the references, penalty is\n    applied.\n        >>> references = [['a'] * 28, ['a'] * 28]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        0.2635971381157267\n    The length of the closest reference is used to compute the penalty. If the\n    length of a hypothesis is 12, and the reference lengths are 13 and 2, the\n    penalty is applied because the hypothesis length (12) is less then the\n    closest reference length (13).\n        >>> references = [['a'] * 13, ['a'] * 2]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len) # doctest: +ELLIPSIS\n        0.9200...\n    The brevity penalty doesn't depend on reference order. More importantly,\n    when two reference sentences are at the same distance, the shortest\n    reference sentence length is used.\n        >>> references = [['a'] * 13, ['a'] * 11]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> bp1 = brevity_penalty(closest_ref_len, hyp_len)\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(reversed(references), hyp_len)\n        >>> bp2 = brevity_penalty(closest_ref_len, hyp_len)\n        >>> bp1 == bp2 == 1\n        True\n    A test example from mteval-v13a.pl (starting from the line 705):\n        >>> references = [['a'] * 11, ['a'] * 8]\n        >>> hypothesis = ['a'] * 7\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len) # doctest: +ELLIPSIS\n        0.8668...\n        >>> references = [['a'] * 11, ['a'] * 8, ['a'] * 6, ['a'] * 7]\n        >>> hypothesis = ['a'] * 7\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    :param hyp_len: The length of the hypothesis for a single sentence OR the\n    sum of all the hypotheses' lengths for a corpus\n    :type hyp_len: int\n    :param closest_ref_len: The length of the closest reference for a single\n    hypothesis OR the sum of all the closest references for every hypotheses.\n    :type closest_ref_len: int\n    :return: BLEU's brevity penalty.\n    :rtype: float\n    \"\"\"\n    if hyp_len > closest_ref_len:\n        return 1\n    # If hypothesis is empty, brevity penalty = 0 should result in BLEU = 0.0\n    elif hyp_len == 0:\n        return 0\n    else:\n        return math.exp(1 - closest_ref_len / hyp_len)\n\n\nclass SmoothingFunction:\n    \"\"\"\n    This is an implementation of the smoothing techniques\n    for segment-level BLEU scores that was presented in\n    Boxing Chen and Collin Cherry (2014) A Systematic Comparison of\n    Smoothing Techniques for Sentence-Level BLEU. In WMT14.\n    http://acl2014.org/acl2014/W14-33/pdf/W14-3346.pdf\n    \"\"\"\n\n    def __init__(self, epsilon=0.1, alpha=5, k=5):\n        \"\"\"\n        This will initialize the parameters required for the various smoothing\n        techniques, the default values are set to the numbers used in the\n        experiments from Chen and Cherry (2014).\n        >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which', 'ensures',\n        ...                 'that', 'the', 'military', 'always', 'obeys', 'the',\n        ...                 'commands', 'of', 'the', 'party']\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that', 'ensures',","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.SmoothingFunction","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.SmoothingFunction#L369-L558","kind":"class","name":"SmoothingFunction","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":369,"end_line":558,"context_start_line":349,"context_end_line":558,"code":"        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    :param hyp_len: The length of the hypothesis for a single sentence OR the\n    sum of all the hypotheses' lengths for a corpus\n    :type hyp_len: int\n    :param closest_ref_len: The length of the closest reference for a single\n    hypothesis OR the sum of all the closest references for every hypotheses.\n    :type closest_ref_len: int\n    :return: BLEU's brevity penalty.\n    :rtype: float\n    \"\"\"\n    if hyp_len > closest_ref_len:\n        return 1\n    # If hypothesis is empty, brevity penalty = 0 should result in BLEU = 0.0\n    elif hyp_len == 0:\n        return 0\n    else:\n        return math.exp(1 - closest_ref_len / hyp_len)\n\n\nclass SmoothingFunction:\n    \"\"\"\n    This is an implementation of the smoothing techniques\n    for segment-level BLEU scores that was presented in\n    Boxing Chen and Collin Cherry (2014) A Systematic Comparison of\n    Smoothing Techniques for Sentence-Level BLEU. In WMT14.\n    http://acl2014.org/acl2014/W14-33/pdf/W14-3346.pdf\n    \"\"\"\n\n    def __init__(self, epsilon=0.1, alpha=5, k=5):\n        \"\"\"\n        This will initialize the parameters required for the various smoothing\n        techniques, the default values are set to the numbers used in the\n        experiments from Chen and Cherry (2014).\n        >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which', 'ensures',\n        ...                 'that', 'the', 'military', 'always', 'obeys', 'the',\n        ...                 'commands', 'of', 'the', 'party']\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that', 'ensures',\n        ...               'that', 'the', 'military', 'will', 'forever', 'heed',\n        ...               'Party', 'commands']\n        >>> chencherry = SmoothingFunction()\n        >>> print(sentence_bleu([reference1], hypothesis1)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method0)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method1)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method2)) # doctest: +ELLIPSIS\n        0.4489...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method3)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method4)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method5)) # doctest: +ELLIPSIS\n        0.4905...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method6)) # doctest: +ELLIPSIS\n        0.4135...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method7)) # doctest: +ELLIPSIS\n        0.4905...\n        :param epsilon: the epsilon value use in method 1\n        :type epsilon: float\n        :param alpha: the alpha value use in method 6\n        :type alpha: int\n        :param k: the k value use in method 4\n        :type k: int\n        \"\"\"\n        self.epsilon = epsilon\n        self.alpha = alpha\n        self.k = k\n\n    def method0(self, p_n, *args, **kwargs):\n        \"\"\"\n        No smoothing.\n        \"\"\"\n        p_n_new = []\n        for i, p_i in enumerate(p_n):\n            if p_i[0] != 0:\n                p_n_new.append(p_i)\n            else:\n                _msg = str(\n                    \"\\nThe hypothesis contains 0 counts of {}-gram overlaps.\\n\"\n                    \"Therefore the BLEU score evaluates to 0, independently of\\n\"\n                    \"how many N-gram overlaps of lower order it contains.\\n\"\n                    \"Consider using lower n-gram order or use \"\n                    \"SmoothingFunction()\"\n                ).format(i + 1)\n                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.\n                # Due to BLEU geometric mean computation in logarithm space,\n                # we we need to take the return sys.float_info.min such that\n                # math.log(sys.float_info.min) returns a 0 precision score.\n                p_n_new.append(sys.float_info.min)\n        return p_n_new\n\n    def method1(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 1: Add *epsilon* counts to precision with 0 counts.\n        \"\"\"\n        return [\n            ((p_i[0] + self.epsilon),  p_i[1])\n            if p_i[0] == 0\n            else p_i\n            for p_i in p_n\n        ]\n\n    def method2(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 2: Add 1 to both numerator and denominator from\n        Chin-Yew Lin and Franz Josef Och (2004) Automatic evaluation of\n        machine translation quality using longest common subsequence and\n        skip-bigram statistics. In ACL04.\n        \"\"\"\n        return [\n            (p_i[0] + 1, p_i[1] + 1)\n            for p_i in p_n\n        ]\n\n    def method3(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 3: NIST geometric sequence smoothing\n        The smoothing is computed by taking 1 / ( 2^k ), instead of 0, for each\n        precision score whose matching n-gram count is null.\n        k is 1 for the first 'n' value for which the n-gram match count is null/\n        For example, if the text contains:\n         - one 2-gram match\n         - and (consequently) two 1-gram matches\n        the n-gram count for each individual precision score would be:\n         - n=1  =>  prec_count = 2     (two unigrams)\n         - n=2  =>  prec_count = 1     (one bigram)\n         - n=3  =>  prec_count = 1/2   (no trigram,  taking 'smoothed' value of 1 / ( 2^k ), with k=1)\n         - n=4  =>  prec_count = 1/4   (no fourgram, taking 'smoothed' value of 1 / ( 2^k ), with k=2)\n        \"\"\"\n        incvnt = 1  # From the mteval-v13a.pl, it's referred to as k.\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0:\n                p_n[i] = 1 / (2 ** incvnt * p_i.denominator)\n                incvnt += 1\n        return p_n\n\n    def method4(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 4:\n        Shorter translations may have inflated precision values due to having\n        smaller denominators; therefore, we give them proportionally\n        smaller smoothed counts. Instead of scaling to 1/(2^k), Chen and Cherry\n        suggests dividing by 1/ln(len(T)), where T is the length of the translation.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0 and hyp_len != 0:\n                incvnt = i + 1 * self.k / math.log(\n                    hyp_len\n                )  # Note that this K is different from the K from NIST.\n                p_n[i] = incvnt / p_i.denominator\n        return p_n\n\n    def method5(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 5:\n        The matched counts for similar values of n should be similar. To a\n        calculate the n-gram matched count, it averages the n−1, n and n+1 gram\n        matched counts.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        m = {}\n        # Requires an precision value for an addition ngram order.\n        p_n_plus1 = p_n + [modified_precision(references, hypothesis, 5)]\n        m[-1] = p_n[0] + 1\n        for i, p_i in enumerate(p_n):\n            p_n[i] = (m[i - 1] + p_i + p_n_plus1[i + 1]) / 3\n            m[i] = p_n[i]\n        return p_n\n\n    def method6(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 6:\n        Interpolates the maximum likelihood estimate of the precision *p_n* with\n        a prior estimate *pi0*. The prior is estimated by assuming that the ratio\n        between pn and pn−1 will be the same as that between pn−1 and pn−2; from\n        Gao and He (2013) Training MRF-Based Phrase Translation Models using\n        Gradient Ascent. In NAACL.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        # This smoothing only works when p_1 and p_2 is non-zero.\n        # Raise an error with an appropriate message when the input is too short\n        # to use this smoothing technique.\n        assert p_n[2], \"This smoothing method requires non-zero precision for bigrams.\"\n        for i, p_i in enumerate(p_n):\n            if i in [0, 1]:  # Skips the first 2 orders of ngrams.\n                continue\n            else:\n                pi0 = 0 if p_n[i - 2] == 0 else p_n[i - 1] ** 2 / p_n[i - 2]\n                # No. of ngrams in translation that matches the reference.\n                m = p_i.numerator\n                # No. of ngrams in translation.\n                l = sum(1 for _ in ngrams(hypothesis, i + 1))\n                # Calculates the interpolated precision.\n                p_n[i] = (m + self.alpha * pi0) / (l + self.alpha)\n        return p_n\n\n    def method7(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 7:\n        Interpolates methods 4 and 5.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        p_n = self.method4(p_n, references, hypothesis, hyp_len)\n        p_n = self.method5(p_n, references, hypothesis, hyp_len)\n        return p_n","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.__init__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.__init__#L378-L417","kind":"function","name":"__init__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":378,"end_line":417,"context_start_line":358,"context_end_line":437,"code":"    :rtype: float\n    \"\"\"\n    if hyp_len > closest_ref_len:\n        return 1\n    # If hypothesis is empty, brevity penalty = 0 should result in BLEU = 0.0\n    elif hyp_len == 0:\n        return 0\n    else:\n        return math.exp(1 - closest_ref_len / hyp_len)\n\n\nclass SmoothingFunction:\n    \"\"\"\n    This is an implementation of the smoothing techniques\n    for segment-level BLEU scores that was presented in\n    Boxing Chen and Collin Cherry (2014) A Systematic Comparison of\n    Smoothing Techniques for Sentence-Level BLEU. In WMT14.\n    http://acl2014.org/acl2014/W14-33/pdf/W14-3346.pdf\n    \"\"\"\n\n    def __init__(self, epsilon=0.1, alpha=5, k=5):\n        \"\"\"\n        This will initialize the parameters required for the various smoothing\n        techniques, the default values are set to the numbers used in the\n        experiments from Chen and Cherry (2014).\n        >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which', 'ensures',\n        ...                 'that', 'the', 'military', 'always', 'obeys', 'the',\n        ...                 'commands', 'of', 'the', 'party']\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that', 'ensures',\n        ...               'that', 'the', 'military', 'will', 'forever', 'heed',\n        ...               'Party', 'commands']\n        >>> chencherry = SmoothingFunction()\n        >>> print(sentence_bleu([reference1], hypothesis1)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method0)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method1)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method2)) # doctest: +ELLIPSIS\n        0.4489...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method3)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method4)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method5)) # doctest: +ELLIPSIS\n        0.4905...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method6)) # doctest: +ELLIPSIS\n        0.4135...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method7)) # doctest: +ELLIPSIS\n        0.4905...\n        :param epsilon: the epsilon value use in method 1\n        :type epsilon: float\n        :param alpha: the alpha value use in method 6\n        :type alpha: int\n        :param k: the k value use in method 4\n        :type k: int\n        \"\"\"\n        self.epsilon = epsilon\n        self.alpha = alpha\n        self.k = k\n\n    def method0(self, p_n, *args, **kwargs):\n        \"\"\"\n        No smoothing.\n        \"\"\"\n        p_n_new = []\n        for i, p_i in enumerate(p_n):\n            if p_i[0] != 0:\n                p_n_new.append(p_i)\n            else:\n                _msg = str(\n                    \"\\nThe hypothesis contains 0 counts of {}-gram overlaps.\\n\"\n                    \"Therefore the BLEU score evaluates to 0, independently of\\n\"\n                    \"how many N-gram overlaps of lower order it contains.\\n\"\n                    \"Consider using lower n-gram order or use \"\n                    \"SmoothingFunction()\"\n                ).format(i + 1)\n                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method0","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method0#L419-L442","kind":"function","name":"method0","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":419,"end_line":442,"context_start_line":399,"context_end_line":462,"code":"        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method4)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method5)) # doctest: +ELLIPSIS\n        0.4905...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method6)) # doctest: +ELLIPSIS\n        0.4135...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method7)) # doctest: +ELLIPSIS\n        0.4905...\n        :param epsilon: the epsilon value use in method 1\n        :type epsilon: float\n        :param alpha: the alpha value use in method 6\n        :type alpha: int\n        :param k: the k value use in method 4\n        :type k: int\n        \"\"\"\n        self.epsilon = epsilon\n        self.alpha = alpha\n        self.k = k\n\n    def method0(self, p_n, *args, **kwargs):\n        \"\"\"\n        No smoothing.\n        \"\"\"\n        p_n_new = []\n        for i, p_i in enumerate(p_n):\n            if p_i[0] != 0:\n                p_n_new.append(p_i)\n            else:\n                _msg = str(\n                    \"\\nThe hypothesis contains 0 counts of {}-gram overlaps.\\n\"\n                    \"Therefore the BLEU score evaluates to 0, independently of\\n\"\n                    \"how many N-gram overlaps of lower order it contains.\\n\"\n                    \"Consider using lower n-gram order or use \"\n                    \"SmoothingFunction()\"\n                ).format(i + 1)\n                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.\n                # Due to BLEU geometric mean computation in logarithm space,\n                # we we need to take the return sys.float_info.min such that\n                # math.log(sys.float_info.min) returns a 0 precision score.\n                p_n_new.append(sys.float_info.min)\n        return p_n_new\n\n    def method1(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 1: Add *epsilon* counts to precision with 0 counts.\n        \"\"\"\n        return [\n            ((p_i[0] + self.epsilon),  p_i[1])\n            if p_i[0] == 0\n            else p_i\n            for p_i in p_n\n        ]\n\n    def method2(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 2: Add 1 to both numerator and denominator from\n        Chin-Yew Lin and Franz Josef Och (2004) Automatic evaluation of\n        machine translation quality using longest common subsequence and\n        skip-bigram statistics. In ACL04.\n        \"\"\"\n        return [","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method1#L444-L453","kind":"function","name":"method1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":444,"end_line":453,"context_start_line":424,"context_end_line":473,"code":"        for i, p_i in enumerate(p_n):\n            if p_i[0] != 0:\n                p_n_new.append(p_i)\n            else:\n                _msg = str(\n                    \"\\nThe hypothesis contains 0 counts of {}-gram overlaps.\\n\"\n                    \"Therefore the BLEU score evaluates to 0, independently of\\n\"\n                    \"how many N-gram overlaps of lower order it contains.\\n\"\n                    \"Consider using lower n-gram order or use \"\n                    \"SmoothingFunction()\"\n                ).format(i + 1)\n                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.\n                # Due to BLEU geometric mean computation in logarithm space,\n                # we we need to take the return sys.float_info.min such that\n                # math.log(sys.float_info.min) returns a 0 precision score.\n                p_n_new.append(sys.float_info.min)\n        return p_n_new\n\n    def method1(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 1: Add *epsilon* counts to precision with 0 counts.\n        \"\"\"\n        return [\n            ((p_i[0] + self.epsilon),  p_i[1])\n            if p_i[0] == 0\n            else p_i\n            for p_i in p_n\n        ]\n\n    def method2(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 2: Add 1 to both numerator and denominator from\n        Chin-Yew Lin and Franz Josef Och (2004) Automatic evaluation of\n        machine translation quality using longest common subsequence and\n        skip-bigram statistics. In ACL04.\n        \"\"\"\n        return [\n            (p_i[0] + 1, p_i[1] + 1)\n            for p_i in p_n\n        ]\n\n    def method3(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 3: NIST geometric sequence smoothing\n        The smoothing is computed by taking 1 / ( 2^k ), instead of 0, for each\n        precision score whose matching n-gram count is null.\n        k is 1 for the first 'n' value for which the n-gram match count is null/\n        For example, if the text contains:","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method2#L455-L465","kind":"function","name":"method2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":455,"end_line":465,"context_start_line":435,"context_end_line":485,"code":"                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.\n                # Due to BLEU geometric mean computation in logarithm space,\n                # we we need to take the return sys.float_info.min such that\n                # math.log(sys.float_info.min) returns a 0 precision score.\n                p_n_new.append(sys.float_info.min)\n        return p_n_new\n\n    def method1(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 1: Add *epsilon* counts to precision with 0 counts.\n        \"\"\"\n        return [\n            ((p_i[0] + self.epsilon),  p_i[1])\n            if p_i[0] == 0\n            else p_i\n            for p_i in p_n\n        ]\n\n    def method2(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 2: Add 1 to both numerator and denominator from\n        Chin-Yew Lin and Franz Josef Och (2004) Automatic evaluation of\n        machine translation quality using longest common subsequence and\n        skip-bigram statistics. In ACL04.\n        \"\"\"\n        return [\n            (p_i[0] + 1, p_i[1] + 1)\n            for p_i in p_n\n        ]\n\n    def method3(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 3: NIST geometric sequence smoothing\n        The smoothing is computed by taking 1 / ( 2^k ), instead of 0, for each\n        precision score whose matching n-gram count is null.\n        k is 1 for the first 'n' value for which the n-gram match count is null/\n        For example, if the text contains:\n         - one 2-gram match\n         - and (consequently) two 1-gram matches\n        the n-gram count for each individual precision score would be:\n         - n=1  =>  prec_count = 2     (two unigrams)\n         - n=2  =>  prec_count = 1     (one bigram)\n         - n=3  =>  prec_count = 1/2   (no trigram,  taking 'smoothed' value of 1 / ( 2^k ), with k=1)\n         - n=4  =>  prec_count = 1/4   (no fourgram, taking 'smoothed' value of 1 / ( 2^k ), with k=2)\n        \"\"\"\n        incvnt = 1  # From the mteval-v13a.pl, it's referred to as k.\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0:\n                p_n[i] = 1 / (2 ** incvnt * p_i.denominator)","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method3","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method3#L467-L487","kind":"function","name":"method3","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":467,"end_line":487,"context_start_line":447,"context_end_line":507,"code":"        \"\"\"\n        return [\n            ((p_i[0] + self.epsilon),  p_i[1])\n            if p_i[0] == 0\n            else p_i\n            for p_i in p_n\n        ]\n\n    def method2(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 2: Add 1 to both numerator and denominator from\n        Chin-Yew Lin and Franz Josef Och (2004) Automatic evaluation of\n        machine translation quality using longest common subsequence and\n        skip-bigram statistics. In ACL04.\n        \"\"\"\n        return [\n            (p_i[0] + 1, p_i[1] + 1)\n            for p_i in p_n\n        ]\n\n    def method3(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 3: NIST geometric sequence smoothing\n        The smoothing is computed by taking 1 / ( 2^k ), instead of 0, for each\n        precision score whose matching n-gram count is null.\n        k is 1 for the first 'n' value for which the n-gram match count is null/\n        For example, if the text contains:\n         - one 2-gram match\n         - and (consequently) two 1-gram matches\n        the n-gram count for each individual precision score would be:\n         - n=1  =>  prec_count = 2     (two unigrams)\n         - n=2  =>  prec_count = 1     (one bigram)\n         - n=3  =>  prec_count = 1/2   (no trigram,  taking 'smoothed' value of 1 / ( 2^k ), with k=1)\n         - n=4  =>  prec_count = 1/4   (no fourgram, taking 'smoothed' value of 1 / ( 2^k ), with k=2)\n        \"\"\"\n        incvnt = 1  # From the mteval-v13a.pl, it's referred to as k.\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0:\n                p_n[i] = 1 / (2 ** incvnt * p_i.denominator)\n                incvnt += 1\n        return p_n\n\n    def method4(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 4:\n        Shorter translations may have inflated precision values due to having\n        smaller denominators; therefore, we give them proportionally\n        smaller smoothed counts. Instead of scaling to 1/(2^k), Chen and Cherry\n        suggests dividing by 1/ln(len(T)), where T is the length of the translation.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0 and hyp_len != 0:\n                incvnt = i + 1 * self.k / math.log(\n                    hyp_len\n                )  # Note that this K is different from the K from NIST.\n                p_n[i] = incvnt / p_i.denominator\n        return p_n\n\n    def method5(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method4","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method4#L489-L504","kind":"function","name":"method4","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":489,"end_line":504,"context_start_line":469,"context_end_line":524,"code":"        Smoothing method 3: NIST geometric sequence smoothing\n        The smoothing is computed by taking 1 / ( 2^k ), instead of 0, for each\n        precision score whose matching n-gram count is null.\n        k is 1 for the first 'n' value for which the n-gram match count is null/\n        For example, if the text contains:\n         - one 2-gram match\n         - and (consequently) two 1-gram matches\n        the n-gram count for each individual precision score would be:\n         - n=1  =>  prec_count = 2     (two unigrams)\n         - n=2  =>  prec_count = 1     (one bigram)\n         - n=3  =>  prec_count = 1/2   (no trigram,  taking 'smoothed' value of 1 / ( 2^k ), with k=1)\n         - n=4  =>  prec_count = 1/4   (no fourgram, taking 'smoothed' value of 1 / ( 2^k ), with k=2)\n        \"\"\"\n        incvnt = 1  # From the mteval-v13a.pl, it's referred to as k.\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0:\n                p_n[i] = 1 / (2 ** incvnt * p_i.denominator)\n                incvnt += 1\n        return p_n\n\n    def method4(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 4:\n        Shorter translations may have inflated precision values due to having\n        smaller denominators; therefore, we give them proportionally\n        smaller smoothed counts. Instead of scaling to 1/(2^k), Chen and Cherry\n        suggests dividing by 1/ln(len(T)), where T is the length of the translation.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0 and hyp_len != 0:\n                incvnt = i + 1 * self.k / math.log(\n                    hyp_len\n                )  # Note that this K is different from the K from NIST.\n                p_n[i] = incvnt / p_i.denominator\n        return p_n\n\n    def method5(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 5:\n        The matched counts for similar values of n should be similar. To a\n        calculate the n-gram matched count, it averages the n−1, n and n+1 gram\n        matched counts.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        m = {}\n        # Requires an precision value for an addition ngram order.\n        p_n_plus1 = p_n + [modified_precision(references, hypothesis, 5)]\n        m[-1] = p_n[0] + 1\n        for i, p_i in enumerate(p_n):\n            p_n[i] = (m[i - 1] + p_i + p_n_plus1[i + 1]) / 3\n            m[i] = p_n[i]\n        return p_n\n\n    def method6(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method5","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method5#L506-L521","kind":"function","name":"method5","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":506,"end_line":521,"context_start_line":486,"context_end_line":541,"code":"                incvnt += 1\n        return p_n\n\n    def method4(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 4:\n        Shorter translations may have inflated precision values due to having\n        smaller denominators; therefore, we give them proportionally\n        smaller smoothed counts. Instead of scaling to 1/(2^k), Chen and Cherry\n        suggests dividing by 1/ln(len(T)), where T is the length of the translation.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0 and hyp_len != 0:\n                incvnt = i + 1 * self.k / math.log(\n                    hyp_len\n                )  # Note that this K is different from the K from NIST.\n                p_n[i] = incvnt / p_i.denominator\n        return p_n\n\n    def method5(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 5:\n        The matched counts for similar values of n should be similar. To a\n        calculate the n-gram matched count, it averages the n−1, n and n+1 gram\n        matched counts.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        m = {}\n        # Requires an precision value for an addition ngram order.\n        p_n_plus1 = p_n + [modified_precision(references, hypothesis, 5)]\n        m[-1] = p_n[0] + 1\n        for i, p_i in enumerate(p_n):\n            p_n[i] = (m[i - 1] + p_i + p_n_plus1[i + 1]) / 3\n            m[i] = p_n[i]\n        return p_n\n\n    def method6(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 6:\n        Interpolates the maximum likelihood estimate of the precision *p_n* with\n        a prior estimate *pi0*. The prior is estimated by assuming that the ratio\n        between pn and pn−1 will be the same as that between pn−1 and pn−2; from\n        Gao and He (2013) Training MRF-Based Phrase Translation Models using\n        Gradient Ascent. In NAACL.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        # This smoothing only works when p_1 and p_2 is non-zero.\n        # Raise an error with an appropriate message when the input is too short\n        # to use this smoothing technique.\n        assert p_n[2], \"This smoothing method requires non-zero precision for bigrams.\"\n        for i, p_i in enumerate(p_n):\n            if i in [0, 1]:  # Skips the first 2 orders of ngrams.\n                continue\n            else:\n                pi0 = 0 if p_n[i - 2] == 0 else p_n[i - 1] ** 2 / p_n[i - 2]","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method6","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method6#L523-L548","kind":"function","name":"method6","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":523,"end_line":548,"context_start_line":503,"context_end_line":558,"code":"                p_n[i] = incvnt / p_i.denominator\n        return p_n\n\n    def method5(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 5:\n        The matched counts for similar values of n should be similar. To a\n        calculate the n-gram matched count, it averages the n−1, n and n+1 gram\n        matched counts.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        m = {}\n        # Requires an precision value for an addition ngram order.\n        p_n_plus1 = p_n + [modified_precision(references, hypothesis, 5)]\n        m[-1] = p_n[0] + 1\n        for i, p_i in enumerate(p_n):\n            p_n[i] = (m[i - 1] + p_i + p_n_plus1[i + 1]) / 3\n            m[i] = p_n[i]\n        return p_n\n\n    def method6(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 6:\n        Interpolates the maximum likelihood estimate of the precision *p_n* with\n        a prior estimate *pi0*. The prior is estimated by assuming that the ratio\n        between pn and pn−1 will be the same as that between pn−1 and pn−2; from\n        Gao and He (2013) Training MRF-Based Phrase Translation Models using\n        Gradient Ascent. In NAACL.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        # This smoothing only works when p_1 and p_2 is non-zero.\n        # Raise an error with an appropriate message when the input is too short\n        # to use this smoothing technique.\n        assert p_n[2], \"This smoothing method requires non-zero precision for bigrams.\"\n        for i, p_i in enumerate(p_n):\n            if i in [0, 1]:  # Skips the first 2 orders of ngrams.\n                continue\n            else:\n                pi0 = 0 if p_n[i - 2] == 0 else p_n[i - 1] ** 2 / p_n[i - 2]\n                # No. of ngrams in translation that matches the reference.\n                m = p_i.numerator\n                # No. of ngrams in translation.\n                l = sum(1 for _ in ngrams(hypothesis, i + 1))\n                # Calculates the interpolated precision.\n                p_n[i] = (m + self.alpha * pi0) / (l + self.alpha)\n        return p_n\n\n    def method7(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 7:\n        Interpolates methods 4 and 5.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        p_n = self.method4(p_n, references, hypothesis, hyp_len)\n        p_n = self.method5(p_n, references, hypothesis, hyp_len)\n        return p_n","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method7","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.method7#L550-L558","kind":"function","name":"method7","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":550,"end_line":558,"context_start_line":530,"context_end_line":558,"code":"        Gradient Ascent. In NAACL.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        # This smoothing only works when p_1 and p_2 is non-zero.\n        # Raise an error with an appropriate message when the input is too short\n        # to use this smoothing technique.\n        assert p_n[2], \"This smoothing method requires non-zero precision for bigrams.\"\n        for i, p_i in enumerate(p_n):\n            if i in [0, 1]:  # Skips the first 2 orders of ngrams.\n                continue\n            else:\n                pi0 = 0 if p_n[i - 2] == 0 else p_n[i - 1] ** 2 / p_n[i - 2]\n                # No. of ngrams in translation that matches the reference.\n                m = p_i.numerator\n                # No. of ngrams in translation.\n                l = sum(1 for _ in ngrams(hypothesis, i + 1))\n                # Calculates the interpolated precision.\n                p_n[i] = (m + self.alpha * pi0) / (l + self.alpha)\n        return p_n\n\n    def method7(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 7:\n        Interpolates methods 4 and 5.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        p_n = self.method4(p_n, references, hypothesis, hyp_len)\n        p_n = self.method5(p_n, references, hypothesis, hyp_len)\n        return p_n","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.weighted_sum","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.weighted_ngram_match.weighted_sum#L244-L248","kind":"function","name":"weighted_sum","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":244,"end_line":248,"context_start_line":224,"context_end_line":268,"code":"    numerator = 0\n    denominator = 0\n\n    counts = Counter(ngrams(hypothesis, n)) if len(hypothesis) >= n else Counter()\n    # Extract a union of references' counts.\n    # max_counts = reduce(or_, [Counter(ngrams(ref, n)) for ref in references])\n    max_counts = {}\n    for reference_and_weights in references:\n        reference = reference_and_weights[0]\n        weights = reference_and_weights[1]\n        reference_counts = (\n            Counter(ngrams(reference, n)) if len(reference) >= n else Counter()\n        )\n        # for ngram in reference_counts:\n        #     max_counts[ngram] = max(max_counts.get(ngram, 0), counts[ngram])\n        clipped_counts = {\n            ngram: min(count, counts[ngram]) for ngram, count in reference_counts.items()\n        }\n        # reweight\n        if n == 1 and len(weights) == len(reference_counts):\n            def weighted_sum(weights, counts):\n                sum_counts = 0\n                for ngram, count in counts.items():\n                    sum_counts += count * (weights[ngram[0]] if ngram[0] in weights else 1)\n                return sum_counts\n\n            numerator += weighted_sum(weights, clipped_counts)\n            denominator += max(1, weighted_sum(weights, reference_counts))\n\n        else:\n            numerator += sum(clipped_counts.values())\n            denominator += max(1, sum(reference_counts.values()))\n\n        # # Assigns the intersection between hypothesis and references' counts.\n        # clipped_counts = {\n        #     ngram: min(count, max_counts[ngram]) for ngram, count in counts.items()\n        # }\n\n        # numerator += sum(clipped_counts.values())\n        # # Ensures that denominator is minimum 1 to avoid ZeroDivisionError.\n        # # Usually this happens when the ngram order is > len(reference).\n        # denominator += max(1, sum(counts.values()))\n\n    #return Fraction(numerator, denominator, _normalize=False)\n    return numerator, denominator","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match#L1-L138","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/dataflow_match.py","language":"python","start_line":1,"end_line":138,"context_start_line":1,"context_end_line":138,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom codebleu.parser import DFG_python\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\nimport pdb\n\ndfg_function={\n    'python':DFG_python\n}\n\ndef calc_dataflow_match(references, candidate, lang):\n    return corpus_dataflow_match([references], [candidate], lang)\n\ndef corpus_dataflow_match(references, candidates, lang):   \n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    parser = [parser,dfg_function[lang]]\n    match_count = 0\n    total_count = 0\n\n    for i in range(len(candidates)):\n        references_sample = references[i]\n        candidate = candidates[i] \n        for reference in references_sample:\n            try:\n                candidate=remove_comments_and_docstrings(candidate,'java')\n            except:\n                pass    \n            try:\n                reference=remove_comments_and_docstrings(reference,'java')\n            except:\n                pass  \n\n            cand_dfg = get_data_flow(candidate, parser)\n            ref_dfg = get_data_flow(reference, parser)\n            \n            normalized_cand_dfg = normalize_dataflow(cand_dfg)\n            normalized_ref_dfg = normalize_dataflow(ref_dfg)\n\n            if len(normalized_ref_dfg) > 0:\n                total_count += len(normalized_ref_dfg)\n                for dataflow in normalized_ref_dfg:\n                    if dataflow in normalized_cand_dfg:\n                            match_count += 1\n                            normalized_cand_dfg.remove(dataflow)    \n\n    if total_count == 0:\n        return 0\n\n    score = match_count / total_count\n    return score\n\ndef get_data_flow(code, parser):\n    try:\n        tree = parser[0].parse(bytes(code,'utf8'))    \n        root_node = tree.root_node  \n        tokens_index=tree_to_token_index(root_node)     \n        code=code.split('\\n')\n        code_tokens=[index_to_code_token(x,code) for x in tokens_index]  \n        index_to_code={}\n        for idx,(index,code) in enumerate(zip(tokens_index,code_tokens)):\n            index_to_code[index]=(idx,code)  \n        try:\n            DFG,_=parser[1](root_node,index_to_code,{}) \n        except:\n            DFG=[]\n        DFG=sorted(DFG,key=lambda x:x[1])\n        indexs=set()\n        for d in DFG:\n            if len(d[-1])!=0:\n                indexs.add(d[1])\n            for x in d[-1]:\n                indexs.add(x)\n        new_DFG=[]\n        for d in DFG:\n            if d[1] in indexs:\n                new_DFG.append(d)\n        codes=code_tokens\n        dfg=new_DFG\n    except:\n        codes=code.split()\n        dfg=[]\n    #merge nodes\n    dic={}\n    for d in dfg:\n        if d[1] not in dic:\n            dic[d[1]]=d\n        else:\n            dic[d[1]]=(d[0],d[1],d[2],list(set(dic[d[1]][3]+d[3])),list(set(dic[d[1]][4]+d[4])))\n    DFG=[]\n    for d in dic:\n        DFG.append(dic[d])\n    dfg=DFG\n    return dfg\n\ndef normalize_dataflow_item(dataflow_item):\n    var_name = dataflow_item[0]\n    var_pos = dataflow_item[1]\n    relationship = dataflow_item[2]\n    par_vars_name_list = dataflow_item[3]\n    par_vars_pos_list = dataflow_item[4]\n\n    var_names = list(set(par_vars_name_list+[var_name]))\n    norm_names = {}\n    for i in range(len(var_names)):\n        norm_names[var_names[i]] = 'var_'+str(i)\n\n    norm_var_name = norm_names[var_name]\n    relationship = dataflow_item[2]\n    norm_par_vars_name_list = [norm_names[x] for x in par_vars_name_list]\n\n    return (norm_var_name, relationship, norm_par_vars_name_list)\n\ndef normalize_dataflow(dataflow):\n    var_dict = {}\n    i = 0\n    normalized_dataflow = []\n    for item in dataflow:\n        var_name = item[0]\n        relationship = item[2]\n        par_vars_name_list = item[3]\n        for name in par_vars_name_list:\n            if name not in var_dict:\n                var_dict[name] = 'var_'+str(i)\n                i += 1\n        if var_name not in var_dict:\n            var_dict[var_name] = 'var_'+str(i)\n            i+= 1\n        normalized_dataflow.append((var_dict[var_name], relationship, [var_dict[x] for x in par_vars_name_list]))\n    return normalized_dataflow\n\n","source_hash":"72304a13041dd77d7b16691cc48d96d584b13818680f9f8b08b94d38eb898841","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match.calc_dataflow_match","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match.calc_dataflow_match#L16-L17","kind":"function","name":"calc_dataflow_match","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/dataflow_match.py","language":"python","start_line":16,"end_line":17,"context_start_line":1,"context_end_line":37,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom codebleu.parser import DFG_python\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\nimport pdb\n\ndfg_function={\n    'python':DFG_python\n}\n\ndef calc_dataflow_match(references, candidate, lang):\n    return corpus_dataflow_match([references], [candidate], lang)\n\ndef corpus_dataflow_match(references, candidates, lang):   \n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    parser = [parser,dfg_function[lang]]\n    match_count = 0\n    total_count = 0\n\n    for i in range(len(candidates)):\n        references_sample = references[i]\n        candidate = candidates[i] \n        for reference in references_sample:\n            try:\n                candidate=remove_comments_and_docstrings(candidate,'java')\n            except:\n                pass    \n            try:\n                reference=remove_comments_and_docstrings(reference,'java')\n            except:","source_hash":"72304a13041dd77d7b16691cc48d96d584b13818680f9f8b08b94d38eb898841","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match.corpus_dataflow_match","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match.corpus_dataflow_match#L19-L57","kind":"function","name":"corpus_dataflow_match","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/dataflow_match.py","language":"python","start_line":19,"end_line":57,"context_start_line":1,"context_end_line":77,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom codebleu.parser import DFG_python\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\nimport pdb\n\ndfg_function={\n    'python':DFG_python\n}\n\ndef calc_dataflow_match(references, candidate, lang):\n    return corpus_dataflow_match([references], [candidate], lang)\n\ndef corpus_dataflow_match(references, candidates, lang):   \n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    parser = [parser,dfg_function[lang]]\n    match_count = 0\n    total_count = 0\n\n    for i in range(len(candidates)):\n        references_sample = references[i]\n        candidate = candidates[i] \n        for reference in references_sample:\n            try:\n                candidate=remove_comments_and_docstrings(candidate,'java')\n            except:\n                pass    \n            try:\n                reference=remove_comments_and_docstrings(reference,'java')\n            except:\n                pass  \n\n            cand_dfg = get_data_flow(candidate, parser)\n            ref_dfg = get_data_flow(reference, parser)\n            \n            normalized_cand_dfg = normalize_dataflow(cand_dfg)\n            normalized_ref_dfg = normalize_dataflow(ref_dfg)\n\n            if len(normalized_ref_dfg) > 0:\n                total_count += len(normalized_ref_dfg)\n                for dataflow in normalized_ref_dfg:\n                    if dataflow in normalized_cand_dfg:\n                            match_count += 1\n                            normalized_cand_dfg.remove(dataflow)    \n\n    if total_count == 0:\n        return 0\n\n    score = match_count / total_count\n    return score\n\ndef get_data_flow(code, parser):\n    try:\n        tree = parser[0].parse(bytes(code,'utf8'))    \n        root_node = tree.root_node  \n        tokens_index=tree_to_token_index(root_node)     \n        code=code.split('\\n')\n        code_tokens=[index_to_code_token(x,code) for x in tokens_index]  \n        index_to_code={}\n        for idx,(index,code) in enumerate(zip(tokens_index,code_tokens)):\n            index_to_code[index]=(idx,code)  \n        try:\n            DFG,_=parser[1](root_node,index_to_code,{}) \n        except:\n            DFG=[]\n        DFG=sorted(DFG,key=lambda x:x[1])\n        indexs=set()\n        for d in DFG:\n            if len(d[-1])!=0:\n                indexs.add(d[1])","source_hash":"72304a13041dd77d7b16691cc48d96d584b13818680f9f8b08b94d38eb898841","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match.get_data_flow","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match.get_data_flow#L59-L100","kind":"function","name":"get_data_flow","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/dataflow_match.py","language":"python","start_line":59,"end_line":100,"context_start_line":39,"context_end_line":120,"code":"\n            cand_dfg = get_data_flow(candidate, parser)\n            ref_dfg = get_data_flow(reference, parser)\n            \n            normalized_cand_dfg = normalize_dataflow(cand_dfg)\n            normalized_ref_dfg = normalize_dataflow(ref_dfg)\n\n            if len(normalized_ref_dfg) > 0:\n                total_count += len(normalized_ref_dfg)\n                for dataflow in normalized_ref_dfg:\n                    if dataflow in normalized_cand_dfg:\n                            match_count += 1\n                            normalized_cand_dfg.remove(dataflow)    \n\n    if total_count == 0:\n        return 0\n\n    score = match_count / total_count\n    return score\n\ndef get_data_flow(code, parser):\n    try:\n        tree = parser[0].parse(bytes(code,'utf8'))    \n        root_node = tree.root_node  \n        tokens_index=tree_to_token_index(root_node)     \n        code=code.split('\\n')\n        code_tokens=[index_to_code_token(x,code) for x in tokens_index]  \n        index_to_code={}\n        for idx,(index,code) in enumerate(zip(tokens_index,code_tokens)):\n            index_to_code[index]=(idx,code)  \n        try:\n            DFG,_=parser[1](root_node,index_to_code,{}) \n        except:\n            DFG=[]\n        DFG=sorted(DFG,key=lambda x:x[1])\n        indexs=set()\n        for d in DFG:\n            if len(d[-1])!=0:\n                indexs.add(d[1])\n            for x in d[-1]:\n                indexs.add(x)\n        new_DFG=[]\n        for d in DFG:\n            if d[1] in indexs:\n                new_DFG.append(d)\n        codes=code_tokens\n        dfg=new_DFG\n    except:\n        codes=code.split()\n        dfg=[]\n    #merge nodes\n    dic={}\n    for d in dfg:\n        if d[1] not in dic:\n            dic[d[1]]=d\n        else:\n            dic[d[1]]=(d[0],d[1],d[2],list(set(dic[d[1]][3]+d[3])),list(set(dic[d[1]][4]+d[4])))\n    DFG=[]\n    for d in dic:\n        DFG.append(dic[d])\n    dfg=DFG\n    return dfg\n\ndef normalize_dataflow_item(dataflow_item):\n    var_name = dataflow_item[0]\n    var_pos = dataflow_item[1]\n    relationship = dataflow_item[2]\n    par_vars_name_list = dataflow_item[3]\n    par_vars_pos_list = dataflow_item[4]\n\n    var_names = list(set(par_vars_name_list+[var_name]))\n    norm_names = {}\n    for i in range(len(var_names)):\n        norm_names[var_names[i]] = 'var_'+str(i)\n\n    norm_var_name = norm_names[var_name]\n    relationship = dataflow_item[2]\n    norm_par_vars_name_list = [norm_names[x] for x in par_vars_name_list]\n\n    return (norm_var_name, relationship, norm_par_vars_name_list)\n\ndef normalize_dataflow(dataflow):","source_hash":"72304a13041dd77d7b16691cc48d96d584b13818680f9f8b08b94d38eb898841","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match.normalize_dataflow_item","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match.normalize_dataflow_item#L102-L118","kind":"function","name":"normalize_dataflow_item","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/dataflow_match.py","language":"python","start_line":102,"end_line":118,"context_start_line":82,"context_end_line":138,"code":"            if d[1] in indexs:\n                new_DFG.append(d)\n        codes=code_tokens\n        dfg=new_DFG\n    except:\n        codes=code.split()\n        dfg=[]\n    #merge nodes\n    dic={}\n    for d in dfg:\n        if d[1] not in dic:\n            dic[d[1]]=d\n        else:\n            dic[d[1]]=(d[0],d[1],d[2],list(set(dic[d[1]][3]+d[3])),list(set(dic[d[1]][4]+d[4])))\n    DFG=[]\n    for d in dic:\n        DFG.append(dic[d])\n    dfg=DFG\n    return dfg\n\ndef normalize_dataflow_item(dataflow_item):\n    var_name = dataflow_item[0]\n    var_pos = dataflow_item[1]\n    relationship = dataflow_item[2]\n    par_vars_name_list = dataflow_item[3]\n    par_vars_pos_list = dataflow_item[4]\n\n    var_names = list(set(par_vars_name_list+[var_name]))\n    norm_names = {}\n    for i in range(len(var_names)):\n        norm_names[var_names[i]] = 'var_'+str(i)\n\n    norm_var_name = norm_names[var_name]\n    relationship = dataflow_item[2]\n    norm_par_vars_name_list = [norm_names[x] for x in par_vars_name_list]\n\n    return (norm_var_name, relationship, norm_par_vars_name_list)\n\ndef normalize_dataflow(dataflow):\n    var_dict = {}\n    i = 0\n    normalized_dataflow = []\n    for item in dataflow:\n        var_name = item[0]\n        relationship = item[2]\n        par_vars_name_list = item[3]\n        for name in par_vars_name_list:\n            if name not in var_dict:\n                var_dict[name] = 'var_'+str(i)\n                i += 1\n        if var_name not in var_dict:\n            var_dict[var_name] = 'var_'+str(i)\n            i+= 1\n        normalized_dataflow.append((var_dict[var_name], relationship, [var_dict[x] for x in par_vars_name_list]))\n    return normalized_dataflow\n\n","source_hash":"72304a13041dd77d7b16691cc48d96d584b13818680f9f8b08b94d38eb898841","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match.normalize_dataflow","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.dataflow_match.normalize_dataflow#L120-L136","kind":"function","name":"normalize_dataflow","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/dataflow_match.py","language":"python","start_line":120,"end_line":136,"context_start_line":100,"context_end_line":138,"code":"    return dfg\n\ndef normalize_dataflow_item(dataflow_item):\n    var_name = dataflow_item[0]\n    var_pos = dataflow_item[1]\n    relationship = dataflow_item[2]\n    par_vars_name_list = dataflow_item[3]\n    par_vars_pos_list = dataflow_item[4]\n\n    var_names = list(set(par_vars_name_list+[var_name]))\n    norm_names = {}\n    for i in range(len(var_names)):\n        norm_names[var_names[i]] = 'var_'+str(i)\n\n    norm_var_name = norm_names[var_name]\n    relationship = dataflow_item[2]\n    norm_par_vars_name_list = [norm_names[x] for x in par_vars_name_list]\n\n    return (norm_var_name, relationship, norm_par_vars_name_list)\n\ndef normalize_dataflow(dataflow):\n    var_dict = {}\n    i = 0\n    normalized_dataflow = []\n    for item in dataflow:\n        var_name = item[0]\n        relationship = item[2]\n        par_vars_name_list = item[3]\n        for name in par_vars_name_list:\n            if name not in var_dict:\n                var_dict[name] = 'var_'+str(i)\n                i += 1\n        if var_name not in var_dict:\n            var_dict[var_name] = 'var_'+str(i)\n            i+= 1\n        normalized_dataflow.append((var_dict[var_name], relationship, [var_dict[x] for x in par_vars_name_list]))\n    return normalized_dataflow\n\n","source_hash":"72304a13041dd77d7b16691cc48d96d584b13818680f9f8b08b94d38eb898841","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.utils","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.utils#L1-L106","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.utils","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/utils.py","language":"python","start_line":1,"end_line":106,"context_start_line":1,"context_end_line":106,"code":"# Natural Language Toolkit: Utility functions\n#\n# Copyright (C) 2001-2020 NLTK Project\n# Author: Steven Bird <stevenbird1@gmail.com>\n# URL: <http://nltk.org/>\n# For license information, see LICENSE.TXT\n\nfrom itertools import chain\n\ndef pad_sequence(\n    sequence,\n    n,\n    pad_left=False,\n    pad_right=False,\n    left_pad_symbol=None,\n    right_pad_symbol=None,\n):\n    \"\"\"\n    Returns a padded sequence of items before ngram extraction.\n        >>> list(pad_sequence([1,2,3,4,5], 2, pad_left=True, pad_right=True, left_pad_symbol='<s>', right_pad_symbol='</s>'))\n        ['<s>', 1, 2, 3, 4, 5, '</s>']\n        >>> list(pad_sequence([1,2,3,4,5], 2, pad_left=True, left_pad_symbol='<s>'))\n        ['<s>', 1, 2, 3, 4, 5]\n        >>> list(pad_sequence([1,2,3,4,5], 2, pad_right=True, right_pad_symbol='</s>'))\n        [1, 2, 3, 4, 5, '</s>']\n    :param sequence: the source data to be padded\n    :type sequence: sequence or iter\n    :param n: the degree of the ngrams\n    :type n: int\n    :param pad_left: whether the ngrams should be left-padded\n    :type pad_left: bool\n    :param pad_right: whether the ngrams should be right-padded\n    :type pad_right: bool\n    :param left_pad_symbol: the symbol to use for left padding (default is None)\n    :type left_pad_symbol: any\n    :param right_pad_symbol: the symbol to use for right padding (default is None)\n    :type right_pad_symbol: any\n    :rtype: sequence or iter\n    \"\"\"\n    sequence = iter(sequence)\n    if pad_left:\n        sequence = chain((left_pad_symbol,) * (n - 1), sequence)\n    if pad_right:\n        sequence = chain(sequence, (right_pad_symbol,) * (n - 1))\n    return sequence\n\n\n# add a flag to pad the sequence so we get peripheral ngrams?\n\n\ndef ngrams(\n    sequence,\n    n,\n    pad_left=False,\n    pad_right=False,\n    left_pad_symbol=None,\n    right_pad_symbol=None,\n):\n    \"\"\"\n    Return the ngrams generated from a sequence of items, as an iterator.\n    For example:\n        >>> from nltk.util import ngrams\n        >>> list(ngrams([1,2,3,4,5], 3))\n        [(1, 2, 3), (2, 3, 4), (3, 4, 5)]\n    Wrap with list for a list version of this function.  Set pad_left\n    or pad_right to true in order to get additional ngrams:\n        >>> list(ngrams([1,2,3,4,5], 2, pad_right=True))\n        [(1, 2), (2, 3), (3, 4), (4, 5), (5, None)]\n        >>> list(ngrams([1,2,3,4,5], 2, pad_right=True, right_pad_symbol='</s>'))\n        [(1, 2), (2, 3), (3, 4), (4, 5), (5, '</s>')]\n        >>> list(ngrams([1,2,3,4,5], 2, pad_left=True, left_pad_symbol='<s>'))\n        [('<s>', 1), (1, 2), (2, 3), (3, 4), (4, 5)]\n        >>> list(ngrams([1,2,3,4,5], 2, pad_left=True, pad_right=True, left_pad_symbol='<s>', right_pad_symbol='</s>'))\n        [('<s>', 1), (1, 2), (2, 3), (3, 4), (4, 5), (5, '</s>')]\n    :param sequence: the source data to be converted into ngrams\n    :type sequence: sequence or iter\n    :param n: the degree of the ngrams\n    :type n: int\n    :param pad_left: whether the ngrams should be left-padded\n    :type pad_left: bool\n    :param pad_right: whether the ngrams should be right-padded\n    :type pad_right: bool\n    :param left_pad_symbol: the symbol to use for left padding (default is None)\n    :type left_pad_symbol: any\n    :param right_pad_symbol: the symbol to use for right padding (default is None)\n    :type right_pad_symbol: any\n    :rtype: sequence or iter\n    \"\"\"\n    sequence = pad_sequence(\n        sequence, n, pad_left, pad_right, left_pad_symbol, right_pad_symbol\n    )\n\n    history = []\n    while n > 1:\n        # PEP 479, prevent RuntimeError from being raised when StopIteration bubbles out of generator\n        try:\n            next_item = next(sequence)\n        except StopIteration:\n            # no more data, terminate the generator\n            return\n        history.append(next_item)\n        n -= 1\n    for item in sequence:\n        history.append(item)\n        yield tuple(history)\n        del history[0]","source_hash":"264e39c7dfb7355617861fed173838a0d4434365662f86b9e9abc2a31b61c455","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.utils.pad_sequence","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.utils.pad_sequence#L10-L45","kind":"function","name":"pad_sequence","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/utils.py","language":"python","start_line":10,"end_line":45,"context_start_line":1,"context_end_line":65,"code":"# Natural Language Toolkit: Utility functions\n#\n# Copyright (C) 2001-2020 NLTK Project\n# Author: Steven Bird <stevenbird1@gmail.com>\n# URL: <http://nltk.org/>\n# For license information, see LICENSE.TXT\n\nfrom itertools import chain\n\ndef pad_sequence(\n    sequence,\n    n,\n    pad_left=False,\n    pad_right=False,\n    left_pad_symbol=None,\n    right_pad_symbol=None,\n):\n    \"\"\"\n    Returns a padded sequence of items before ngram extraction.\n        >>> list(pad_sequence([1,2,3,4,5], 2, pad_left=True, pad_right=True, left_pad_symbol='<s>', right_pad_symbol='</s>'))\n        ['<s>', 1, 2, 3, 4, 5, '</s>']\n        >>> list(pad_sequence([1,2,3,4,5], 2, pad_left=True, left_pad_symbol='<s>'))\n        ['<s>', 1, 2, 3, 4, 5]\n        >>> list(pad_sequence([1,2,3,4,5], 2, pad_right=True, right_pad_symbol='</s>'))\n        [1, 2, 3, 4, 5, '</s>']\n    :param sequence: the source data to be padded\n    :type sequence: sequence or iter\n    :param n: the degree of the ngrams\n    :type n: int\n    :param pad_left: whether the ngrams should be left-padded\n    :type pad_left: bool\n    :param pad_right: whether the ngrams should be right-padded\n    :type pad_right: bool\n    :param left_pad_symbol: the symbol to use for left padding (default is None)\n    :type left_pad_symbol: any\n    :param right_pad_symbol: the symbol to use for right padding (default is None)\n    :type right_pad_symbol: any\n    :rtype: sequence or iter\n    \"\"\"\n    sequence = iter(sequence)\n    if pad_left:\n        sequence = chain((left_pad_symbol,) * (n - 1), sequence)\n    if pad_right:\n        sequence = chain(sequence, (right_pad_symbol,) * (n - 1))\n    return sequence\n\n\n# add a flag to pad the sequence so we get peripheral ngrams?\n\n\ndef ngrams(\n    sequence,\n    n,\n    pad_left=False,\n    pad_right=False,\n    left_pad_symbol=None,\n    right_pad_symbol=None,\n):\n    \"\"\"\n    Return the ngrams generated from a sequence of items, as an iterator.\n    For example:\n        >>> from nltk.util import ngrams\n        >>> list(ngrams([1,2,3,4,5], 3))\n        [(1, 2, 3), (2, 3, 4), (3, 4, 5)]\n    Wrap with list for a list version of this function.  Set pad_left","source_hash":"264e39c7dfb7355617861fed173838a0d4434365662f86b9e9abc2a31b61c455","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.utils.ngrams","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.utils.ngrams#L51-L106","kind":"function","name":"ngrams","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/utils.py","language":"python","start_line":51,"end_line":106,"context_start_line":31,"context_end_line":106,"code":"    :type pad_left: bool\n    :param pad_right: whether the ngrams should be right-padded\n    :type pad_right: bool\n    :param left_pad_symbol: the symbol to use for left padding (default is None)\n    :type left_pad_symbol: any\n    :param right_pad_symbol: the symbol to use for right padding (default is None)\n    :type right_pad_symbol: any\n    :rtype: sequence or iter\n    \"\"\"\n    sequence = iter(sequence)\n    if pad_left:\n        sequence = chain((left_pad_symbol,) * (n - 1), sequence)\n    if pad_right:\n        sequence = chain(sequence, (right_pad_symbol,) * (n - 1))\n    return sequence\n\n\n# add a flag to pad the sequence so we get peripheral ngrams?\n\n\ndef ngrams(\n    sequence,\n    n,\n    pad_left=False,\n    pad_right=False,\n    left_pad_symbol=None,\n    right_pad_symbol=None,\n):\n    \"\"\"\n    Return the ngrams generated from a sequence of items, as an iterator.\n    For example:\n        >>> from nltk.util import ngrams\n        >>> list(ngrams([1,2,3,4,5], 3))\n        [(1, 2, 3), (2, 3, 4), (3, 4, 5)]\n    Wrap with list for a list version of this function.  Set pad_left\n    or pad_right to true in order to get additional ngrams:\n        >>> list(ngrams([1,2,3,4,5], 2, pad_right=True))\n        [(1, 2), (2, 3), (3, 4), (4, 5), (5, None)]\n        >>> list(ngrams([1,2,3,4,5], 2, pad_right=True, right_pad_symbol='</s>'))\n        [(1, 2), (2, 3), (3, 4), (4, 5), (5, '</s>')]\n        >>> list(ngrams([1,2,3,4,5], 2, pad_left=True, left_pad_symbol='<s>'))\n        [('<s>', 1), (1, 2), (2, 3), (3, 4), (4, 5)]\n        >>> list(ngrams([1,2,3,4,5], 2, pad_left=True, pad_right=True, left_pad_symbol='<s>', right_pad_symbol='</s>'))\n        [('<s>', 1), (1, 2), (2, 3), (3, 4), (4, 5), (5, '</s>')]\n    :param sequence: the source data to be converted into ngrams\n    :type sequence: sequence or iter\n    :param n: the degree of the ngrams\n    :type n: int\n    :param pad_left: whether the ngrams should be left-padded\n    :type pad_left: bool\n    :param pad_right: whether the ngrams should be right-padded\n    :type pad_right: bool\n    :param left_pad_symbol: the symbol to use for left padding (default is None)\n    :type left_pad_symbol: any\n    :param right_pad_symbol: the symbol to use for right padding (default is None)\n    :type right_pad_symbol: any\n    :rtype: sequence or iter\n    \"\"\"\n    sequence = pad_sequence(\n        sequence, n, pad_left, pad_right, left_pad_symbol, right_pad_symbol\n    )\n\n    history = []\n    while n > 1:\n        # PEP 479, prevent RuntimeError from being raised when StopIteration bubbles out of generator\n        try:\n            next_item = next(sequence)\n        except StopIteration:\n            # no more data, terminate the generator\n            return\n        history.append(next_item)\n        n -= 1\n    for item in sequence:\n        history.append(item)\n        yield tuple(history)\n        del history[0]","source_hash":"264e39c7dfb7355617861fed173838a0d4434365662f86b9e9abc2a31b61c455","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_match","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_match#L1-L97","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_match","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_match.py","language":"python","start_line":1,"end_line":97,"context_start_line":1,"context_end_line":97,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\ndef calc_syntax_match(references, candidate, lang):\n    return corpus_syntax_match([references], [candidate], lang)\n\ndef corpus_syntax_match(references, candidates, lang):   \n    '''\n    Language.build_library(# Store the library in the `build` directory\n        'build/my-languages.so',\n        'vendor/tree-sitter-python')\n    '''\n    # LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    LANGUAGE = Language('./build/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    match_count = 0\n    total_count = 0\n\n    for i in range(len(candidates)):\n        references_sample = references[i]\n        candidate = candidates[i] \n        for reference in references_sample:\n            try:\n                candidate=remove_comments_and_docstrings(candidate,'java')\n            except:\n                pass    \n            try:\n                reference=remove_comments_and_docstrings(reference,'java')\n            except:\n                pass  \n\n            candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n\n            reference_tree = parser.parse(bytes(reference,'utf8')).root_node\n            '''\n            print(candidate)\n            print(reference_tree.sexp())\n            print(candidate_tree.sexp())\n            print(reference_tree.sexp() == candidate_tree.sexp())\n            exit()\n            '''\n\n            '''\n            if candidate_tree.sexp() == reference_tree.sexp():\n                total_match = 1/0.3\n            else:\n                total_match = -1/0.3\n            '''\n            total_match = 0\n\n            def get_all_sub_trees(root_node):\n                node_stack = []\n                sub_tree_sexp_list = []\n                depth = 1\n                node_stack.append([root_node, depth])\n                while len(node_stack) != 0:\n                    cur_node, cur_depth = node_stack.pop()\n                    sub_tree_sexp_list.append([cur_node.sexp(), cur_depth])\n                    for child_node in cur_node.children:\n                        if len(child_node.children) != 0:\n                            depth = cur_depth + 1\n                            node_stack.append([child_node, depth])\n                return sub_tree_sexp_list\n            cand_sexps = [x[0] for x in get_all_sub_trees(candidate_tree)]\n            ref_sexps = get_all_sub_trees(reference_tree)\n\n            # print(cand_sexps)\n            # print(ref_sexps)\n            \n            for sub_tree, depth in ref_sexps:\n                if sub_tree in cand_sexps:\n                     match_count += 1\n            total_count += len(ref_sexps)  \n        \n    if total_count == 0:\n        return 0 + total_match\n\n    score = match_count / total_count + total_match\n    return score","source_hash":"81d088ac5225c2f00cbb49515483052cb3b2e032666654349f0c701caee43556","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_match.calc_syntax_match","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_match.calc_syntax_match#L21-L22","kind":"function","name":"calc_syntax_match","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_match.py","language":"python","start_line":21,"end_line":22,"context_start_line":1,"context_end_line":42,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\ndef calc_syntax_match(references, candidate, lang):\n    return corpus_syntax_match([references], [candidate], lang)\n\ndef corpus_syntax_match(references, candidates, lang):   \n    '''\n    Language.build_library(# Store the library in the `build` directory\n        'build/my-languages.so',\n        'vendor/tree-sitter-python')\n    '''\n    # LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    LANGUAGE = Language('./build/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    match_count = 0\n    total_count = 0\n\n    for i in range(len(candidates)):\n        references_sample = references[i]\n        candidate = candidates[i] \n        for reference in references_sample:\n            try:\n                candidate=remove_comments_and_docstrings(candidate,'java')","source_hash":"81d088ac5225c2f00cbb49515483052cb3b2e032666654349f0c701caee43556","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_match.corpus_syntax_match","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_match.corpus_syntax_match#L24-L97","kind":"function","name":"corpus_syntax_match","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_match.py","language":"python","start_line":24,"end_line":97,"context_start_line":4,"context_end_line":97,"code":"from codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\ndef calc_syntax_match(references, candidate, lang):\n    return corpus_syntax_match([references], [candidate], lang)\n\ndef corpus_syntax_match(references, candidates, lang):   \n    '''\n    Language.build_library(# Store the library in the `build` directory\n        'build/my-languages.so',\n        'vendor/tree-sitter-python')\n    '''\n    # LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    LANGUAGE = Language('./build/my-languages.so', lang)\n    parser = Parser()\n    parser.set_language(LANGUAGE)\n    match_count = 0\n    total_count = 0\n\n    for i in range(len(candidates)):\n        references_sample = references[i]\n        candidate = candidates[i] \n        for reference in references_sample:\n            try:\n                candidate=remove_comments_and_docstrings(candidate,'java')\n            except:\n                pass    \n            try:\n                reference=remove_comments_and_docstrings(reference,'java')\n            except:\n                pass  \n\n            candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n\n            reference_tree = parser.parse(bytes(reference,'utf8')).root_node\n            '''\n            print(candidate)\n            print(reference_tree.sexp())\n            print(candidate_tree.sexp())\n            print(reference_tree.sexp() == candidate_tree.sexp())\n            exit()\n            '''\n\n            '''\n            if candidate_tree.sexp() == reference_tree.sexp():\n                total_match = 1/0.3\n            else:\n                total_match = -1/0.3\n            '''\n            total_match = 0\n\n            def get_all_sub_trees(root_node):\n                node_stack = []\n                sub_tree_sexp_list = []\n                depth = 1\n                node_stack.append([root_node, depth])\n                while len(node_stack) != 0:\n                    cur_node, cur_depth = node_stack.pop()\n                    sub_tree_sexp_list.append([cur_node.sexp(), cur_depth])\n                    for child_node in cur_node.children:\n                        if len(child_node.children) != 0:\n                            depth = cur_depth + 1\n                            node_stack.append([child_node, depth])\n                return sub_tree_sexp_list\n            cand_sexps = [x[0] for x in get_all_sub_trees(candidate_tree)]\n            ref_sexps = get_all_sub_trees(reference_tree)\n\n            # print(cand_sexps)\n            # print(ref_sexps)\n            \n            for sub_tree, depth in ref_sexps:\n                if sub_tree in cand_sexps:\n                     match_count += 1\n            total_count += len(ref_sexps)  \n        \n    if total_count == 0:\n        return 0 + total_match\n\n    score = match_count / total_count + total_match\n    return score","source_hash":"81d088ac5225c2f00cbb49515483052cb3b2e032666654349f0c701caee43556","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_match.get_all_sub_trees","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.syntax_match.get_all_sub_trees#L69-L81","kind":"function","name":"get_all_sub_trees","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_match.py","language":"python","start_line":69,"end_line":81,"context_start_line":49,"context_end_line":97,"code":"\n            candidate_tree = parser.parse(bytes(candidate,'utf8')).root_node\n\n            reference_tree = parser.parse(bytes(reference,'utf8')).root_node\n            '''\n            print(candidate)\n            print(reference_tree.sexp())\n            print(candidate_tree.sexp())\n            print(reference_tree.sexp() == candidate_tree.sexp())\n            exit()\n            '''\n\n            '''\n            if candidate_tree.sexp() == reference_tree.sexp():\n                total_match = 1/0.3\n            else:\n                total_match = -1/0.3\n            '''\n            total_match = 0\n\n            def get_all_sub_trees(root_node):\n                node_stack = []\n                sub_tree_sexp_list = []\n                depth = 1\n                node_stack.append([root_node, depth])\n                while len(node_stack) != 0:\n                    cur_node, cur_depth = node_stack.pop()\n                    sub_tree_sexp_list.append([cur_node.sexp(), cur_depth])\n                    for child_node in cur_node.children:\n                        if len(child_node.children) != 0:\n                            depth = cur_depth + 1\n                            node_stack.append([child_node, depth])\n                return sub_tree_sexp_list\n            cand_sexps = [x[0] for x in get_all_sub_trees(candidate_tree)]\n            ref_sexps = get_all_sub_trees(reference_tree)\n\n            # print(cand_sexps)\n            # print(ref_sexps)\n            \n            for sub_tree, depth in ref_sexps:\n                if sub_tree in cand_sexps:\n                     match_count += 1\n            total_count += len(ref_sexps)  \n        \n    if total_count == 0:\n        return 0 + total_match\n\n    score = match_count / total_count + total_match\n    return score","source_hash":"81d088ac5225c2f00cbb49515483052cb3b2e032666654349f0c701caee43556","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu#L1-L590","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":1,"end_line":590,"context_start_line":1,"context_end_line":590,"code":"# -*- coding: utf-8 -*-\n# Natural Language Toolkit: BLEU Score\n#\n# Copyright (C) 2001-2020 NLTK Project\n# Authors: Chin Yee Lee, Hengfeng Li, Ruxin Hou, Calvin Tanujaya Lim\n# Contributors: Björn Mattsson, Dmitrijs Milajevs, Liling Tan\n# URL: <http://nltk.org/>\n# For license information, see LICENSE.TXT\n\n\"\"\"BLEU score implementation.\"\"\"\n\nimport math\nimport sys\nfrom fractions import Fraction\nimport warnings\nfrom collections import Counter\n\nfrom codebleu.utils import ngrams\nimport pdb\n\n\ndef sentence_bleu(\n    references,\n    hypothesis,\n    weights=(0.25, 0.25, 0.25, 0.25),\n    smoothing_function=None,\n    auto_reweigh=False,\n):\n    \"\"\"\n    Calculate BLEU score (Bilingual Evaluation Understudy) from\n    Papineni, Kishore, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002.\n    \"BLEU: a method for automatic evaluation of machine translation.\"\n    In Proceedings of ACL. http://www.aclweb.org/anthology/P02-1040.pdf\n    >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n    ...               'ensures', 'that', 'the', 'military', 'always',\n    ...               'obeys', 'the', 'commands', 'of', 'the', 'party']\n    >>> hypothesis2 = ['It', 'is', 'to', 'insure', 'the', 'troops',\n    ...               'forever', 'hearing', 'the', 'activity', 'guidebook',\n    ...               'that', 'party', 'direct']\n    >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n    ...               'ensures', 'that', 'the', 'military', 'will', 'forever',\n    ...               'heed', 'Party', 'commands']\n    >>> reference2 = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n    ...               'guarantees', 'the', 'military', 'forces', 'always',\n    ...               'being', 'under', 'the', 'command', 'of', 'the',\n    ...               'Party']\n    >>> reference3 = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n    ...               'army', 'always', 'to', 'heed', 'the', 'directions',\n    ...               'of', 'the', 'party']\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis1) # doctest: +ELLIPSIS\n    0.5045...\n    If there is no ngrams overlap for any order of n-grams, BLEU returns the\n    value 0. This is because the precision for the order of n-grams without\n    overlap is 0, and the geometric mean in the final BLEU score computation\n    multiplies the 0 with the precision of other n-grams. This results in 0\n    (independently of the precision of the othe n-gram orders). The following\n    example has zero 3-gram and 4-gram overlaps:\n    >>> round(sentence_bleu([reference1, reference2, reference3], hypothesis2),4) # doctest: +ELLIPSIS\n    0.0\n    To avoid this harsh behaviour when no ngram overlaps are found a smoothing\n    function can be used.\n    >>> chencherry = SmoothingFunction()\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis2,\n    ...     smoothing_function=chencherry.method1) # doctest: +ELLIPSIS\n    0.0370...\n    The default BLEU calculates a score for up to 4-grams using uniform\n    weights (this is called BLEU-4). To evaluate your translations with\n    higher/lower order ngrams, use customized weights. E.g. when accounting\n    for up to 5-grams with uniform weights (this is called BLEU-5) use:\n    >>> weights = (1./5., 1./5., 1./5., 1./5., 1./5.)\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis1, weights) # doctest: +ELLIPSIS\n    0.3920...\n    :param references: reference sentences\n    :type references: list(list(str))\n    :param hypothesis: a hypothesis sentence\n    :type hypothesis: list(str)\n    :param weights: weights for unigrams, bigrams, trigrams and so on\n    :type weights: list(float)\n    :param smoothing_function:\n    :type smoothing_function: SmoothingFunction\n    :param auto_reweigh: Option to re-normalize the weights uniformly.\n    :type auto_reweigh: bool\n    :return: The sentence-level BLEU score.\n    :rtype: float\n    \"\"\"\n    return corpus_bleu(\n        [references], [hypothesis], weights, smoothing_function, auto_reweigh\n    )\n\n\ndef corpus_bleu(\n    list_of_references,\n    hypotheses,\n    weights=(0.25, 0.25, 0.25, 0.25),\n    smoothing_function=None,\n    auto_reweigh=False,\n):\n    \"\"\"\n    Calculate a single corpus-level BLEU score (aka. system-level BLEU) for all\n    the hypotheses and their respective references.\n    Instead of averaging the sentence level BLEU scores (i.e. marco-average\n    precision), the original BLEU metric (Papineni et al. 2002) accounts for\n    the micro-average precision (i.e. summing the numerators and denominators\n    for each hypothesis-reference(s) pairs before the division).\n    >>> hyp1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n    ...         'ensures', 'that', 'the', 'military', 'always',\n    ...         'obeys', 'the', 'commands', 'of', 'the', 'party']\n    >>> ref1a = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n    ...          'ensures', 'that', 'the', 'military', 'will', 'forever',\n    ...          'heed', 'Party', 'commands']\n    >>> ref1b = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n    ...          'guarantees', 'the', 'military', 'forces', 'always',\n    ...          'being', 'under', 'the', 'command', 'of', 'the', 'Party']\n    >>> ref1c = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n    ...          'army', 'always', 'to', 'heed', 'the', 'directions',\n    ...          'of', 'the', 'party']\n    >>> hyp2 = ['he', 'read', 'the', 'book', 'because', 'he', 'was',\n    ...         'interested', 'in', 'world', 'history']\n    >>> ref2a = ['he', 'was', 'interested', 'in', 'world', 'history',\n    ...          'because', 'he', 'read', 'the', 'book']\n    >>> list_of_references = [[ref1a, ref1b, ref1c], [ref2a]]\n    >>> hypotheses = [hyp1, hyp2]\n    >>> corpus_bleu(list_of_references, hypotheses) # doctest: +ELLIPSIS\n    0.5920...\n    The example below show that corpus_bleu() is different from averaging\n    sentence_bleu() for hypotheses\n    >>> score1 = sentence_bleu([ref1a, ref1b, ref1c], hyp1)\n    >>> score2 = sentence_bleu([ref2a], hyp2)\n    >>> (score1 + score2) / 2 # doctest: +ELLIPSIS\n    0.6223...\n    :param list_of_references: a corpus of lists of reference sentences, w.r.t. hypotheses\n    :type list_of_references: list(list(list(str)))\n    :param hypotheses: a list of hypothesis sentences\n    :type hypotheses: list(list(str))\n    :param weights: weights for unigrams, bigrams, trigrams and so on\n    :type weights: list(float)\n    :param smoothing_function:\n    :type smoothing_function: SmoothingFunction\n    :param auto_reweigh: Option to re-normalize the weights uniformly.\n    :type auto_reweigh: bool\n    :return: The corpus-level BLEU score.\n    :rtype: float\n    \"\"\"\n    # Before proceeding to compute BLEU, perform sanity checks.\n\n    p_numerators = Counter()  # Key = ngram order, and value = no. of ngram matches.\n    p_denominators = Counter()  # Key = ngram order, and value = no. of ngram in ref.\n    hyp_lengths, ref_lengths = 0, 0\n\n    assert len(list_of_references) == len(hypotheses), (\n        \"The number of hypotheses and their reference(s) should be the \" \"same \"\n    )\n\n    # Iterate through each hypothesis and their corresponding references.\n    for references, hypothesis in zip(list_of_references, hypotheses):\n        # For each order of ngram, calculate the numerator and\n        # denominator for the corpus-level modified precision.\n        for i, _ in enumerate(weights, start=1):\n            p_i = modified_precision(references, hypothesis, i)\n            p_numerators[i] += p_i.numerator\n            p_denominators[i] += p_i.denominator\n\n        # Calculate the hypothesis length and the closest reference length.\n        # Adds them to the corpus-level hypothesis and reference counts.\n        hyp_len = len(hypothesis)\n        hyp_lengths += hyp_len\n        ref_lengths += closest_ref_length(references, hyp_len)\n\n    # Calculate corpus-level brevity penalty.\n    bp = brevity_penalty(ref_lengths, hyp_lengths)\n\n    # Uniformly re-weighting based on maximum hypothesis lengths if largest\n    # order of n-grams < 4 and weights is set at default.\n    if auto_reweigh:\n        if hyp_lengths < 4 and weights == (0.25, 0.25, 0.25, 0.25):\n            weights = (1 / hyp_lengths,) * hyp_lengths\n\n    # Collects the various precision values for the different ngram orders.\n    p_n = [\n        Fraction(p_numerators[i], p_denominators[i], _normalize=False)\n        for i, _ in enumerate(weights, start=1)\n    ]\n\n    # Returns 0 if there's no matching n-grams\n    # We only need to check for p_numerators[1] == 0, since if there's\n    # no unigrams, there won't be any higher order ngrams.\n    if p_numerators[1] == 0:\n        return 0\n\n    # If there's no smoothing, set use method0 from SmoothinFunction class.\n    if not smoothing_function:\n        smoothing_function = SmoothingFunction().method1\n    # Smoothen the modified precision.\n    # Note: smoothing_function() may convert values into floats;\n    #       it tries to retain the Fraction object as much as the\n    #       smoothing method allows.\n    p_n = smoothing_function(\n        p_n, references=references, hypothesis=hypothesis, hyp_len=hyp_lengths\n    )\n    s = (w_i * math.log(p_i) for w_i, p_i in zip(weights, p_n))\n    s = bp * math.exp(math.fsum(s))\n    return s\n\n\ndef modified_precision(references, hypothesis, n):\n    \"\"\"\n    Calculate modified ngram precision.\n    The normal precision method may lead to some wrong translations with\n    high-precision, e.g., the translation, in which a word of reference\n    repeats several times, has very high precision.\n    This function only returns the Fraction object that contains the numerator\n    and denominator necessary to calculate the corpus-level precision.\n    To calculate the modified precision for a single pair of hypothesis and\n    references, cast the Fraction object into a float.\n    The famous \"the the the ... \" example shows that you can get BLEU precision\n    by duplicating high frequency words.\n        >>> reference1 = 'the cat is on the mat'.split()\n        >>> reference2 = 'there is a cat on the mat'.split()\n        >>> hypothesis1 = 'the the the the the the the'.split()\n        >>> references = [reference1, reference2]\n        >>> float(modified_precision(references, hypothesis1, n=1)) # doctest: +ELLIPSIS\n        0.2857...\n    In the modified n-gram precision, a reference word will be considered\n    exhausted after a matching hypothesis word is identified, e.g.\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n        ...               'ensures', 'that', 'the', 'military', 'will',\n        ...               'forever', 'heed', 'Party', 'commands']\n        >>> reference2 = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n        ...               'guarantees', 'the', 'military', 'forces', 'always',\n        ...               'being', 'under', 'the', 'command', 'of', 'the',\n        ...               'Party']\n        >>> reference3 = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n        ...               'army', 'always', 'to', 'heed', 'the', 'directions',\n        ...               'of', 'the', 'party']\n        >>> hypothesis = 'of the'.split()\n        >>> references = [reference1, reference2, reference3]\n        >>> float(modified_precision(references, hypothesis, n=1))\n        1.0\n        >>> float(modified_precision(references, hypothesis, n=2))\n        1.0\n    An example of a normal machine translation hypothesis:\n        >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n        ...               'ensures', 'that', 'the', 'military', 'always',\n        ...               'obeys', 'the', 'commands', 'of', 'the', 'party']\n        >>> hypothesis2 = ['It', 'is', 'to', 'insure', 'the', 'troops',\n        ...               'forever', 'hearing', 'the', 'activity', 'guidebook',\n        ...               'that', 'party', 'direct']\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n        ...               'ensures', 'that', 'the', 'military', 'will',\n        ...               'forever', 'heed', 'Party', 'commands']\n        >>> reference2 = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n        ...               'guarantees', 'the', 'military', 'forces', 'always',\n        ...               'being', 'under', 'the', 'command', 'of', 'the',\n        ...               'Party']\n        >>> reference3 = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n        ...               'army', 'always', 'to', 'heed', 'the', 'directions',\n        ...               'of', 'the', 'party']\n        >>> references = [reference1, reference2, reference3]\n        >>> float(modified_precision(references, hypothesis1, n=1)) # doctest: +ELLIPSIS\n        0.9444...\n        >>> float(modified_precision(references, hypothesis2, n=1)) # doctest: +ELLIPSIS\n        0.5714...\n        >>> float(modified_precision(references, hypothesis1, n=2)) # doctest: +ELLIPSIS\n        0.5882352941176471\n        >>> float(modified_precision(references, hypothesis2, n=2)) # doctest: +ELLIPSIS\n        0.07692...\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hypothesis: A hypothesis translation.\n    :type hypothesis: list(str)\n    :param n: The ngram order.\n    :type n: int\n    :return: BLEU's modified precision for the nth order ngram.\n    :rtype: Fraction\n    \"\"\"\n    # Extracts all ngrams in hypothesis\n    # Set an empty Counter if hypothesis is empty.\n\n    counts = Counter(ngrams(hypothesis, n)) if len(hypothesis) >= n else Counter()\n    # Extract a union of references' counts.\n    # max_counts = reduce(or_, [Counter(ngrams(ref, n)) for ref in references])\n    max_counts = {}\n    for reference in references:\n        reference_counts = (\n            Counter(ngrams(reference, n)) if len(reference) >= n else Counter()\n        )\n        for ngram in counts:\n            max_counts[ngram] = max(max_counts.get(ngram, 0), reference_counts[ngram])\n\n    # Assigns the intersection between hypothesis and references' counts.\n    clipped_counts = {\n        ngram: min(count, max_counts[ngram]) for ngram, count in counts.items()\n    }\n\n    numerator = sum(clipped_counts.values())\n    # Ensures that denominator is minimum 1 to avoid ZeroDivisionError.\n    # Usually this happens when the ngram order is > len(reference).\n    denominator = max(1, sum(counts.values()))\n\n    return Fraction(numerator, denominator, _normalize=False)\n\n\ndef closest_ref_length(references, hyp_len):\n    \"\"\"\n    This function finds the reference that is the closest length to the\n    hypothesis. The closest reference length is referred to as *r* variable\n    from the brevity penalty formula in Papineni et. al. (2002)\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hyp_len: The length of the hypothesis.\n    :type hyp_len: int\n    :return: The length of the reference that's closest to the hypothesis.\n    :rtype: int\n    \"\"\"\n    ref_lens = (len(reference) for reference in references)\n    closest_ref_len = min(\n        ref_lens, key=lambda ref_len: (abs(ref_len - hyp_len), ref_len)\n    )\n    return closest_ref_len\n\n\ndef brevity_penalty(closest_ref_len, hyp_len):\n    \"\"\"\n    Calculate brevity penalty.\n    As the modified n-gram precision still has the problem from the short\n    length sentence, brevity penalty is used to modify the overall BLEU\n    score according to length.\n    An example from the paper. There are three references with length 12, 15\n    and 17. And a concise hypothesis of the length 12. The brevity penalty is 1.\n        >>> reference1 = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> reference2 = list('aaaaaaaaaaaaaaa')   # i.e. ['a'] * 15\n        >>> reference3 = list('aaaaaaaaaaaaaaaaa') # i.e. ['a'] * 17\n        >>> hypothesis = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> references = [reference1, reference2, reference3]\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    In case a hypothesis translation is shorter than the references, penalty is\n    applied.\n        >>> references = [['a'] * 28, ['a'] * 28]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        0.2635971381157267\n    The length of the closest reference is used to compute the penalty. If the\n    length of a hypothesis is 12, and the reference lengths are 13 and 2, the\n    penalty is applied because the hypothesis length (12) is less then the\n    closest reference length (13).\n        >>> references = [['a'] * 13, ['a'] * 2]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len) # doctest: +ELLIPSIS\n        0.9200...\n    The brevity penalty doesn't depend on reference order. More importantly,\n    when two reference sentences are at the same distance, the shortest\n    reference sentence length is used.\n        >>> references = [['a'] * 13, ['a'] * 11]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> bp1 = brevity_penalty(closest_ref_len, hyp_len)\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(reversed(references), hyp_len)\n        >>> bp2 = brevity_penalty(closest_ref_len, hyp_len)\n        >>> bp1 == bp2 == 1\n        True\n    A test example from mteval-v13a.pl (starting from the line 705):\n        >>> references = [['a'] * 11, ['a'] * 8]\n        >>> hypothesis = ['a'] * 7\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len) # doctest: +ELLIPSIS\n        0.8668...\n        >>> references = [['a'] * 11, ['a'] * 8, ['a'] * 6, ['a'] * 7]\n        >>> hypothesis = ['a'] * 7\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    :param hyp_len: The length of the hypothesis for a single sentence OR the\n    sum of all the hypotheses' lengths for a corpus\n    :type hyp_len: int\n    :param closest_ref_len: The length of the closest reference for a single\n    hypothesis OR the sum of all the closest references for every hypotheses.\n    :type closest_ref_len: int\n    :return: BLEU's brevity penalty.\n    :rtype: float\n    \"\"\"\n    if hyp_len > closest_ref_len:\n        return 1\n    # If hypothesis is empty, brevity penalty = 0 should result in BLEU = 0.0\n    elif hyp_len == 0:\n        return 0\n    else:\n        return math.exp(1 - closest_ref_len / hyp_len)\n\n\nclass SmoothingFunction:\n    \"\"\"\n    This is an implementation of the smoothing techniques\n    for segment-level BLEU scores that was presented in\n    Boxing Chen and Collin Cherry (2014) A Systematic Comparison of\n    Smoothing Techniques for Sentence-Level BLEU. In WMT14.\n    http://acl2014.org/acl2014/W14-33/pdf/W14-3346.pdf\n    \"\"\"\n\n    def __init__(self, epsilon=0.1, alpha=5, k=5):\n        \"\"\"\n        This will initialize the parameters required for the various smoothing\n        techniques, the default values are set to the numbers used in the\n        experiments from Chen and Cherry (2014).\n        >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which', 'ensures',\n        ...                 'that', 'the', 'military', 'always', 'obeys', 'the',\n        ...                 'commands', 'of', 'the', 'party']\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that', 'ensures',\n        ...               'that', 'the', 'military', 'will', 'forever', 'heed',\n# ... truncated ...","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":true}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.sentence_bleu","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.sentence_bleu#L22-L88","kind":"function","name":"sentence_bleu","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":22,"end_line":88,"context_start_line":2,"context_end_line":108,"code":"# Natural Language Toolkit: BLEU Score\n#\n# Copyright (C) 2001-2020 NLTK Project\n# Authors: Chin Yee Lee, Hengfeng Li, Ruxin Hou, Calvin Tanujaya Lim\n# Contributors: Björn Mattsson, Dmitrijs Milajevs, Liling Tan\n# URL: <http://nltk.org/>\n# For license information, see LICENSE.TXT\n\n\"\"\"BLEU score implementation.\"\"\"\n\nimport math\nimport sys\nfrom fractions import Fraction\nimport warnings\nfrom collections import Counter\n\nfrom codebleu.utils import ngrams\nimport pdb\n\n\ndef sentence_bleu(\n    references,\n    hypothesis,\n    weights=(0.25, 0.25, 0.25, 0.25),\n    smoothing_function=None,\n    auto_reweigh=False,\n):\n    \"\"\"\n    Calculate BLEU score (Bilingual Evaluation Understudy) from\n    Papineni, Kishore, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002.\n    \"BLEU: a method for automatic evaluation of machine translation.\"\n    In Proceedings of ACL. http://www.aclweb.org/anthology/P02-1040.pdf\n    >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n    ...               'ensures', 'that', 'the', 'military', 'always',\n    ...               'obeys', 'the', 'commands', 'of', 'the', 'party']\n    >>> hypothesis2 = ['It', 'is', 'to', 'insure', 'the', 'troops',\n    ...               'forever', 'hearing', 'the', 'activity', 'guidebook',\n    ...               'that', 'party', 'direct']\n    >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n    ...               'ensures', 'that', 'the', 'military', 'will', 'forever',\n    ...               'heed', 'Party', 'commands']\n    >>> reference2 = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n    ...               'guarantees', 'the', 'military', 'forces', 'always',\n    ...               'being', 'under', 'the', 'command', 'of', 'the',\n    ...               'Party']\n    >>> reference3 = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n    ...               'army', 'always', 'to', 'heed', 'the', 'directions',\n    ...               'of', 'the', 'party']\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis1) # doctest: +ELLIPSIS\n    0.5045...\n    If there is no ngrams overlap for any order of n-grams, BLEU returns the\n    value 0. This is because the precision for the order of n-grams without\n    overlap is 0, and the geometric mean in the final BLEU score computation\n    multiplies the 0 with the precision of other n-grams. This results in 0\n    (independently of the precision of the othe n-gram orders). The following\n    example has zero 3-gram and 4-gram overlaps:\n    >>> round(sentence_bleu([reference1, reference2, reference3], hypothesis2),4) # doctest: +ELLIPSIS\n    0.0\n    To avoid this harsh behaviour when no ngram overlaps are found a smoothing\n    function can be used.\n    >>> chencherry = SmoothingFunction()\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis2,\n    ...     smoothing_function=chencherry.method1) # doctest: +ELLIPSIS\n    0.0370...\n    The default BLEU calculates a score for up to 4-grams using uniform\n    weights (this is called BLEU-4). To evaluate your translations with\n    higher/lower order ngrams, use customized weights. E.g. when accounting\n    for up to 5-grams with uniform weights (this is called BLEU-5) use:\n    >>> weights = (1./5., 1./5., 1./5., 1./5., 1./5.)\n    >>> sentence_bleu([reference1, reference2, reference3], hypothesis1, weights) # doctest: +ELLIPSIS\n    0.3920...\n    :param references: reference sentences\n    :type references: list(list(str))\n    :param hypothesis: a hypothesis sentence\n    :type hypothesis: list(str)\n    :param weights: weights for unigrams, bigrams, trigrams and so on\n    :type weights: list(float)\n    :param smoothing_function:\n    :type smoothing_function: SmoothingFunction\n    :param auto_reweigh: Option to re-normalize the weights uniformly.\n    :type auto_reweigh: bool\n    :return: The sentence-level BLEU score.\n    :rtype: float\n    \"\"\"\n    return corpus_bleu(\n        [references], [hypothesis], weights, smoothing_function, auto_reweigh\n    )\n\n\ndef corpus_bleu(\n    list_of_references,\n    hypotheses,\n    weights=(0.25, 0.25, 0.25, 0.25),\n    smoothing_function=None,\n    auto_reweigh=False,\n):\n    \"\"\"\n    Calculate a single corpus-level BLEU score (aka. system-level BLEU) for all\n    the hypotheses and their respective references.\n    Instead of averaging the sentence level BLEU scores (i.e. marco-average\n    precision), the original BLEU metric (Papineni et al. 2002) accounts for\n    the micro-average precision (i.e. summing the numerators and denominators\n    for each hypothesis-reference(s) pairs before the division).\n    >>> hyp1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n    ...         'ensures', 'that', 'the', 'military', 'always',\n    ...         'obeys', 'the', 'commands', 'of', 'the', 'party']\n    >>> ref1a = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.corpus_bleu","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.corpus_bleu#L91-L202","kind":"function","name":"corpus_bleu","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":91,"end_line":202,"context_start_line":71,"context_end_line":222,"code":"    >>> sentence_bleu([reference1, reference2, reference3], hypothesis1, weights) # doctest: +ELLIPSIS\n    0.3920...\n    :param references: reference sentences\n    :type references: list(list(str))\n    :param hypothesis: a hypothesis sentence\n    :type hypothesis: list(str)\n    :param weights: weights for unigrams, bigrams, trigrams and so on\n    :type weights: list(float)\n    :param smoothing_function:\n    :type smoothing_function: SmoothingFunction\n    :param auto_reweigh: Option to re-normalize the weights uniformly.\n    :type auto_reweigh: bool\n    :return: The sentence-level BLEU score.\n    :rtype: float\n    \"\"\"\n    return corpus_bleu(\n        [references], [hypothesis], weights, smoothing_function, auto_reweigh\n    )\n\n\ndef corpus_bleu(\n    list_of_references,\n    hypotheses,\n    weights=(0.25, 0.25, 0.25, 0.25),\n    smoothing_function=None,\n    auto_reweigh=False,\n):\n    \"\"\"\n    Calculate a single corpus-level BLEU score (aka. system-level BLEU) for all\n    the hypotheses and their respective references.\n    Instead of averaging the sentence level BLEU scores (i.e. marco-average\n    precision), the original BLEU metric (Papineni et al. 2002) accounts for\n    the micro-average precision (i.e. summing the numerators and denominators\n    for each hypothesis-reference(s) pairs before the division).\n    >>> hyp1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n    ...         'ensures', 'that', 'the', 'military', 'always',\n    ...         'obeys', 'the', 'commands', 'of', 'the', 'party']\n    >>> ref1a = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n    ...          'ensures', 'that', 'the', 'military', 'will', 'forever',\n    ...          'heed', 'Party', 'commands']\n    >>> ref1b = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n    ...          'guarantees', 'the', 'military', 'forces', 'always',\n    ...          'being', 'under', 'the', 'command', 'of', 'the', 'Party']\n    >>> ref1c = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n    ...          'army', 'always', 'to', 'heed', 'the', 'directions',\n    ...          'of', 'the', 'party']\n    >>> hyp2 = ['he', 'read', 'the', 'book', 'because', 'he', 'was',\n    ...         'interested', 'in', 'world', 'history']\n    >>> ref2a = ['he', 'was', 'interested', 'in', 'world', 'history',\n    ...          'because', 'he', 'read', 'the', 'book']\n    >>> list_of_references = [[ref1a, ref1b, ref1c], [ref2a]]\n    >>> hypotheses = [hyp1, hyp2]\n    >>> corpus_bleu(list_of_references, hypotheses) # doctest: +ELLIPSIS\n    0.5920...\n    The example below show that corpus_bleu() is different from averaging\n    sentence_bleu() for hypotheses\n    >>> score1 = sentence_bleu([ref1a, ref1b, ref1c], hyp1)\n    >>> score2 = sentence_bleu([ref2a], hyp2)\n    >>> (score1 + score2) / 2 # doctest: +ELLIPSIS\n    0.6223...\n    :param list_of_references: a corpus of lists of reference sentences, w.r.t. hypotheses\n    :type list_of_references: list(list(list(str)))\n    :param hypotheses: a list of hypothesis sentences\n    :type hypotheses: list(list(str))\n    :param weights: weights for unigrams, bigrams, trigrams and so on\n    :type weights: list(float)\n    :param smoothing_function:\n    :type smoothing_function: SmoothingFunction\n    :param auto_reweigh: Option to re-normalize the weights uniformly.\n    :type auto_reweigh: bool\n    :return: The corpus-level BLEU score.\n    :rtype: float\n    \"\"\"\n    # Before proceeding to compute BLEU, perform sanity checks.\n\n    p_numerators = Counter()  # Key = ngram order, and value = no. of ngram matches.\n    p_denominators = Counter()  # Key = ngram order, and value = no. of ngram in ref.\n    hyp_lengths, ref_lengths = 0, 0\n\n    assert len(list_of_references) == len(hypotheses), (\n        \"The number of hypotheses and their reference(s) should be the \" \"same \"\n    )\n\n    # Iterate through each hypothesis and their corresponding references.\n    for references, hypothesis in zip(list_of_references, hypotheses):\n        # For each order of ngram, calculate the numerator and\n        # denominator for the corpus-level modified precision.\n        for i, _ in enumerate(weights, start=1):\n            p_i = modified_precision(references, hypothesis, i)\n            p_numerators[i] += p_i.numerator\n            p_denominators[i] += p_i.denominator\n\n        # Calculate the hypothesis length and the closest reference length.\n        # Adds them to the corpus-level hypothesis and reference counts.\n        hyp_len = len(hypothesis)\n        hyp_lengths += hyp_len\n        ref_lengths += closest_ref_length(references, hyp_len)\n\n    # Calculate corpus-level brevity penalty.\n    bp = brevity_penalty(ref_lengths, hyp_lengths)\n\n    # Uniformly re-weighting based on maximum hypothesis lengths if largest\n    # order of n-grams < 4 and weights is set at default.\n    if auto_reweigh:\n        if hyp_lengths < 4 and weights == (0.25, 0.25, 0.25, 0.25):\n            weights = (1 / hyp_lengths,) * hyp_lengths\n\n    # Collects the various precision values for the different ngram orders.\n    p_n = [\n        Fraction(p_numerators[i], p_denominators[i], _normalize=False)\n        for i, _ in enumerate(weights, start=1)\n    ]\n\n    # Returns 0 if there's no matching n-grams\n    # We only need to check for p_numerators[1] == 0, since if there's\n    # no unigrams, there won't be any higher order ngrams.\n    if p_numerators[1] == 0:\n        return 0\n\n    # If there's no smoothing, set use method0 from SmoothinFunction class.\n    if not smoothing_function:\n        smoothing_function = SmoothingFunction().method1\n    # Smoothen the modified precision.\n    # Note: smoothing_function() may convert values into floats;\n    #       it tries to retain the Fraction object as much as the\n    #       smoothing method allows.\n    p_n = smoothing_function(\n        p_n, references=references, hypothesis=hypothesis, hyp_len=hyp_lengths\n    )\n    s = (w_i * math.log(p_i) for w_i, p_i in zip(weights, p_n))\n    s = bp * math.exp(math.fsum(s))\n    return s\n\n\ndef modified_precision(references, hypothesis, n):\n    \"\"\"\n    Calculate modified ngram precision.\n    The normal precision method may lead to some wrong translations with\n    high-precision, e.g., the translation, in which a word of reference\n    repeats several times, has very high precision.\n    This function only returns the Fraction object that contains the numerator\n    and denominator necessary to calculate the corpus-level precision.\n    To calculate the modified precision for a single pair of hypothesis and\n    references, cast the Fraction object into a float.\n    The famous \"the the the ... \" example shows that you can get BLEU precision\n    by duplicating high frequency words.\n        >>> reference1 = 'the cat is on the mat'.split()\n        >>> reference2 = 'there is a cat on the mat'.split()\n        >>> hypothesis1 = 'the the the the the the the'.split()\n        >>> references = [reference1, reference2]\n        >>> float(modified_precision(references, hypothesis1, n=1)) # doctest: +ELLIPSIS\n        0.2857...","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.modified_precision","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.modified_precision#L205-L300","kind":"function","name":"modified_precision","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":205,"end_line":300,"context_start_line":185,"context_end_line":320,"code":"    # We only need to check for p_numerators[1] == 0, since if there's\n    # no unigrams, there won't be any higher order ngrams.\n    if p_numerators[1] == 0:\n        return 0\n\n    # If there's no smoothing, set use method0 from SmoothinFunction class.\n    if not smoothing_function:\n        smoothing_function = SmoothingFunction().method1\n    # Smoothen the modified precision.\n    # Note: smoothing_function() may convert values into floats;\n    #       it tries to retain the Fraction object as much as the\n    #       smoothing method allows.\n    p_n = smoothing_function(\n        p_n, references=references, hypothesis=hypothesis, hyp_len=hyp_lengths\n    )\n    s = (w_i * math.log(p_i) for w_i, p_i in zip(weights, p_n))\n    s = bp * math.exp(math.fsum(s))\n    return s\n\n\ndef modified_precision(references, hypothesis, n):\n    \"\"\"\n    Calculate modified ngram precision.\n    The normal precision method may lead to some wrong translations with\n    high-precision, e.g., the translation, in which a word of reference\n    repeats several times, has very high precision.\n    This function only returns the Fraction object that contains the numerator\n    and denominator necessary to calculate the corpus-level precision.\n    To calculate the modified precision for a single pair of hypothesis and\n    references, cast the Fraction object into a float.\n    The famous \"the the the ... \" example shows that you can get BLEU precision\n    by duplicating high frequency words.\n        >>> reference1 = 'the cat is on the mat'.split()\n        >>> reference2 = 'there is a cat on the mat'.split()\n        >>> hypothesis1 = 'the the the the the the the'.split()\n        >>> references = [reference1, reference2]\n        >>> float(modified_precision(references, hypothesis1, n=1)) # doctest: +ELLIPSIS\n        0.2857...\n    In the modified n-gram precision, a reference word will be considered\n    exhausted after a matching hypothesis word is identified, e.g.\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n        ...               'ensures', 'that', 'the', 'military', 'will',\n        ...               'forever', 'heed', 'Party', 'commands']\n        >>> reference2 = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n        ...               'guarantees', 'the', 'military', 'forces', 'always',\n        ...               'being', 'under', 'the', 'command', 'of', 'the',\n        ...               'Party']\n        >>> reference3 = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n        ...               'army', 'always', 'to', 'heed', 'the', 'directions',\n        ...               'of', 'the', 'party']\n        >>> hypothesis = 'of the'.split()\n        >>> references = [reference1, reference2, reference3]\n        >>> float(modified_precision(references, hypothesis, n=1))\n        1.0\n        >>> float(modified_precision(references, hypothesis, n=2))\n        1.0\n    An example of a normal machine translation hypothesis:\n        >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which',\n        ...               'ensures', 'that', 'the', 'military', 'always',\n        ...               'obeys', 'the', 'commands', 'of', 'the', 'party']\n        >>> hypothesis2 = ['It', 'is', 'to', 'insure', 'the', 'troops',\n        ...               'forever', 'hearing', 'the', 'activity', 'guidebook',\n        ...               'that', 'party', 'direct']\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that',\n        ...               'ensures', 'that', 'the', 'military', 'will',\n        ...               'forever', 'heed', 'Party', 'commands']\n        >>> reference2 = ['It', 'is', 'the', 'guiding', 'principle', 'which',\n        ...               'guarantees', 'the', 'military', 'forces', 'always',\n        ...               'being', 'under', 'the', 'command', 'of', 'the',\n        ...               'Party']\n        >>> reference3 = ['It', 'is', 'the', 'practical', 'guide', 'for', 'the',\n        ...               'army', 'always', 'to', 'heed', 'the', 'directions',\n        ...               'of', 'the', 'party']\n        >>> references = [reference1, reference2, reference3]\n        >>> float(modified_precision(references, hypothesis1, n=1)) # doctest: +ELLIPSIS\n        0.9444...\n        >>> float(modified_precision(references, hypothesis2, n=1)) # doctest: +ELLIPSIS\n        0.5714...\n        >>> float(modified_precision(references, hypothesis1, n=2)) # doctest: +ELLIPSIS\n        0.5882352941176471\n        >>> float(modified_precision(references, hypothesis2, n=2)) # doctest: +ELLIPSIS\n        0.07692...\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hypothesis: A hypothesis translation.\n    :type hypothesis: list(str)\n    :param n: The ngram order.\n    :type n: int\n    :return: BLEU's modified precision for the nth order ngram.\n    :rtype: Fraction\n    \"\"\"\n    # Extracts all ngrams in hypothesis\n    # Set an empty Counter if hypothesis is empty.\n\n    counts = Counter(ngrams(hypothesis, n)) if len(hypothesis) >= n else Counter()\n    # Extract a union of references' counts.\n    # max_counts = reduce(or_, [Counter(ngrams(ref, n)) for ref in references])\n    max_counts = {}\n    for reference in references:\n        reference_counts = (\n            Counter(ngrams(reference, n)) if len(reference) >= n else Counter()\n        )\n        for ngram in counts:\n            max_counts[ngram] = max(max_counts.get(ngram, 0), reference_counts[ngram])\n\n    # Assigns the intersection between hypothesis and references' counts.\n    clipped_counts = {\n        ngram: min(count, max_counts[ngram]) for ngram, count in counts.items()\n    }\n\n    numerator = sum(clipped_counts.values())\n    # Ensures that denominator is minimum 1 to avoid ZeroDivisionError.\n    # Usually this happens when the ngram order is > len(reference).\n    denominator = max(1, sum(counts.values()))\n\n    return Fraction(numerator, denominator, _normalize=False)\n\n\ndef closest_ref_length(references, hyp_len):\n    \"\"\"\n    This function finds the reference that is the closest length to the\n    hypothesis. The closest reference length is referred to as *r* variable\n    from the brevity penalty formula in Papineni et. al. (2002)\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hyp_len: The length of the hypothesis.\n    :type hyp_len: int\n    :return: The length of the reference that's closest to the hypothesis.\n    :rtype: int\n    \"\"\"\n    ref_lens = (len(reference) for reference in references)\n    closest_ref_len = min(\n        ref_lens, key=lambda ref_len: (abs(ref_len - hyp_len), ref_len)\n    )\n    return closest_ref_len\n","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.closest_ref_length","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.closest_ref_length#L303-L319","kind":"function","name":"closest_ref_length","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":303,"end_line":319,"context_start_line":283,"context_end_line":339,"code":"    for reference in references:\n        reference_counts = (\n            Counter(ngrams(reference, n)) if len(reference) >= n else Counter()\n        )\n        for ngram in counts:\n            max_counts[ngram] = max(max_counts.get(ngram, 0), reference_counts[ngram])\n\n    # Assigns the intersection between hypothesis and references' counts.\n    clipped_counts = {\n        ngram: min(count, max_counts[ngram]) for ngram, count in counts.items()\n    }\n\n    numerator = sum(clipped_counts.values())\n    # Ensures that denominator is minimum 1 to avoid ZeroDivisionError.\n    # Usually this happens when the ngram order is > len(reference).\n    denominator = max(1, sum(counts.values()))\n\n    return Fraction(numerator, denominator, _normalize=False)\n\n\ndef closest_ref_length(references, hyp_len):\n    \"\"\"\n    This function finds the reference that is the closest length to the\n    hypothesis. The closest reference length is referred to as *r* variable\n    from the brevity penalty formula in Papineni et. al. (2002)\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hyp_len: The length of the hypothesis.\n    :type hyp_len: int\n    :return: The length of the reference that's closest to the hypothesis.\n    :rtype: int\n    \"\"\"\n    ref_lens = (len(reference) for reference in references)\n    closest_ref_len = min(\n        ref_lens, key=lambda ref_len: (abs(ref_len - hyp_len), ref_len)\n    )\n    return closest_ref_len\n\n\ndef brevity_penalty(closest_ref_len, hyp_len):\n    \"\"\"\n    Calculate brevity penalty.\n    As the modified n-gram precision still has the problem from the short\n    length sentence, brevity penalty is used to modify the overall BLEU\n    score according to length.\n    An example from the paper. There are three references with length 12, 15\n    and 17. And a concise hypothesis of the length 12. The brevity penalty is 1.\n        >>> reference1 = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> reference2 = list('aaaaaaaaaaaaaaa')   # i.e. ['a'] * 15\n        >>> reference3 = list('aaaaaaaaaaaaaaaaa') # i.e. ['a'] * 17\n        >>> hypothesis = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> references = [reference1, reference2, reference3]\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    In case a hypothesis translation is shorter than the references, penalty is","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.brevity_penalty","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.brevity_penalty#L322-L398","kind":"function","name":"brevity_penalty","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":322,"end_line":398,"context_start_line":302,"context_end_line":418,"code":"\ndef closest_ref_length(references, hyp_len):\n    \"\"\"\n    This function finds the reference that is the closest length to the\n    hypothesis. The closest reference length is referred to as *r* variable\n    from the brevity penalty formula in Papineni et. al. (2002)\n    :param references: A list of reference translations.\n    :type references: list(list(str))\n    :param hyp_len: The length of the hypothesis.\n    :type hyp_len: int\n    :return: The length of the reference that's closest to the hypothesis.\n    :rtype: int\n    \"\"\"\n    ref_lens = (len(reference) for reference in references)\n    closest_ref_len = min(\n        ref_lens, key=lambda ref_len: (abs(ref_len - hyp_len), ref_len)\n    )\n    return closest_ref_len\n\n\ndef brevity_penalty(closest_ref_len, hyp_len):\n    \"\"\"\n    Calculate brevity penalty.\n    As the modified n-gram precision still has the problem from the short\n    length sentence, brevity penalty is used to modify the overall BLEU\n    score according to length.\n    An example from the paper. There are three references with length 12, 15\n    and 17. And a concise hypothesis of the length 12. The brevity penalty is 1.\n        >>> reference1 = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> reference2 = list('aaaaaaaaaaaaaaa')   # i.e. ['a'] * 15\n        >>> reference3 = list('aaaaaaaaaaaaaaaaa') # i.e. ['a'] * 17\n        >>> hypothesis = list('aaaaaaaaaaaa')      # i.e. ['a'] * 12\n        >>> references = [reference1, reference2, reference3]\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    In case a hypothesis translation is shorter than the references, penalty is\n    applied.\n        >>> references = [['a'] * 28, ['a'] * 28]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        0.2635971381157267\n    The length of the closest reference is used to compute the penalty. If the\n    length of a hypothesis is 12, and the reference lengths are 13 and 2, the\n    penalty is applied because the hypothesis length (12) is less then the\n    closest reference length (13).\n        >>> references = [['a'] * 13, ['a'] * 2]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len) # doctest: +ELLIPSIS\n        0.9200...\n    The brevity penalty doesn't depend on reference order. More importantly,\n    when two reference sentences are at the same distance, the shortest\n    reference sentence length is used.\n        >>> references = [['a'] * 13, ['a'] * 11]\n        >>> hypothesis = ['a'] * 12\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> bp1 = brevity_penalty(closest_ref_len, hyp_len)\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(reversed(references), hyp_len)\n        >>> bp2 = brevity_penalty(closest_ref_len, hyp_len)\n        >>> bp1 == bp2 == 1\n        True\n    A test example from mteval-v13a.pl (starting from the line 705):\n        >>> references = [['a'] * 11, ['a'] * 8]\n        >>> hypothesis = ['a'] * 7\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len) # doctest: +ELLIPSIS\n        0.8668...\n        >>> references = [['a'] * 11, ['a'] * 8, ['a'] * 6, ['a'] * 7]\n        >>> hypothesis = ['a'] * 7\n        >>> hyp_len = len(hypothesis)\n        >>> closest_ref_len =  closest_ref_length(references, hyp_len)\n        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    :param hyp_len: The length of the hypothesis for a single sentence OR the\n    sum of all the hypotheses' lengths for a corpus\n    :type hyp_len: int\n    :param closest_ref_len: The length of the closest reference for a single\n    hypothesis OR the sum of all the closest references for every hypotheses.\n    :type closest_ref_len: int\n    :return: BLEU's brevity penalty.\n    :rtype: float\n    \"\"\"\n    if hyp_len > closest_ref_len:\n        return 1\n    # If hypothesis is empty, brevity penalty = 0 should result in BLEU = 0.0\n    elif hyp_len == 0:\n        return 0\n    else:\n        return math.exp(1 - closest_ref_len / hyp_len)\n\n\nclass SmoothingFunction:\n    \"\"\"\n    This is an implementation of the smoothing techniques\n    for segment-level BLEU scores that was presented in\n    Boxing Chen and Collin Cherry (2014) A Systematic Comparison of\n    Smoothing Techniques for Sentence-Level BLEU. In WMT14.\n    http://acl2014.org/acl2014/W14-33/pdf/W14-3346.pdf\n    \"\"\"\n\n    def __init__(self, epsilon=0.1, alpha=5, k=5):\n        \"\"\"\n        This will initialize the parameters required for the various smoothing\n        techniques, the default values are set to the numbers used in the\n        experiments from Chen and Cherry (2014).\n        >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which', 'ensures',\n        ...                 'that', 'the', 'military', 'always', 'obeys', 'the',\n        ...                 'commands', 'of', 'the', 'party']\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that', 'ensures',","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.SmoothingFunction","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.SmoothingFunction#L401-L590","kind":"class","name":"SmoothingFunction","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":401,"end_line":590,"context_start_line":381,"context_end_line":590,"code":"        >>> brevity_penalty(closest_ref_len, hyp_len)\n        1.0\n    :param hyp_len: The length of the hypothesis for a single sentence OR the\n    sum of all the hypotheses' lengths for a corpus\n    :type hyp_len: int\n    :param closest_ref_len: The length of the closest reference for a single\n    hypothesis OR the sum of all the closest references for every hypotheses.\n    :type closest_ref_len: int\n    :return: BLEU's brevity penalty.\n    :rtype: float\n    \"\"\"\n    if hyp_len > closest_ref_len:\n        return 1\n    # If hypothesis is empty, brevity penalty = 0 should result in BLEU = 0.0\n    elif hyp_len == 0:\n        return 0\n    else:\n        return math.exp(1 - closest_ref_len / hyp_len)\n\n\nclass SmoothingFunction:\n    \"\"\"\n    This is an implementation of the smoothing techniques\n    for segment-level BLEU scores that was presented in\n    Boxing Chen and Collin Cherry (2014) A Systematic Comparison of\n    Smoothing Techniques for Sentence-Level BLEU. In WMT14.\n    http://acl2014.org/acl2014/W14-33/pdf/W14-3346.pdf\n    \"\"\"\n\n    def __init__(self, epsilon=0.1, alpha=5, k=5):\n        \"\"\"\n        This will initialize the parameters required for the various smoothing\n        techniques, the default values are set to the numbers used in the\n        experiments from Chen and Cherry (2014).\n        >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which', 'ensures',\n        ...                 'that', 'the', 'military', 'always', 'obeys', 'the',\n        ...                 'commands', 'of', 'the', 'party']\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that', 'ensures',\n        ...               'that', 'the', 'military', 'will', 'forever', 'heed',\n        ...               'Party', 'commands']\n        >>> chencherry = SmoothingFunction()\n        >>> print(sentence_bleu([reference1], hypothesis1)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method0)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method1)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method2)) # doctest: +ELLIPSIS\n        0.4489...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method3)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method4)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method5)) # doctest: +ELLIPSIS\n        0.4905...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method6)) # doctest: +ELLIPSIS\n        0.4135...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method7)) # doctest: +ELLIPSIS\n        0.4905...\n        :param epsilon: the epsilon value use in method 1\n        :type epsilon: float\n        :param alpha: the alpha value use in method 6\n        :type alpha: int\n        :param k: the k value use in method 4\n        :type k: int\n        \"\"\"\n        self.epsilon = epsilon\n        self.alpha = alpha\n        self.k = k\n\n    def method0(self, p_n, *args, **kwargs):\n        \"\"\"\n        No smoothing.\n        \"\"\"\n        p_n_new = []\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator != 0:\n                p_n_new.append(p_i)\n            else:\n                _msg = str(\n                    \"\\nThe hypothesis contains 0 counts of {}-gram overlaps.\\n\"\n                    \"Therefore the BLEU score evaluates to 0, independently of\\n\"\n                    \"how many N-gram overlaps of lower order it contains.\\n\"\n                    \"Consider using lower n-gram order or use \"\n                    \"SmoothingFunction()\"\n                ).format(i + 1)\n                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.\n                # Due to BLEU geometric mean computation in logarithm space,\n                # we we need to take the return sys.float_info.min such that\n                # math.log(sys.float_info.min) returns a 0 precision score.\n                p_n_new.append(sys.float_info.min)\n        return p_n_new\n\n    def method1(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 1: Add *epsilon* counts to precision with 0 counts.\n        \"\"\"\n        return [\n            (p_i.numerator + self.epsilon) / p_i.denominator\n            if p_i.numerator == 0\n            else p_i\n            for p_i in p_n\n        ]\n\n    def method2(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 2: Add 1 to both numerator and denominator from\n        Chin-Yew Lin and Franz Josef Och (2004) Automatic evaluation of\n        machine translation quality using longest common subsequence and\n        skip-bigram statistics. In ACL04.\n        \"\"\"\n        return [\n            Fraction(p_i.numerator + 1, p_i.denominator + 1, _normalize=False)\n            for p_i in p_n\n        ]\n\n    def method3(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 3: NIST geometric sequence smoothing\n        The smoothing is computed by taking 1 / ( 2^k ), instead of 0, for each\n        precision score whose matching n-gram count is null.\n        k is 1 for the first 'n' value for which the n-gram match count is null/\n        For example, if the text contains:\n         - one 2-gram match\n         - and (consequently) two 1-gram matches\n        the n-gram count for each individual precision score would be:\n         - n=1  =>  prec_count = 2     (two unigrams)\n         - n=2  =>  prec_count = 1     (one bigram)\n         - n=3  =>  prec_count = 1/2   (no trigram,  taking 'smoothed' value of 1 / ( 2^k ), with k=1)\n         - n=4  =>  prec_count = 1/4   (no fourgram, taking 'smoothed' value of 1 / ( 2^k ), with k=2)\n        \"\"\"\n        incvnt = 1  # From the mteval-v13a.pl, it's referred to as k.\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0:\n                p_n[i] = 1 / (2 ** incvnt * p_i.denominator)\n                incvnt += 1\n        return p_n\n\n    def method4(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 4:\n        Shorter translations may have inflated precision values due to having\n        smaller denominators; therefore, we give them proportionally\n        smaller smoothed counts. Instead of scaling to 1/(2^k), Chen and Cherry\n        suggests dividing by 1/ln(len(T)), where T is the length of the translation.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0 and hyp_len != 0:\n                incvnt = i + 1 * self.k / math.log(\n                    hyp_len\n                )  # Note that this K is different from the K from NIST.\n                p_n[i] = incvnt / p_i.denominator\n        return p_n\n\n    def method5(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 5:\n        The matched counts for similar values of n should be similar. To a\n        calculate the n-gram matched count, it averages the n−1, n and n+1 gram\n        matched counts.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        m = {}\n        # Requires an precision value for an addition ngram order.\n        p_n_plus1 = p_n + [modified_precision(references, hypothesis, 5)]\n        m[-1] = p_n[0] + 1\n        for i, p_i in enumerate(p_n):\n            p_n[i] = (m[i - 1] + p_i + p_n_plus1[i + 1]) / 3\n            m[i] = p_n[i]\n        return p_n\n\n    def method6(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 6:\n        Interpolates the maximum likelihood estimate of the precision *p_n* with\n        a prior estimate *pi0*. The prior is estimated by assuming that the ratio\n        between pn and pn−1 will be the same as that between pn−1 and pn−2; from\n        Gao and He (2013) Training MRF-Based Phrase Translation Models using\n        Gradient Ascent. In NAACL.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        # This smoothing only works when p_1 and p_2 is non-zero.\n        # Raise an error with an appropriate message when the input is too short\n        # to use this smoothing technique.\n        assert p_n[2], \"This smoothing method requires non-zero precision for bigrams.\"\n        for i, p_i in enumerate(p_n):\n            if i in [0, 1]:  # Skips the first 2 orders of ngrams.\n                continue\n            else:\n                pi0 = 0 if p_n[i - 2] == 0 else p_n[i - 1] ** 2 / p_n[i - 2]\n                # No. of ngrams in translation that matches the reference.\n                m = p_i.numerator\n                # No. of ngrams in translation.\n                l = sum(1 for _ in ngrams(hypothesis, i + 1))\n                # Calculates the interpolated precision.\n                p_n[i] = (m + self.alpha * pi0) / (l + self.alpha)\n        return p_n\n\n    def method7(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 7:\n        Interpolates methods 4 and 5.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        p_n = self.method4(p_n, references, hypothesis, hyp_len)\n        p_n = self.method5(p_n, references, hypothesis, hyp_len)\n        return p_n","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.__init__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.__init__#L410-L449","kind":"function","name":"__init__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":410,"end_line":449,"context_start_line":390,"context_end_line":469,"code":"    :rtype: float\n    \"\"\"\n    if hyp_len > closest_ref_len:\n        return 1\n    # If hypothesis is empty, brevity penalty = 0 should result in BLEU = 0.0\n    elif hyp_len == 0:\n        return 0\n    else:\n        return math.exp(1 - closest_ref_len / hyp_len)\n\n\nclass SmoothingFunction:\n    \"\"\"\n    This is an implementation of the smoothing techniques\n    for segment-level BLEU scores that was presented in\n    Boxing Chen and Collin Cherry (2014) A Systematic Comparison of\n    Smoothing Techniques for Sentence-Level BLEU. In WMT14.\n    http://acl2014.org/acl2014/W14-33/pdf/W14-3346.pdf\n    \"\"\"\n\n    def __init__(self, epsilon=0.1, alpha=5, k=5):\n        \"\"\"\n        This will initialize the parameters required for the various smoothing\n        techniques, the default values are set to the numbers used in the\n        experiments from Chen and Cherry (2014).\n        >>> hypothesis1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'which', 'ensures',\n        ...                 'that', 'the', 'military', 'always', 'obeys', 'the',\n        ...                 'commands', 'of', 'the', 'party']\n        >>> reference1 = ['It', 'is', 'a', 'guide', 'to', 'action', 'that', 'ensures',\n        ...               'that', 'the', 'military', 'will', 'forever', 'heed',\n        ...               'Party', 'commands']\n        >>> chencherry = SmoothingFunction()\n        >>> print(sentence_bleu([reference1], hypothesis1)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method0)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method1)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method2)) # doctest: +ELLIPSIS\n        0.4489...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method3)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method4)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method5)) # doctest: +ELLIPSIS\n        0.4905...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method6)) # doctest: +ELLIPSIS\n        0.4135...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method7)) # doctest: +ELLIPSIS\n        0.4905...\n        :param epsilon: the epsilon value use in method 1\n        :type epsilon: float\n        :param alpha: the alpha value use in method 6\n        :type alpha: int\n        :param k: the k value use in method 4\n        :type k: int\n        \"\"\"\n        self.epsilon = epsilon\n        self.alpha = alpha\n        self.k = k\n\n    def method0(self, p_n, *args, **kwargs):\n        \"\"\"\n        No smoothing.\n        \"\"\"\n        p_n_new = []\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator != 0:\n                p_n_new.append(p_i)\n            else:\n                _msg = str(\n                    \"\\nThe hypothesis contains 0 counts of {}-gram overlaps.\\n\"\n                    \"Therefore the BLEU score evaluates to 0, independently of\\n\"\n                    \"how many N-gram overlaps of lower order it contains.\\n\"\n                    \"Consider using lower n-gram order or use \"\n                    \"SmoothingFunction()\"\n                ).format(i + 1)\n                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method0","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method0#L451-L474","kind":"function","name":"method0","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":451,"end_line":474,"context_start_line":431,"context_end_line":494,"code":"        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method4)) # doctest: +ELLIPSIS\n        0.4118...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method5)) # doctest: +ELLIPSIS\n        0.4905...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method6)) # doctest: +ELLIPSIS\n        0.4135...\n        >>> print(sentence_bleu([reference1], hypothesis1, smoothing_function=chencherry.method7)) # doctest: +ELLIPSIS\n        0.4905...\n        :param epsilon: the epsilon value use in method 1\n        :type epsilon: float\n        :param alpha: the alpha value use in method 6\n        :type alpha: int\n        :param k: the k value use in method 4\n        :type k: int\n        \"\"\"\n        self.epsilon = epsilon\n        self.alpha = alpha\n        self.k = k\n\n    def method0(self, p_n, *args, **kwargs):\n        \"\"\"\n        No smoothing.\n        \"\"\"\n        p_n_new = []\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator != 0:\n                p_n_new.append(p_i)\n            else:\n                _msg = str(\n                    \"\\nThe hypothesis contains 0 counts of {}-gram overlaps.\\n\"\n                    \"Therefore the BLEU score evaluates to 0, independently of\\n\"\n                    \"how many N-gram overlaps of lower order it contains.\\n\"\n                    \"Consider using lower n-gram order or use \"\n                    \"SmoothingFunction()\"\n                ).format(i + 1)\n                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.\n                # Due to BLEU geometric mean computation in logarithm space,\n                # we we need to take the return sys.float_info.min such that\n                # math.log(sys.float_info.min) returns a 0 precision score.\n                p_n_new.append(sys.float_info.min)\n        return p_n_new\n\n    def method1(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 1: Add *epsilon* counts to precision with 0 counts.\n        \"\"\"\n        return [\n            (p_i.numerator + self.epsilon) / p_i.denominator\n            if p_i.numerator == 0\n            else p_i\n            for p_i in p_n\n        ]\n\n    def method2(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 2: Add 1 to both numerator and denominator from\n        Chin-Yew Lin and Franz Josef Och (2004) Automatic evaluation of\n        machine translation quality using longest common subsequence and\n        skip-bigram statistics. In ACL04.\n        \"\"\"\n        return [","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method1#L476-L485","kind":"function","name":"method1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":476,"end_line":485,"context_start_line":456,"context_end_line":505,"code":"        for i, p_i in enumerate(p_n):\n            if p_i.numerator != 0:\n                p_n_new.append(p_i)\n            else:\n                _msg = str(\n                    \"\\nThe hypothesis contains 0 counts of {}-gram overlaps.\\n\"\n                    \"Therefore the BLEU score evaluates to 0, independently of\\n\"\n                    \"how many N-gram overlaps of lower order it contains.\\n\"\n                    \"Consider using lower n-gram order or use \"\n                    \"SmoothingFunction()\"\n                ).format(i + 1)\n                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.\n                # Due to BLEU geometric mean computation in logarithm space,\n                # we we need to take the return sys.float_info.min such that\n                # math.log(sys.float_info.min) returns a 0 precision score.\n                p_n_new.append(sys.float_info.min)\n        return p_n_new\n\n    def method1(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 1: Add *epsilon* counts to precision with 0 counts.\n        \"\"\"\n        return [\n            (p_i.numerator + self.epsilon) / p_i.denominator\n            if p_i.numerator == 0\n            else p_i\n            for p_i in p_n\n        ]\n\n    def method2(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 2: Add 1 to both numerator and denominator from\n        Chin-Yew Lin and Franz Josef Och (2004) Automatic evaluation of\n        machine translation quality using longest common subsequence and\n        skip-bigram statistics. In ACL04.\n        \"\"\"\n        return [\n            Fraction(p_i.numerator + 1, p_i.denominator + 1, _normalize=False)\n            for p_i in p_n\n        ]\n\n    def method3(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 3: NIST geometric sequence smoothing\n        The smoothing is computed by taking 1 / ( 2^k ), instead of 0, for each\n        precision score whose matching n-gram count is null.\n        k is 1 for the first 'n' value for which the n-gram match count is null/\n        For example, if the text contains:","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method2#L487-L497","kind":"function","name":"method2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":487,"end_line":497,"context_start_line":467,"context_end_line":517,"code":"                warnings.warn(_msg)\n                # When numerator==0 where denonminator==0 or !=0, the result\n                # for the precision score should be equal to 0 or undefined.\n                # Due to BLEU geometric mean computation in logarithm space,\n                # we we need to take the return sys.float_info.min such that\n                # math.log(sys.float_info.min) returns a 0 precision score.\n                p_n_new.append(sys.float_info.min)\n        return p_n_new\n\n    def method1(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 1: Add *epsilon* counts to precision with 0 counts.\n        \"\"\"\n        return [\n            (p_i.numerator + self.epsilon) / p_i.denominator\n            if p_i.numerator == 0\n            else p_i\n            for p_i in p_n\n        ]\n\n    def method2(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 2: Add 1 to both numerator and denominator from\n        Chin-Yew Lin and Franz Josef Och (2004) Automatic evaluation of\n        machine translation quality using longest common subsequence and\n        skip-bigram statistics. In ACL04.\n        \"\"\"\n        return [\n            Fraction(p_i.numerator + 1, p_i.denominator + 1, _normalize=False)\n            for p_i in p_n\n        ]\n\n    def method3(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 3: NIST geometric sequence smoothing\n        The smoothing is computed by taking 1 / ( 2^k ), instead of 0, for each\n        precision score whose matching n-gram count is null.\n        k is 1 for the first 'n' value for which the n-gram match count is null/\n        For example, if the text contains:\n         - one 2-gram match\n         - and (consequently) two 1-gram matches\n        the n-gram count for each individual precision score would be:\n         - n=1  =>  prec_count = 2     (two unigrams)\n         - n=2  =>  prec_count = 1     (one bigram)\n         - n=3  =>  prec_count = 1/2   (no trigram,  taking 'smoothed' value of 1 / ( 2^k ), with k=1)\n         - n=4  =>  prec_count = 1/4   (no fourgram, taking 'smoothed' value of 1 / ( 2^k ), with k=2)\n        \"\"\"\n        incvnt = 1  # From the mteval-v13a.pl, it's referred to as k.\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0:\n                p_n[i] = 1 / (2 ** incvnt * p_i.denominator)","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method3","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method3#L499-L519","kind":"function","name":"method3","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":499,"end_line":519,"context_start_line":479,"context_end_line":539,"code":"        \"\"\"\n        return [\n            (p_i.numerator + self.epsilon) / p_i.denominator\n            if p_i.numerator == 0\n            else p_i\n            for p_i in p_n\n        ]\n\n    def method2(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 2: Add 1 to both numerator and denominator from\n        Chin-Yew Lin and Franz Josef Och (2004) Automatic evaluation of\n        machine translation quality using longest common subsequence and\n        skip-bigram statistics. In ACL04.\n        \"\"\"\n        return [\n            Fraction(p_i.numerator + 1, p_i.denominator + 1, _normalize=False)\n            for p_i in p_n\n        ]\n\n    def method3(self, p_n, *args, **kwargs):\n        \"\"\"\n        Smoothing method 3: NIST geometric sequence smoothing\n        The smoothing is computed by taking 1 / ( 2^k ), instead of 0, for each\n        precision score whose matching n-gram count is null.\n        k is 1 for the first 'n' value for which the n-gram match count is null/\n        For example, if the text contains:\n         - one 2-gram match\n         - and (consequently) two 1-gram matches\n        the n-gram count for each individual precision score would be:\n         - n=1  =>  prec_count = 2     (two unigrams)\n         - n=2  =>  prec_count = 1     (one bigram)\n         - n=3  =>  prec_count = 1/2   (no trigram,  taking 'smoothed' value of 1 / ( 2^k ), with k=1)\n         - n=4  =>  prec_count = 1/4   (no fourgram, taking 'smoothed' value of 1 / ( 2^k ), with k=2)\n        \"\"\"\n        incvnt = 1  # From the mteval-v13a.pl, it's referred to as k.\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0:\n                p_n[i] = 1 / (2 ** incvnt * p_i.denominator)\n                incvnt += 1\n        return p_n\n\n    def method4(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 4:\n        Shorter translations may have inflated precision values due to having\n        smaller denominators; therefore, we give them proportionally\n        smaller smoothed counts. Instead of scaling to 1/(2^k), Chen and Cherry\n        suggests dividing by 1/ln(len(T)), where T is the length of the translation.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0 and hyp_len != 0:\n                incvnt = i + 1 * self.k / math.log(\n                    hyp_len\n                )  # Note that this K is different from the K from NIST.\n                p_n[i] = incvnt / p_i.denominator\n        return p_n\n\n    def method5(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method4","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method4#L521-L536","kind":"function","name":"method4","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":521,"end_line":536,"context_start_line":501,"context_end_line":556,"code":"        Smoothing method 3: NIST geometric sequence smoothing\n        The smoothing is computed by taking 1 / ( 2^k ), instead of 0, for each\n        precision score whose matching n-gram count is null.\n        k is 1 for the first 'n' value for which the n-gram match count is null/\n        For example, if the text contains:\n         - one 2-gram match\n         - and (consequently) two 1-gram matches\n        the n-gram count for each individual precision score would be:\n         - n=1  =>  prec_count = 2     (two unigrams)\n         - n=2  =>  prec_count = 1     (one bigram)\n         - n=3  =>  prec_count = 1/2   (no trigram,  taking 'smoothed' value of 1 / ( 2^k ), with k=1)\n         - n=4  =>  prec_count = 1/4   (no fourgram, taking 'smoothed' value of 1 / ( 2^k ), with k=2)\n        \"\"\"\n        incvnt = 1  # From the mteval-v13a.pl, it's referred to as k.\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0:\n                p_n[i] = 1 / (2 ** incvnt * p_i.denominator)\n                incvnt += 1\n        return p_n\n\n    def method4(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 4:\n        Shorter translations may have inflated precision values due to having\n        smaller denominators; therefore, we give them proportionally\n        smaller smoothed counts. Instead of scaling to 1/(2^k), Chen and Cherry\n        suggests dividing by 1/ln(len(T)), where T is the length of the translation.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0 and hyp_len != 0:\n                incvnt = i + 1 * self.k / math.log(\n                    hyp_len\n                )  # Note that this K is different from the K from NIST.\n                p_n[i] = incvnt / p_i.denominator\n        return p_n\n\n    def method5(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 5:\n        The matched counts for similar values of n should be similar. To a\n        calculate the n-gram matched count, it averages the n−1, n and n+1 gram\n        matched counts.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        m = {}\n        # Requires an precision value for an addition ngram order.\n        p_n_plus1 = p_n + [modified_precision(references, hypothesis, 5)]\n        m[-1] = p_n[0] + 1\n        for i, p_i in enumerate(p_n):\n            p_n[i] = (m[i - 1] + p_i + p_n_plus1[i + 1]) / 3\n            m[i] = p_n[i]\n        return p_n\n\n    def method6(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method5","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method5#L538-L553","kind":"function","name":"method5","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":538,"end_line":553,"context_start_line":518,"context_end_line":573,"code":"                incvnt += 1\n        return p_n\n\n    def method4(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 4:\n        Shorter translations may have inflated precision values due to having\n        smaller denominators; therefore, we give them proportionally\n        smaller smoothed counts. Instead of scaling to 1/(2^k), Chen and Cherry\n        suggests dividing by 1/ln(len(T)), where T is the length of the translation.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        for i, p_i in enumerate(p_n):\n            if p_i.numerator == 0 and hyp_len != 0:\n                incvnt = i + 1 * self.k / math.log(\n                    hyp_len\n                )  # Note that this K is different from the K from NIST.\n                p_n[i] = incvnt / p_i.denominator\n        return p_n\n\n    def method5(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 5:\n        The matched counts for similar values of n should be similar. To a\n        calculate the n-gram matched count, it averages the n−1, n and n+1 gram\n        matched counts.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        m = {}\n        # Requires an precision value for an addition ngram order.\n        p_n_plus1 = p_n + [modified_precision(references, hypothesis, 5)]\n        m[-1] = p_n[0] + 1\n        for i, p_i in enumerate(p_n):\n            p_n[i] = (m[i - 1] + p_i + p_n_plus1[i + 1]) / 3\n            m[i] = p_n[i]\n        return p_n\n\n    def method6(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 6:\n        Interpolates the maximum likelihood estimate of the precision *p_n* with\n        a prior estimate *pi0*. The prior is estimated by assuming that the ratio\n        between pn and pn−1 will be the same as that between pn−1 and pn−2; from\n        Gao and He (2013) Training MRF-Based Phrase Translation Models using\n        Gradient Ascent. In NAACL.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        # This smoothing only works when p_1 and p_2 is non-zero.\n        # Raise an error with an appropriate message when the input is too short\n        # to use this smoothing technique.\n        assert p_n[2], \"This smoothing method requires non-zero precision for bigrams.\"\n        for i, p_i in enumerate(p_n):\n            if i in [0, 1]:  # Skips the first 2 orders of ngrams.\n                continue\n            else:\n                pi0 = 0 if p_n[i - 2] == 0 else p_n[i - 1] ** 2 / p_n[i - 2]","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method6","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method6#L555-L580","kind":"function","name":"method6","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":555,"end_line":580,"context_start_line":535,"context_end_line":590,"code":"                p_n[i] = incvnt / p_i.denominator\n        return p_n\n\n    def method5(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 5:\n        The matched counts for similar values of n should be similar. To a\n        calculate the n-gram matched count, it averages the n−1, n and n+1 gram\n        matched counts.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        m = {}\n        # Requires an precision value for an addition ngram order.\n        p_n_plus1 = p_n + [modified_precision(references, hypothesis, 5)]\n        m[-1] = p_n[0] + 1\n        for i, p_i in enumerate(p_n):\n            p_n[i] = (m[i - 1] + p_i + p_n_plus1[i + 1]) / 3\n            m[i] = p_n[i]\n        return p_n\n\n    def method6(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 6:\n        Interpolates the maximum likelihood estimate of the precision *p_n* with\n        a prior estimate *pi0*. The prior is estimated by assuming that the ratio\n        between pn and pn−1 will be the same as that between pn−1 and pn−2; from\n        Gao and He (2013) Training MRF-Based Phrase Translation Models using\n        Gradient Ascent. In NAACL.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        # This smoothing only works when p_1 and p_2 is non-zero.\n        # Raise an error with an appropriate message when the input is too short\n        # to use this smoothing technique.\n        assert p_n[2], \"This smoothing method requires non-zero precision for bigrams.\"\n        for i, p_i in enumerate(p_n):\n            if i in [0, 1]:  # Skips the first 2 orders of ngrams.\n                continue\n            else:\n                pi0 = 0 if p_n[i - 2] == 0 else p_n[i - 1] ** 2 / p_n[i - 2]\n                # No. of ngrams in translation that matches the reference.\n                m = p_i.numerator\n                # No. of ngrams in translation.\n                l = sum(1 for _ in ngrams(hypothesis, i + 1))\n                # Calculates the interpolated precision.\n                p_n[i] = (m + self.alpha * pi0) / (l + self.alpha)\n        return p_n\n\n    def method7(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 7:\n        Interpolates methods 4 and 5.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        p_n = self.method4(p_n, references, hypothesis, hyp_len)\n        p_n = self.method5(p_n, references, hypothesis, hyp_len)\n        return p_n","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method7","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.bleu.method7#L582-L590","kind":"function","name":"method7","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":582,"end_line":590,"context_start_line":562,"context_end_line":590,"code":"        Gradient Ascent. In NAACL.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        # This smoothing only works when p_1 and p_2 is non-zero.\n        # Raise an error with an appropriate message when the input is too short\n        # to use this smoothing technique.\n        assert p_n[2], \"This smoothing method requires non-zero precision for bigrams.\"\n        for i, p_i in enumerate(p_n):\n            if i in [0, 1]:  # Skips the first 2 orders of ngrams.\n                continue\n            else:\n                pi0 = 0 if p_n[i - 2] == 0 else p_n[i - 1] ** 2 / p_n[i - 2]\n                # No. of ngrams in translation that matches the reference.\n                m = p_i.numerator\n                # No. of ngrams in translation.\n                l = sum(1 for _ in ngrams(hypothesis, i + 1))\n                # Calculates the interpolated precision.\n                p_n[i] = (m + self.alpha * pi0) / (l + self.alpha)\n        return p_n\n\n    def method7(self, p_n, references, hypothesis, hyp_len=None, *args, **kwargs):\n        \"\"\"\n        Smoothing method 7:\n        Interpolates methods 4 and 5.\n        \"\"\"\n        hyp_len = hyp_len if hyp_len else len(hypothesis)\n        p_n = self.method4(p_n, references, hypothesis, hyp_len)\n        p_n = self.method5(p_n, references, hypothesis, hyp_len)\n        return p_n","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG#L1-L1179","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","language":"python","start_line":1,"end_line":1179,"context_start_line":1,"context_end_line":1179,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom tree_sitter import Language, Parser\nfrom .utils import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\n\n\ndef DFG_python(root_node,index_to_code,states):\n    assignment=['assignment','augmented_assignment','for_in_clause']\n    if_statement=['if_statement']\n    for_statement=['for_statement']\n    while_statement=['while_statement']\n    do_first_statement=['for_in_clause'] \n    def_statement=['default_parameter']\n    states=states.copy() \n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':        \n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:\n        name=root_node.child_by_field_name('name')\n        value=root_node.child_by_field_name('value')\n        DFG=[]\n        if value is None:\n            indexs=tree_to_variable_index(name,index_to_code)\n            for index in indexs:\n                idx,code=index_to_code[index]\n                DFG.append((code,idx,'comesFrom',[],[]))\n                states[code]=[idx]\n            return sorted(DFG,key=lambda x:x[1]),states\n        else:\n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)\n            temp,states=DFG_python(value,index_to_code,states)\n            DFG+=temp            \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'comesFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            return sorted(DFG,key=lambda x:x[1]),states        \n    elif root_node.type in assignment:\n        if root_node.type=='for_in_clause':\n            right_nodes=[root_node.children[-1]]\n            left_nodes=[root_node.child_by_field_name('left')]\n        else:\n            if root_node.child_by_field_name('right') is None:\n                return [],states\n            left_nodes=[x for x in root_node.child_by_field_name('left').children if x.type!=',']\n            right_nodes=[x for x in root_node.child_by_field_name('right').children if x.type!=',']\n            if len(right_nodes)!=len(left_nodes):\n                left_nodes=[root_node.child_by_field_name('left')]\n                right_nodes=[root_node.child_by_field_name('right')]\n            if len(left_nodes)==0:\n                left_nodes=[root_node.child_by_field_name('left')]\n            if len(right_nodes)==0:\n                right_nodes=[root_node.child_by_field_name('right')]\n        DFG=[]\n        for node in right_nodes:\n            temp,states=DFG_python(node,index_to_code,states)\n            DFG+=temp\n            \n        for left_node,right_node in zip(left_nodes,right_nodes):\n            left_tokens_index=tree_to_variable_index(left_node,index_to_code)\n            right_tokens_index=tree_to_variable_index(right_node,index_to_code)\n            temp=[]\n            for token1_index in left_tokens_index:\n                idx1,code1=index_to_code[token1_index]\n                temp.append((code1,idx1,'computedFrom',[index_to_code[x][1] for x in right_tokens_index],\n                             [index_to_code[x][0] for x in right_tokens_index]))\n                states[code1]=[idx1]\n            DFG+=temp        \n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in if_statement:\n        DFG=[]\n        current_states=states.copy()\n        others_states=[]\n        tag=False\n        if 'else' in root_node.type:\n            tag=True\n        for child in root_node.children:\n            if 'else' in child.type:\n                tag=True\n            if child.type not in ['elif_clause','else_clause']:\n                temp,current_states=DFG_python(child,index_to_code,current_states)\n                DFG+=temp\n            else:\n                temp,new_states=DFG_python(child,index_to_code,states)\n                DFG+=temp\n                others_states.append(new_states)\n        others_states.append(current_states)\n        if tag is False:\n            others_states.append(states)\n        new_states={}\n        for dic in others_states:\n            for key in dic:\n                if key not in new_states:\n                    new_states[key]=dic[key].copy()\n                else:\n                    new_states[key]+=dic[key]\n        for key in new_states:\n            new_states[key]=sorted(list(set(new_states[key])))\n        return sorted(DFG,key=lambda x:x[1]),new_states\n    elif root_node.type in for_statement:\n        DFG=[]\n        for i in range(2):\n            right_nodes=[x for x in root_node.child_by_field_name('right').children if x.type!=',']\n            left_nodes=[x for x in root_node.child_by_field_name('left').children if x.type!=',']\n            if len(right_nodes)!=len(left_nodes):\n                left_nodes=[root_node.child_by_field_name('left')]\n                right_nodes=[root_node.child_by_field_name('right')]\n            if len(left_nodes)==0:\n                left_nodes=[root_node.child_by_field_name('left')]\n            if len(right_nodes)==0:\n                right_nodes=[root_node.child_by_field_name('right')]\n            for node in right_nodes:\n                temp,states=DFG_python(node,index_to_code,states)\n                DFG+=temp\n            for left_node,right_node in zip(left_nodes,right_nodes):\n                left_tokens_index=tree_to_variable_index(left_node,index_to_code)\n                right_tokens_index=tree_to_variable_index(right_node,index_to_code)\n                temp=[]\n                for token1_index in left_tokens_index:\n                    idx1,code1=index_to_code[token1_index]\n                    temp.append((code1,idx1,'computedFrom',[index_to_code[x][1] for x in right_tokens_index],\n                                 [index_to_code[x][0] for x in right_tokens_index]))\n                    states[code1]=[idx1]\n                DFG+=temp   \n            if  root_node.children[-1].type==\"block\":\n                temp,states=DFG_python(root_node.children[-1],index_to_code,states)\n                DFG+=temp \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in while_statement:  \n        DFG=[]\n        for i in range(2):\n            for child in root_node.children:\n                temp,states=DFG_python(child,index_to_code,states)\n                DFG+=temp    \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_python(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_python(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n        \n\n######## OTHER LANGUAGES #########\n\ndef DFG_java(root_node,index_to_code,states):\n    assignment=['assignment_expression']\n    def_statement=['variable_declarator']\n    increment_statement=['update_expression']\n    if_statement=['if_statement','else']\n    for_statement=['for_statement']\n    enhanced_for_statement=['enhanced_for_statement']\n    while_statement=['while_statement']\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:\n        name=root_node.child_by_field_name('name')\n        value=root_node.child_by_field_name('value')\n        DFG=[]\n        if value is None:\n            indexs=tree_to_variable_index(name,index_to_code)\n            for index in indexs:\n                idx,code=index_to_code[index]\n                DFG.append((code,idx,'comesFrom',[],[]))\n                states[code]=[idx]\n            return sorted(DFG,key=lambda x:x[1]),states\n        else:\n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)\n            temp,states=DFG_java(value,index_to_code,states)\n            DFG+=temp            \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'comesFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in assignment:\n        left_nodes=root_node.child_by_field_name('left')\n        right_nodes=root_node.child_by_field_name('right')\n        DFG=[]\n        temp,states=DFG_java(right_nodes,index_to_code,states)\n        DFG+=temp            \n        name_indexs=tree_to_variable_index(left_nodes,index_to_code)\n        value_indexs=tree_to_variable_index(right_nodes,index_to_code)        \n        for index1 in name_indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in value_indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]   \n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in increment_statement:\n        DFG=[]\n        indexs=tree_to_variable_index(root_node,index_to_code)\n        for index1 in indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]\n        return sorted(DFG,key=lambda x:x[1]),states   \n    elif root_node.type in if_statement:\n        DFG=[]\n        current_states=states.copy()\n        others_states=[]\n        flag=False\n        tag=False\n        if 'else' in root_node.type:\n            tag=True\n        for child in root_node.children:\n            if 'else' in child.type:\n                tag=True\n            if child.type not in if_statement and flag is False:\n                temp,current_states=DFG_java(child,index_to_code,current_states)\n                DFG+=temp\n            else:\n                flag=True\n                temp,new_states=DFG_java(child,index_to_code,states)\n                DFG+=temp\n                others_states.append(new_states)\n        others_states.append(current_states)\n        if tag is False:\n            others_states.append(states)\n        new_states={}\n        for dic in others_states:\n            for key in dic:\n                if key not in new_states:\n                    new_states[key]=dic[key].copy()\n                else:\n                    new_states[key]+=dic[key]\n        for key in new_states:\n            new_states[key]=sorted(list(set(new_states[key])))\n        return sorted(DFG,key=lambda x:x[1]),new_states\n    elif root_node.type in for_statement:\n        DFG=[]\n        for child in root_node.children:\n            temp,states=DFG_java(child,index_to_code,states)\n            DFG+=temp\n        flag=False\n        for child in root_node.children:\n            if flag:\n                temp,states=DFG_java(child,index_to_code,states)\n                DFG+=temp                \n            elif child.type==\"local_variable_declaration\":\n                flag=True\n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in enhanced_for_statement:\n        name=root_node.child_by_field_name('name')\n        value=root_node.child_by_field_name('value')\n        body=root_node.child_by_field_name('body')\n        DFG=[]\n        for i in range(2):\n            temp,states=DFG_java(value,index_to_code,states)\n            DFG+=temp       \n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)        \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            temp,states=DFG_java(body,index_to_code,states)\n            DFG+=temp                       \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in while_statement:  \n        DFG=[]\n        for i in range(2):\n            for child in root_node.children:\n                temp,states=DFG_java(child,index_to_code,states)\n                DFG+=temp    \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_java(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_java(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n\ndef DFG_csharp(root_node,index_to_code,states):\n    assignment=['assignment_expression']\n    def_statement=['variable_declarator']\n    increment_statement=['postfix_unary_expression']\n    if_statement=['if_statement','else']\n    for_statement=['for_statement']\n    enhanced_for_statement=['for_each_statement']\n    while_statement=['while_statement']\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:\n        if len(root_node.children)==2:\n            name=root_node.children[0]\n            value=root_node.children[1]\n        else:\n            name=root_node.children[0]\n            value=None\n        DFG=[]\n        if value is None:\n            indexs=tree_to_variable_index(name,index_to_code)\n            for index in indexs:\n                idx,code=index_to_code[index]\n                DFG.append((code,idx,'comesFrom',[],[]))\n                states[code]=[idx]\n            return sorted(DFG,key=lambda x:x[1]),states\n        else:\n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)\n            temp,states=DFG_csharp(value,index_to_code,states)\n            DFG+=temp            \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'comesFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in assignment:\n        left_nodes=root_node.child_by_field_name('left')\n        right_nodes=root_node.child_by_field_name('right')\n        DFG=[]\n        temp,states=DFG_csharp(right_nodes,index_to_code,states)\n        DFG+=temp            \n        name_indexs=tree_to_variable_index(left_nodes,index_to_code)\n        value_indexs=tree_to_variable_index(right_nodes,index_to_code)        \n        for index1 in name_indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in value_indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]   \n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in increment_statement:\n        DFG=[]\n        indexs=tree_to_variable_index(root_node,index_to_code)\n        for index1 in indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]\n        return sorted(DFG,key=lambda x:x[1]),states   \n    elif root_node.type in if_statement:\n        DFG=[]\n        current_states=states.copy()\n        others_states=[]\n        flag=False\n        tag=False\n        if 'else' in root_node.type:\n            tag=True\n        for child in root_node.children:\n            if 'else' in child.type:\n                tag=True\n            if child.type not in if_statement and flag is False:\n                temp,current_states=DFG_csharp(child,index_to_code,current_states)\n                DFG+=temp\n            else:\n                flag=True\n# ... truncated ...","source_hash":"aa834b9452568da78d6e37907ac83c94801d9d9630febced21d6c19ec3e414f4","truncated":true}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_python","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_python#L11-L177","kind":"function","name":"DFG_python","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","language":"python","start_line":11,"end_line":177,"context_start_line":1,"context_end_line":197,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom tree_sitter import Language, Parser\nfrom .utils import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\n\n\ndef DFG_python(root_node,index_to_code,states):\n    assignment=['assignment','augmented_assignment','for_in_clause']\n    if_statement=['if_statement']\n    for_statement=['for_statement']\n    while_statement=['while_statement']\n    do_first_statement=['for_in_clause'] \n    def_statement=['default_parameter']\n    states=states.copy() \n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':        \n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:\n        name=root_node.child_by_field_name('name')\n        value=root_node.child_by_field_name('value')\n        DFG=[]\n        if value is None:\n            indexs=tree_to_variable_index(name,index_to_code)\n            for index in indexs:\n                idx,code=index_to_code[index]\n                DFG.append((code,idx,'comesFrom',[],[]))\n                states[code]=[idx]\n            return sorted(DFG,key=lambda x:x[1]),states\n        else:\n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)\n            temp,states=DFG_python(value,index_to_code,states)\n            DFG+=temp            \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'comesFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            return sorted(DFG,key=lambda x:x[1]),states        \n    elif root_node.type in assignment:\n        if root_node.type=='for_in_clause':\n            right_nodes=[root_node.children[-1]]\n            left_nodes=[root_node.child_by_field_name('left')]\n        else:\n            if root_node.child_by_field_name('right') is None:\n                return [],states\n            left_nodes=[x for x in root_node.child_by_field_name('left').children if x.type!=',']\n            right_nodes=[x for x in root_node.child_by_field_name('right').children if x.type!=',']\n            if len(right_nodes)!=len(left_nodes):\n                left_nodes=[root_node.child_by_field_name('left')]\n                right_nodes=[root_node.child_by_field_name('right')]\n            if len(left_nodes)==0:\n                left_nodes=[root_node.child_by_field_name('left')]\n            if len(right_nodes)==0:\n                right_nodes=[root_node.child_by_field_name('right')]\n        DFG=[]\n        for node in right_nodes:\n            temp,states=DFG_python(node,index_to_code,states)\n            DFG+=temp\n            \n        for left_node,right_node in zip(left_nodes,right_nodes):\n            left_tokens_index=tree_to_variable_index(left_node,index_to_code)\n            right_tokens_index=tree_to_variable_index(right_node,index_to_code)\n            temp=[]\n            for token1_index in left_tokens_index:\n                idx1,code1=index_to_code[token1_index]\n                temp.append((code1,idx1,'computedFrom',[index_to_code[x][1] for x in right_tokens_index],\n                             [index_to_code[x][0] for x in right_tokens_index]))\n                states[code1]=[idx1]\n            DFG+=temp        \n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in if_statement:\n        DFG=[]\n        current_states=states.copy()\n        others_states=[]\n        tag=False\n        if 'else' in root_node.type:\n            tag=True\n        for child in root_node.children:\n            if 'else' in child.type:\n                tag=True\n            if child.type not in ['elif_clause','else_clause']:\n                temp,current_states=DFG_python(child,index_to_code,current_states)\n                DFG+=temp\n            else:\n                temp,new_states=DFG_python(child,index_to_code,states)\n                DFG+=temp\n                others_states.append(new_states)\n        others_states.append(current_states)\n        if tag is False:\n            others_states.append(states)\n        new_states={}\n        for dic in others_states:\n            for key in dic:\n                if key not in new_states:\n                    new_states[key]=dic[key].copy()\n                else:\n                    new_states[key]+=dic[key]\n        for key in new_states:\n            new_states[key]=sorted(list(set(new_states[key])))\n        return sorted(DFG,key=lambda x:x[1]),new_states\n    elif root_node.type in for_statement:\n        DFG=[]\n        for i in range(2):\n            right_nodes=[x for x in root_node.child_by_field_name('right').children if x.type!=',']\n            left_nodes=[x for x in root_node.child_by_field_name('left').children if x.type!=',']\n            if len(right_nodes)!=len(left_nodes):\n                left_nodes=[root_node.child_by_field_name('left')]\n                right_nodes=[root_node.child_by_field_name('right')]\n            if len(left_nodes)==0:\n                left_nodes=[root_node.child_by_field_name('left')]\n            if len(right_nodes)==0:\n                right_nodes=[root_node.child_by_field_name('right')]\n            for node in right_nodes:\n                temp,states=DFG_python(node,index_to_code,states)\n                DFG+=temp\n            for left_node,right_node in zip(left_nodes,right_nodes):\n                left_tokens_index=tree_to_variable_index(left_node,index_to_code)\n                right_tokens_index=tree_to_variable_index(right_node,index_to_code)\n                temp=[]\n                for token1_index in left_tokens_index:\n                    idx1,code1=index_to_code[token1_index]\n                    temp.append((code1,idx1,'computedFrom',[index_to_code[x][1] for x in right_tokens_index],\n                                 [index_to_code[x][0] for x in right_tokens_index]))\n                    states[code1]=[idx1]\n                DFG+=temp   \n            if  root_node.children[-1].type==\"block\":\n                temp,states=DFG_python(root_node.children[-1],index_to_code,states)\n                DFG+=temp \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in while_statement:  \n        DFG=[]\n        for i in range(2):\n            for child in root_node.children:\n                temp,states=DFG_python(child,index_to_code,states)\n                DFG+=temp    \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_python(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_python(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n        \n\n######## OTHER LANGUAGES #########\n\ndef DFG_java(root_node,index_to_code,states):\n    assignment=['assignment_expression']\n    def_statement=['variable_declarator']\n    increment_statement=['update_expression']\n    if_statement=['if_statement','else']\n    for_statement=['for_statement']\n    enhanced_for_statement=['enhanced_for_statement']\n    while_statement=['while_statement']\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states","source_hash":"aa834b9452568da78d6e37907ac83c94801d9d9630febced21d6c19ec3e414f4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_java","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_java#L182-L356","kind":"function","name":"DFG_java","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","language":"python","start_line":182,"end_line":356,"context_start_line":162,"context_end_line":376,"code":"                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_python(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_python(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n        \n\n######## OTHER LANGUAGES #########\n\ndef DFG_java(root_node,index_to_code,states):\n    assignment=['assignment_expression']\n    def_statement=['variable_declarator']\n    increment_statement=['update_expression']\n    if_statement=['if_statement','else']\n    for_statement=['for_statement']\n    enhanced_for_statement=['enhanced_for_statement']\n    while_statement=['while_statement']\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:\n        name=root_node.child_by_field_name('name')\n        value=root_node.child_by_field_name('value')\n        DFG=[]\n        if value is None:\n            indexs=tree_to_variable_index(name,index_to_code)\n            for index in indexs:\n                idx,code=index_to_code[index]\n                DFG.append((code,idx,'comesFrom',[],[]))\n                states[code]=[idx]\n            return sorted(DFG,key=lambda x:x[1]),states\n        else:\n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)\n            temp,states=DFG_java(value,index_to_code,states)\n            DFG+=temp            \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'comesFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in assignment:\n        left_nodes=root_node.child_by_field_name('left')\n        right_nodes=root_node.child_by_field_name('right')\n        DFG=[]\n        temp,states=DFG_java(right_nodes,index_to_code,states)\n        DFG+=temp            \n        name_indexs=tree_to_variable_index(left_nodes,index_to_code)\n        value_indexs=tree_to_variable_index(right_nodes,index_to_code)        \n        for index1 in name_indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in value_indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]   \n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in increment_statement:\n        DFG=[]\n        indexs=tree_to_variable_index(root_node,index_to_code)\n        for index1 in indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]\n        return sorted(DFG,key=lambda x:x[1]),states   \n    elif root_node.type in if_statement:\n        DFG=[]\n        current_states=states.copy()\n        others_states=[]\n        flag=False\n        tag=False\n        if 'else' in root_node.type:\n            tag=True\n        for child in root_node.children:\n            if 'else' in child.type:\n                tag=True\n            if child.type not in if_statement and flag is False:\n                temp,current_states=DFG_java(child,index_to_code,current_states)\n                DFG+=temp\n            else:\n                flag=True\n                temp,new_states=DFG_java(child,index_to_code,states)\n                DFG+=temp\n                others_states.append(new_states)\n        others_states.append(current_states)\n        if tag is False:\n            others_states.append(states)\n        new_states={}\n        for dic in others_states:\n            for key in dic:\n                if key not in new_states:\n                    new_states[key]=dic[key].copy()\n                else:\n                    new_states[key]+=dic[key]\n        for key in new_states:\n            new_states[key]=sorted(list(set(new_states[key])))\n        return sorted(DFG,key=lambda x:x[1]),new_states\n    elif root_node.type in for_statement:\n        DFG=[]\n        for child in root_node.children:\n            temp,states=DFG_java(child,index_to_code,states)\n            DFG+=temp\n        flag=False\n        for child in root_node.children:\n            if flag:\n                temp,states=DFG_java(child,index_to_code,states)\n                DFG+=temp                \n            elif child.type==\"local_variable_declaration\":\n                flag=True\n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in enhanced_for_statement:\n        name=root_node.child_by_field_name('name')\n        value=root_node.child_by_field_name('value')\n        body=root_node.child_by_field_name('body')\n        DFG=[]\n        for i in range(2):\n            temp,states=DFG_java(value,index_to_code,states)\n            DFG+=temp       \n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)        \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            temp,states=DFG_java(body,index_to_code,states)\n            DFG+=temp                       \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in while_statement:  \n        DFG=[]\n        for i in range(2):\n            for child in root_node.children:\n                temp,states=DFG_java(child,index_to_code,states)\n                DFG+=temp    \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_java(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_java(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n\ndef DFG_csharp(root_node,index_to_code,states):\n    assignment=['assignment_expression']\n    def_statement=['variable_declarator']\n    increment_statement=['postfix_unary_expression']\n    if_statement=['if_statement','else']\n    for_statement=['for_statement']\n    enhanced_for_statement=['for_each_statement']\n    while_statement=['while_statement']\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]","source_hash":"aa834b9452568da78d6e37907ac83c94801d9d9630febced21d6c19ec3e414f4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_csharp","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_csharp#L358-L536","kind":"function","name":"DFG_csharp","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","language":"python","start_line":358,"end_line":536,"context_start_line":338,"context_end_line":556,"code":"            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_java(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_java(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n\ndef DFG_csharp(root_node,index_to_code,states):\n    assignment=['assignment_expression']\n    def_statement=['variable_declarator']\n    increment_statement=['postfix_unary_expression']\n    if_statement=['if_statement','else']\n    for_statement=['for_statement']\n    enhanced_for_statement=['for_each_statement']\n    while_statement=['while_statement']\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:\n        if len(root_node.children)==2:\n            name=root_node.children[0]\n            value=root_node.children[1]\n        else:\n            name=root_node.children[0]\n            value=None\n        DFG=[]\n        if value is None:\n            indexs=tree_to_variable_index(name,index_to_code)\n            for index in indexs:\n                idx,code=index_to_code[index]\n                DFG.append((code,idx,'comesFrom',[],[]))\n                states[code]=[idx]\n            return sorted(DFG,key=lambda x:x[1]),states\n        else:\n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)\n            temp,states=DFG_csharp(value,index_to_code,states)\n            DFG+=temp            \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'comesFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in assignment:\n        left_nodes=root_node.child_by_field_name('left')\n        right_nodes=root_node.child_by_field_name('right')\n        DFG=[]\n        temp,states=DFG_csharp(right_nodes,index_to_code,states)\n        DFG+=temp            \n        name_indexs=tree_to_variable_index(left_nodes,index_to_code)\n        value_indexs=tree_to_variable_index(right_nodes,index_to_code)        \n        for index1 in name_indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in value_indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]   \n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in increment_statement:\n        DFG=[]\n        indexs=tree_to_variable_index(root_node,index_to_code)\n        for index1 in indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]\n        return sorted(DFG,key=lambda x:x[1]),states   \n    elif root_node.type in if_statement:\n        DFG=[]\n        current_states=states.copy()\n        others_states=[]\n        flag=False\n        tag=False\n        if 'else' in root_node.type:\n            tag=True\n        for child in root_node.children:\n            if 'else' in child.type:\n                tag=True\n            if child.type not in if_statement and flag is False:\n                temp,current_states=DFG_csharp(child,index_to_code,current_states)\n                DFG+=temp\n            else:\n                flag=True\n                temp,new_states=DFG_csharp(child,index_to_code,states)\n                DFG+=temp\n                others_states.append(new_states)\n        others_states.append(current_states)\n        if tag is False:\n            others_states.append(states)\n        new_states={}\n        for dic in others_states:\n            for key in dic:\n                if key not in new_states:\n                    new_states[key]=dic[key].copy()\n                else:\n                    new_states[key]+=dic[key]\n        for key in new_states:\n            new_states[key]=sorted(list(set(new_states[key])))\n        return sorted(DFG,key=lambda x:x[1]),new_states\n    elif root_node.type in for_statement:\n        DFG=[]\n        for child in root_node.children:\n            temp,states=DFG_csharp(child,index_to_code,states)\n            DFG+=temp\n        flag=False\n        for child in root_node.children:\n            if flag:\n                temp,states=DFG_csharp(child,index_to_code,states)\n                DFG+=temp                \n            elif child.type==\"local_variable_declaration\":\n                flag=True\n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in enhanced_for_statement:\n        name=root_node.child_by_field_name('left')\n        value=root_node.child_by_field_name('right')\n        body=root_node.child_by_field_name('body')\n        DFG=[]\n        for i in range(2):\n            temp,states=DFG_csharp(value,index_to_code,states)\n            DFG+=temp       \n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)        \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            temp,states=DFG_csharp(body,index_to_code,states)\n            DFG+=temp                       \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in while_statement:  \n        DFG=[]\n        for i in range(2):\n            for child in root_node.children:\n                temp,states=DFG_csharp(child,index_to_code,states)\n                DFG+=temp    \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_csharp(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_csharp(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states  \n    \ndef DFG_ruby(root_node,index_to_code,states):\n    assignment=['assignment','operator_assignment']\n    if_statement=['if','elsif','else','unless','when']\n    for_statement=['for']\n    while_statement=['while_modifier','until']\n    do_first_statement=[] \n    def_statement=['keyword_parameter']\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        states=states.copy()\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:","source_hash":"aa834b9452568da78d6e37907ac83c94801d9d9630febced21d6c19ec3e414f4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_ruby","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_ruby#L538-L695","kind":"function","name":"DFG_ruby","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","language":"python","start_line":538,"end_line":695,"context_start_line":518,"context_end_line":715,"code":"            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_csharp(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_csharp(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states  \n    \ndef DFG_ruby(root_node,index_to_code,states):\n    assignment=['assignment','operator_assignment']\n    if_statement=['if','elsif','else','unless','when']\n    for_statement=['for']\n    while_statement=['while_modifier','until']\n    do_first_statement=[] \n    def_statement=['keyword_parameter']\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        states=states.copy()\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:\n        name=root_node.child_by_field_name('name')\n        value=root_node.child_by_field_name('value')\n        DFG=[]\n        if value is None:\n            indexs=tree_to_variable_index(name,index_to_code)\n            for index in indexs:\n                idx,code=index_to_code[index]\n                DFG.append((code,idx,'comesFrom',[],[]))\n                states[code]=[idx]\n            return sorted(DFG,key=lambda x:x[1]),states\n        else:\n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)\n            temp,states=DFG_ruby(value,index_to_code,states)\n            DFG+=temp            \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'comesFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            return sorted(DFG,key=lambda x:x[1]),states        \n    elif root_node.type in assignment:\n        left_nodes=[x for x in root_node.child_by_field_name('left').children if x.type!=',']\n        right_nodes=[x for x in root_node.child_by_field_name('right').children if x.type!=',']\n        if len(right_nodes)!=len(left_nodes):\n            left_nodes=[root_node.child_by_field_name('left')]\n            right_nodes=[root_node.child_by_field_name('right')]\n        if len(left_nodes)==0:\n            left_nodes=[root_node.child_by_field_name('left')]\n        if len(right_nodes)==0:\n            right_nodes=[root_node.child_by_field_name('right')]\n        if root_node.type==\"operator_assignment\":\n            left_nodes=[root_node.children[0]]\n            right_nodes=[root_node.children[-1]]\n\n        DFG=[]\n        for node in right_nodes:\n            temp,states=DFG_ruby(node,index_to_code,states)\n            DFG+=temp\n            \n        for left_node,right_node in zip(left_nodes,right_nodes):\n            left_tokens_index=tree_to_variable_index(left_node,index_to_code)\n            right_tokens_index=tree_to_variable_index(right_node,index_to_code)\n            temp=[]\n            for token1_index in left_tokens_index:\n                idx1,code1=index_to_code[token1_index]\n                temp.append((code1,idx1,'computedFrom',[index_to_code[x][1] for x in right_tokens_index],\n                             [index_to_code[x][0] for x in right_tokens_index]))\n                states[code1]=[idx1]\n            DFG+=temp        \n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in if_statement:\n        DFG=[]\n        current_states=states.copy()\n        others_states=[]\n        tag=False\n        if 'else' in root_node.type:\n            tag=True\n        for child in root_node.children:\n            if 'else' in child.type:\n                tag=True\n            if child.type not in if_statement:\n                temp,current_states=DFG_ruby(child,index_to_code,current_states)\n                DFG+=temp\n            else:\n                temp,new_states=DFG_ruby(child,index_to_code,states)\n                DFG+=temp\n                others_states.append(new_states)\n        others_states.append(current_states)\n        if tag is False:\n            others_states.append(states)\n        new_states={}\n        for dic in others_states:\n            for key in dic:\n                if key not in new_states:\n                    new_states[key]=dic[key].copy()\n                else:\n                    new_states[key]+=dic[key]\n        for key in new_states:\n            new_states[key]=sorted(list(set(new_states[key])))\n        return sorted(DFG,key=lambda x:x[1]),new_states\n    elif root_node.type in for_statement:\n        DFG=[]\n        for i in range(2):\n            left_nodes=[root_node.child_by_field_name('pattern')]\n            right_nodes=[root_node.child_by_field_name('value')]\n            assert len(right_nodes)==len(left_nodes)\n            for node in right_nodes:\n                temp,states=DFG_ruby(node,index_to_code,states)\n                DFG+=temp\n            for left_node,right_node in zip(left_nodes,right_nodes):\n                left_tokens_index=tree_to_variable_index(left_node,index_to_code)\n                right_tokens_index=tree_to_variable_index(right_node,index_to_code)\n                temp=[]\n                for token1_index in left_tokens_index:\n                    idx1,code1=index_to_code[token1_index]\n                    temp.append((code1,idx1,'computedFrom',[index_to_code[x][1] for x in right_tokens_index],\n                                 [index_to_code[x][0] for x in right_tokens_index]))\n                    states[code1]=[idx1]\n                DFG+=temp \n            temp,states=DFG_ruby(root_node.child_by_field_name('body'),index_to_code,states)\n            DFG+=temp \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in while_statement:  \n        DFG=[]\n        for i in range(2):\n            for child in root_node.children:\n                temp,states=DFG_ruby(child,index_to_code,states)\n                DFG+=temp    \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_ruby(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_ruby(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n\ndef DFG_go(root_node,index_to_code,states):\n    assignment=['assignment_statement',]\n    def_statement=['var_spec']\n    increment_statement=['inc_statement']\n    if_statement=['if_statement','else']\n    for_statement=['for_statement']\n    enhanced_for_statement=[]\n    while_statement=[]\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]","source_hash":"aa834b9452568da78d6e37907ac83c94801d9d9630febced21d6c19ec3e414f4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_go","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_go#L697-L837","kind":"function","name":"DFG_go","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","language":"python","start_line":697,"end_line":837,"context_start_line":677,"context_end_line":857,"code":"            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_ruby(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_ruby(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n\ndef DFG_go(root_node,index_to_code,states):\n    assignment=['assignment_statement',]\n    def_statement=['var_spec']\n    increment_statement=['inc_statement']\n    if_statement=['if_statement','else']\n    for_statement=['for_statement']\n    enhanced_for_statement=[]\n    while_statement=[]\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:\n        name=root_node.child_by_field_name('name')\n        value=root_node.child_by_field_name('value')\n        DFG=[]\n        if value is None:\n            indexs=tree_to_variable_index(name,index_to_code)\n            for index in indexs:\n                idx,code=index_to_code[index]\n                DFG.append((code,idx,'comesFrom',[],[]))\n                states[code]=[idx]\n            return sorted(DFG,key=lambda x:x[1]),states\n        else:\n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)\n            temp,states=DFG_go(value,index_to_code,states)\n            DFG+=temp            \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'comesFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in assignment:\n        left_nodes=root_node.child_by_field_name('left')\n        right_nodes=root_node.child_by_field_name('right')\n        DFG=[]\n        temp,states=DFG_go(right_nodes,index_to_code,states)\n        DFG+=temp            \n        name_indexs=tree_to_variable_index(left_nodes,index_to_code)\n        value_indexs=tree_to_variable_index(right_nodes,index_to_code)        \n        for index1 in name_indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in value_indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]   \n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in increment_statement:\n        DFG=[]\n        indexs=tree_to_variable_index(root_node,index_to_code)\n        for index1 in indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]\n        return sorted(DFG,key=lambda x:x[1]),states   \n    elif root_node.type in if_statement:\n        DFG=[]\n        current_states=states.copy()\n        others_states=[]\n        flag=False\n        tag=False\n        if 'else' in root_node.type:\n            tag=True\n        for child in root_node.children:\n            if 'else' in child.type:\n                tag=True\n            if child.type not in if_statement and flag is False:\n                temp,current_states=DFG_go(child,index_to_code,current_states)\n                DFG+=temp\n            else:\n                flag=True\n                temp,new_states=DFG_go(child,index_to_code,states)\n                DFG+=temp\n                others_states.append(new_states)\n        others_states.append(current_states)\n        if tag is False:\n            others_states.append(states)\n        new_states={}\n        for dic in others_states:\n            for key in dic:\n                if key not in new_states:\n                    new_states[key]=dic[key].copy()\n                else:\n                    new_states[key]+=dic[key]\n        for key in states:\n            if key not in new_states:\n                new_states[key]=states[key]\n            else:\n                new_states[key]+=states[key]\n        for key in new_states:\n            new_states[key]=sorted(list(set(new_states[key])))\n        return sorted(DFG,key=lambda x:x[1]),new_states\n    elif root_node.type in for_statement:\n        DFG=[]\n        for child in root_node.children:\n            temp,states=DFG_go(child,index_to_code,states)\n            DFG+=temp\n        flag=False\n        for child in root_node.children:\n            if flag:\n                temp,states=DFG_go(child,index_to_code,states)\n                DFG+=temp                \n            elif child.type==\"for_clause\":\n                if child.child_by_field_name('update') is not None:\n                    temp,states=DFG_go(child.child_by_field_name('update'),index_to_code,states)\n                    DFG+=temp                 \n                flag=True\n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_go(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_go(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n\ndef DFG_php(root_node,index_to_code,states):\n    assignment=['assignment_expression','augmented_assignment_expression']\n    def_statement=['simple_parameter']\n    increment_statement=['update_expression']\n    if_statement=['if_statement','else_clause']\n    for_statement=['for_statement']\n    enhanced_for_statement=['foreach_statement']\n    while_statement=['while_statement']\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]","source_hash":"aa834b9452568da78d6e37907ac83c94801d9d9630febced21d6c19ec3e414f4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_php","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_php#L839-L1022","kind":"function","name":"DFG_php","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","language":"python","start_line":839,"end_line":1022,"context_start_line":819,"context_end_line":1042,"code":"            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_go(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_go(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n\ndef DFG_php(root_node,index_to_code,states):\n    assignment=['assignment_expression','augmented_assignment_expression']\n    def_statement=['simple_parameter']\n    increment_statement=['update_expression']\n    if_statement=['if_statement','else_clause']\n    for_statement=['for_statement']\n    enhanced_for_statement=['foreach_statement']\n    while_statement=['while_statement']\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:\n        name=root_node.child_by_field_name('name')\n        value=root_node.child_by_field_name('default_value')\n        DFG=[]\n        if value is None:\n            indexs=tree_to_variable_index(name,index_to_code)\n            for index in indexs:\n                idx,code=index_to_code[index]\n                DFG.append((code,idx,'comesFrom',[],[]))\n                states[code]=[idx]\n            return sorted(DFG,key=lambda x:x[1]),states\n        else:\n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)\n            temp,states=DFG_php(value,index_to_code,states)\n            DFG+=temp            \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'comesFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in assignment:\n        left_nodes=root_node.child_by_field_name('left')\n        right_nodes=root_node.child_by_field_name('right')\n        DFG=[]\n        temp,states=DFG_php(right_nodes,index_to_code,states)\n        DFG+=temp            \n        name_indexs=tree_to_variable_index(left_nodes,index_to_code)\n        value_indexs=tree_to_variable_index(right_nodes,index_to_code)        \n        for index1 in name_indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in value_indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]   \n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in increment_statement:\n        DFG=[]\n        indexs=tree_to_variable_index(root_node,index_to_code)\n        for index1 in indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]\n        return sorted(DFG,key=lambda x:x[1]),states   \n    elif root_node.type in if_statement:\n        DFG=[]\n        current_states=states.copy()\n        others_states=[]\n        flag=False\n        tag=False\n        if 'else' in root_node.type:\n            tag=True\n        for child in root_node.children:\n            if 'else' in child.type:\n                tag=True\n            if child.type not in if_statement and flag is False:\n                temp,current_states=DFG_php(child,index_to_code,current_states)\n                DFG+=temp\n            else:\n                flag=True\n                temp,new_states=DFG_php(child,index_to_code,states)\n                DFG+=temp\n                others_states.append(new_states)\n        others_states.append(current_states)\n        new_states={}\n        for dic in others_states:\n            for key in dic:\n                if key not in new_states:\n                    new_states[key]=dic[key].copy()\n                else:\n                    new_states[key]+=dic[key]\n        for key in states:\n            if key not in new_states:\n                new_states[key]=states[key]\n            else:\n                new_states[key]+=states[key]\n        for key in new_states:\n            new_states[key]=sorted(list(set(new_states[key])))\n        return sorted(DFG,key=lambda x:x[1]),new_states\n    elif root_node.type in for_statement:\n        DFG=[]\n        for child in root_node.children:\n            temp,states=DFG_php(child,index_to_code,states)\n            DFG+=temp\n        flag=False\n        for child in root_node.children:\n            if flag:\n                temp,states=DFG_php(child,index_to_code,states)\n                DFG+=temp                \n            elif child.type==\"assignment_expression\":               \n                flag=True\n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in enhanced_for_statement:\n        name=None\n        value=None\n        for child in root_node.children:\n            if child.type=='variable_name' and value is None:\n                value=child\n            elif child.type=='variable_name' and name is None:\n                name=child\n                break\n        body=root_node.child_by_field_name('body')\n        DFG=[]\n        for i in range(2):\n            temp,states=DFG_php(value,index_to_code,states)\n            DFG+=temp       \n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)        \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            temp,states=DFG_php(body,index_to_code,states)\n            DFG+=temp                       \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in while_statement:  \n        DFG=[]\n        for i in range(2):\n            for child in root_node.children:\n                temp,states=DFG_php(child,index_to_code,states)\n                DFG+=temp    \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_php(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_php(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n\ndef DFG_javascript(root_node,index_to_code,states):\n    assignment=['assignment_pattern','augmented_assignment_expression']\n    def_statement=['variable_declarator']\n    increment_statement=['update_expression']\n    if_statement=['if_statement','else']\n    for_statement=['for_statement']\n    enhanced_for_statement=[]\n    while_statement=['while_statement']\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]","source_hash":"aa834b9452568da78d6e37907ac83c94801d9d9630febced21d6c19ec3e414f4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_javascript","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.DFG.DFG_javascript#L1024-L1176","kind":"function","name":"DFG_javascript","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","language":"python","start_line":1024,"end_line":1176,"context_start_line":1004,"context_end_line":1179,"code":"            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states        \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_php(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_php(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n\ndef DFG_javascript(root_node,index_to_code,states):\n    assignment=['assignment_pattern','augmented_assignment_expression']\n    def_statement=['variable_declarator']\n    increment_statement=['update_expression']\n    if_statement=['if_statement','else']\n    for_statement=['for_statement']\n    enhanced_for_statement=[]\n    while_statement=['while_statement']\n    do_first_statement=[]    \n    states=states.copy()\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:\n            return [],states\n        elif code in states:\n            return [(code,idx,'comesFrom',[code],states[code].copy())],states\n        else:\n            if root_node.type=='identifier':\n                states[code]=[idx]\n            return [(code,idx,'comesFrom',[],[])],states\n    elif root_node.type in def_statement:\n        name=root_node.child_by_field_name('name')\n        value=root_node.child_by_field_name('value')\n        DFG=[]\n        if value is None:\n            indexs=tree_to_variable_index(name,index_to_code)\n            for index in indexs:\n                idx,code=index_to_code[index]\n                DFG.append((code,idx,'comesFrom',[],[]))\n                states[code]=[idx]\n            return sorted(DFG,key=lambda x:x[1]),states\n        else:\n            name_indexs=tree_to_variable_index(name,index_to_code)\n            value_indexs=tree_to_variable_index(value,index_to_code)\n            temp,states=DFG_javascript(value,index_to_code,states)\n            DFG+=temp            \n            for index1 in name_indexs:\n                idx1,code1=index_to_code[index1]\n                for index2 in value_indexs:\n                    idx2,code2=index_to_code[index2]\n                    DFG.append((code1,idx1,'comesFrom',[code2],[idx2]))\n                states[code1]=[idx1]   \n            return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in assignment:\n        left_nodes=root_node.child_by_field_name('left')\n        right_nodes=root_node.child_by_field_name('right')\n        DFG=[]\n        temp,states=DFG_javascript(right_nodes,index_to_code,states)\n        DFG+=temp            \n        name_indexs=tree_to_variable_index(left_nodes,index_to_code)\n        value_indexs=tree_to_variable_index(right_nodes,index_to_code)        \n        for index1 in name_indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in value_indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]   \n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in increment_statement:\n        DFG=[]\n        indexs=tree_to_variable_index(root_node,index_to_code)\n        for index1 in indexs:\n            idx1,code1=index_to_code[index1]\n            for index2 in indexs:\n                idx2,code2=index_to_code[index2]\n                DFG.append((code1,idx1,'computedFrom',[code2],[idx2]))\n            states[code1]=[idx1]\n        return sorted(DFG,key=lambda x:x[1]),states   \n    elif root_node.type in if_statement:\n        DFG=[]\n        current_states=states.copy()\n        others_states=[]\n        flag=False\n        tag=False\n        if 'else' in root_node.type:\n            tag=True\n        for child in root_node.children:\n            if 'else' in child.type:\n                tag=True\n            if child.type not in if_statement and flag is False:\n                temp,current_states=DFG_javascript(child,index_to_code,current_states)\n                DFG+=temp\n            else:\n                flag=True\n                temp,new_states=DFG_javascript(child,index_to_code,states)\n                DFG+=temp\n                others_states.append(new_states)\n        others_states.append(current_states)\n        if tag is False:\n            others_states.append(states)        \n        new_states={}\n        for dic in others_states:\n            for key in dic:\n                if key not in new_states:\n                    new_states[key]=dic[key].copy()\n                else:\n                    new_states[key]+=dic[key]\n        for key in states:\n            if key not in new_states:\n                new_states[key]=states[key]\n            else:\n                new_states[key]+=states[key]\n        for key in new_states:\n            new_states[key]=sorted(list(set(new_states[key])))\n        return sorted(DFG,key=lambda x:x[1]),new_states\n    elif root_node.type in for_statement:\n        DFG=[]\n        for child in root_node.children:\n            temp,states=DFG_javascript(child,index_to_code,states)\n            DFG+=temp\n        flag=False\n        for child in root_node.children:\n            if flag:\n                temp,states=DFG_javascript(child,index_to_code,states)\n                DFG+=temp                \n            elif child.type==\"variable_declaration\":               \n                flag=True\n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states\n    elif root_node.type in while_statement:  \n        DFG=[]\n        for i in range(2):\n            for child in root_node.children:\n                temp,states=DFG_javascript(child,index_to_code,states)\n                DFG+=temp    \n        dic={}\n        for x in DFG:\n            if (x[0],x[1],x[2]) not in dic:\n                dic[(x[0],x[1],x[2])]=[x[3],x[4]]\n            else:\n                dic[(x[0],x[1],x[2])][0]=list(set(dic[(x[0],x[1],x[2])][0]+x[3]))\n                dic[(x[0],x[1],x[2])][1]=sorted(list(set(dic[(x[0],x[1],x[2])][1]+x[4])))\n        DFG=[(x[0],x[1],x[2],y[0],y[1]) for x,y in sorted(dic.items(),key=lambda t:t[0][1])]\n        return sorted(DFG,key=lambda x:x[1]),states    \n    else:\n        DFG=[]\n        for child in root_node.children:\n            if child.type in do_first_statement:\n                temp,states=DFG_javascript(child,index_to_code,states)\n                DFG+=temp\n        for child in root_node.children:\n            if child.type not in do_first_statement:\n                temp,states=DFG_javascript(child,index_to_code,states)\n                DFG+=temp\n        \n        return sorted(DFG,key=lambda x:x[1]),states\n\n\n     ","source_hash":"aa834b9452568da78d6e37907ac83c94801d9d9630febced21d6c19ec3e414f4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils#L1-L102","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/utils.py","language":"python","start_line":1,"end_line":102,"context_start_line":1,"context_end_line":102,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nimport re\nfrom io import StringIO\nimport  tokenize\n\ndef remove_comments_and_docstrings(source,lang):\n    if lang in ['python']:\n        \"\"\"\n        Returns 'source' minus comments and docstrings.\n        \"\"\"\n        io_obj = StringIO(source)\n        out = \"\"\n        prev_toktype = tokenize.INDENT\n        last_lineno = -1\n        last_col = 0\n        for tok in tokenize.generate_tokens(io_obj.readline):\n            token_type = tok[0]\n            token_string = tok[1]\n            start_line, start_col = tok[2]\n            end_line, end_col = tok[3]\n            ltext = tok[4]\n            if start_line > last_lineno:\n                last_col = 0\n            if start_col > last_col:\n                out += (\" \" * (start_col - last_col))\n            # Remove comments:\n            if token_type == tokenize.COMMENT:\n                pass\n            # This series of conditionals removes docstrings:\n            elif token_type == tokenize.STRING:\n                if prev_toktype != tokenize.INDENT:\n            # This is likely a docstring; double-check we're not inside an operator:\n                    if prev_toktype != tokenize.NEWLINE:\n                        if start_col > 0:\n                            out += token_string\n            else:\n                out += token_string\n            prev_toktype = token_type\n            last_col = end_col\n            last_lineno = end_line\n        temp=[]\n        for x in out.split('\\n'):\n            if x.strip()!=\"\":\n                temp.append(x)\n        return '\\n'.join(temp)\n    elif lang in ['ruby']:\n        return source\n    else:\n        def replacer(match):\n            s = match.group(0)\n            if s.startswith('/'):\n                return \" \" # note: a space and not an empty string\n            else:\n                return s\n        pattern = re.compile(\n            r'//.*?$|/\\*.*?\\*/|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"',\n            re.DOTALL | re.MULTILINE\n        )\n        temp=[]\n        for x in re.sub(pattern, replacer, source).split('\\n'):\n            if x.strip()!=\"\":\n                temp.append(x)\n        return '\\n'.join(temp)\n\ndef tree_to_token_index(root_node):\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        return [(root_node.start_point,root_node.end_point)]\n    else:\n        code_tokens=[]\n        for child in root_node.children:\n            code_tokens+=tree_to_token_index(child)\n        return code_tokens\n    \ndef tree_to_variable_index(root_node,index_to_code):\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        index=(root_node.start_point,root_node.end_point)\n        _,code=index_to_code[index]\n        if root_node.type!=code:\n            return [(root_node.start_point,root_node.end_point)]\n        else:\n            return []\n    else:\n        code_tokens=[]\n        for child in root_node.children:\n            code_tokens+=tree_to_variable_index(child,index_to_code)\n        return code_tokens    \n\ndef index_to_code_token(index,code):\n    start_point=index[0]\n    end_point=index[1]\n    if start_point[0]==end_point[0]:\n        s=code[start_point[0]][start_point[1]:end_point[1]]\n    else:\n        s=\"\"\n        s+=code[start_point[0]][start_point[1]:]\n        for i in range(start_point[0]+1,end_point[0]):\n            s+=code[i]\n        s+=code[end_point[0]][:end_point[1]]   \n    return s\n   ","source_hash":"2ecb4ed83e010e05ab32e112e2cdae4eb0c050a5e1b8ecf0eb0533565c17c479","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils.remove_comments_and_docstrings","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils.remove_comments_and_docstrings#L8-L65","kind":"function","name":"remove_comments_and_docstrings","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/utils.py","language":"python","start_line":8,"end_line":65,"context_start_line":1,"context_end_line":85,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nimport re\nfrom io import StringIO\nimport  tokenize\n\ndef remove_comments_and_docstrings(source,lang):\n    if lang in ['python']:\n        \"\"\"\n        Returns 'source' minus comments and docstrings.\n        \"\"\"\n        io_obj = StringIO(source)\n        out = \"\"\n        prev_toktype = tokenize.INDENT\n        last_lineno = -1\n        last_col = 0\n        for tok in tokenize.generate_tokens(io_obj.readline):\n            token_type = tok[0]\n            token_string = tok[1]\n            start_line, start_col = tok[2]\n            end_line, end_col = tok[3]\n            ltext = tok[4]\n            if start_line > last_lineno:\n                last_col = 0\n            if start_col > last_col:\n                out += (\" \" * (start_col - last_col))\n            # Remove comments:\n            if token_type == tokenize.COMMENT:\n                pass\n            # This series of conditionals removes docstrings:\n            elif token_type == tokenize.STRING:\n                if prev_toktype != tokenize.INDENT:\n            # This is likely a docstring; double-check we're not inside an operator:\n                    if prev_toktype != tokenize.NEWLINE:\n                        if start_col > 0:\n                            out += token_string\n            else:\n                out += token_string\n            prev_toktype = token_type\n            last_col = end_col\n            last_lineno = end_line\n        temp=[]\n        for x in out.split('\\n'):\n            if x.strip()!=\"\":\n                temp.append(x)\n        return '\\n'.join(temp)\n    elif lang in ['ruby']:\n        return source\n    else:\n        def replacer(match):\n            s = match.group(0)\n            if s.startswith('/'):\n                return \" \" # note: a space and not an empty string\n            else:\n                return s\n        pattern = re.compile(\n            r'//.*?$|/\\*.*?\\*/|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"',\n            re.DOTALL | re.MULTILINE\n        )\n        temp=[]\n        for x in re.sub(pattern, replacer, source).split('\\n'):\n            if x.strip()!=\"\":\n                temp.append(x)\n        return '\\n'.join(temp)\n\ndef tree_to_token_index(root_node):\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        return [(root_node.start_point,root_node.end_point)]\n    else:\n        code_tokens=[]\n        for child in root_node.children:\n            code_tokens+=tree_to_token_index(child)\n        return code_tokens\n    \ndef tree_to_variable_index(root_node,index_to_code):\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        index=(root_node.start_point,root_node.end_point)\n        _,code=index_to_code[index]\n        if root_node.type!=code:\n            return [(root_node.start_point,root_node.end_point)]\n        else:\n            return []\n    else:\n        code_tokens=[]","source_hash":"2ecb4ed83e010e05ab32e112e2cdae4eb0c050a5e1b8ecf0eb0533565c17c479","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils.tree_to_token_index","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils.tree_to_token_index#L67-L74","kind":"function","name":"tree_to_token_index","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/utils.py","language":"python","start_line":67,"end_line":74,"context_start_line":47,"context_end_line":94,"code":"        return '\\n'.join(temp)\n    elif lang in ['ruby']:\n        return source\n    else:\n        def replacer(match):\n            s = match.group(0)\n            if s.startswith('/'):\n                return \" \" # note: a space and not an empty string\n            else:\n                return s\n        pattern = re.compile(\n            r'//.*?$|/\\*.*?\\*/|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"',\n            re.DOTALL | re.MULTILINE\n        )\n        temp=[]\n        for x in re.sub(pattern, replacer, source).split('\\n'):\n            if x.strip()!=\"\":\n                temp.append(x)\n        return '\\n'.join(temp)\n\ndef tree_to_token_index(root_node):\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        return [(root_node.start_point,root_node.end_point)]\n    else:\n        code_tokens=[]\n        for child in root_node.children:\n            code_tokens+=tree_to_token_index(child)\n        return code_tokens\n    \ndef tree_to_variable_index(root_node,index_to_code):\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        index=(root_node.start_point,root_node.end_point)\n        _,code=index_to_code[index]\n        if root_node.type!=code:\n            return [(root_node.start_point,root_node.end_point)]\n        else:\n            return []\n    else:\n        code_tokens=[]\n        for child in root_node.children:\n            code_tokens+=tree_to_variable_index(child,index_to_code)\n        return code_tokens    \n\ndef index_to_code_token(index,code):\n    start_point=index[0]\n    end_point=index[1]\n    if start_point[0]==end_point[0]:\n        s=code[start_point[0]][start_point[1]:end_point[1]]","source_hash":"2ecb4ed83e010e05ab32e112e2cdae4eb0c050a5e1b8ecf0eb0533565c17c479","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils.tree_to_variable_index","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils.tree_to_variable_index#L76-L88","kind":"function","name":"tree_to_variable_index","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/utils.py","language":"python","start_line":76,"end_line":88,"context_start_line":56,"context_end_line":102,"code":"                return s\n        pattern = re.compile(\n            r'//.*?$|/\\*.*?\\*/|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"',\n            re.DOTALL | re.MULTILINE\n        )\n        temp=[]\n        for x in re.sub(pattern, replacer, source).split('\\n'):\n            if x.strip()!=\"\":\n                temp.append(x)\n        return '\\n'.join(temp)\n\ndef tree_to_token_index(root_node):\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        return [(root_node.start_point,root_node.end_point)]\n    else:\n        code_tokens=[]\n        for child in root_node.children:\n            code_tokens+=tree_to_token_index(child)\n        return code_tokens\n    \ndef tree_to_variable_index(root_node,index_to_code):\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        index=(root_node.start_point,root_node.end_point)\n        _,code=index_to_code[index]\n        if root_node.type!=code:\n            return [(root_node.start_point,root_node.end_point)]\n        else:\n            return []\n    else:\n        code_tokens=[]\n        for child in root_node.children:\n            code_tokens+=tree_to_variable_index(child,index_to_code)\n        return code_tokens    \n\ndef index_to_code_token(index,code):\n    start_point=index[0]\n    end_point=index[1]\n    if start_point[0]==end_point[0]:\n        s=code[start_point[0]][start_point[1]:end_point[1]]\n    else:\n        s=\"\"\n        s+=code[start_point[0]][start_point[1]:]\n        for i in range(start_point[0]+1,end_point[0]):\n            s+=code[i]\n        s+=code[end_point[0]][:end_point[1]]   \n    return s\n   ","source_hash":"2ecb4ed83e010e05ab32e112e2cdae4eb0c050a5e1b8ecf0eb0533565c17c479","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils.index_to_code_token","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils.index_to_code_token#L90-L101","kind":"function","name":"index_to_code_token","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/utils.py","language":"python","start_line":90,"end_line":101,"context_start_line":70,"context_end_line":102,"code":"    else:\n        code_tokens=[]\n        for child in root_node.children:\n            code_tokens+=tree_to_token_index(child)\n        return code_tokens\n    \ndef tree_to_variable_index(root_node,index_to_code):\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        index=(root_node.start_point,root_node.end_point)\n        _,code=index_to_code[index]\n        if root_node.type!=code:\n            return [(root_node.start_point,root_node.end_point)]\n        else:\n            return []\n    else:\n        code_tokens=[]\n        for child in root_node.children:\n            code_tokens+=tree_to_variable_index(child,index_to_code)\n        return code_tokens    \n\ndef index_to_code_token(index,code):\n    start_point=index[0]\n    end_point=index[1]\n    if start_point[0]==end_point[0]:\n        s=code[start_point[0]][start_point[1]:end_point[1]]\n    else:\n        s=\"\"\n        s+=code[start_point[0]][start_point[1]:]\n        for i in range(start_point[0]+1,end_point[0]):\n            s+=code[i]\n        s+=code[end_point[0]][:end_point[1]]   \n    return s\n   ","source_hash":"2ecb4ed83e010e05ab32e112e2cdae4eb0c050a5e1b8ecf0eb0533565c17c479","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils.replacer","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.utils.replacer#L51-L56","kind":"function","name":"replacer","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/utils.py","language":"python","start_line":51,"end_line":56,"context_start_line":31,"context_end_line":76,"code":"            # This series of conditionals removes docstrings:\n            elif token_type == tokenize.STRING:\n                if prev_toktype != tokenize.INDENT:\n            # This is likely a docstring; double-check we're not inside an operator:\n                    if prev_toktype != tokenize.NEWLINE:\n                        if start_col > 0:\n                            out += token_string\n            else:\n                out += token_string\n            prev_toktype = token_type\n            last_col = end_col\n            last_lineno = end_line\n        temp=[]\n        for x in out.split('\\n'):\n            if x.strip()!=\"\":\n                temp.append(x)\n        return '\\n'.join(temp)\n    elif lang in ['ruby']:\n        return source\n    else:\n        def replacer(match):\n            s = match.group(0)\n            if s.startswith('/'):\n                return \" \" # note: a space and not an empty string\n            else:\n                return s\n        pattern = re.compile(\n            r'//.*?$|/\\*.*?\\*/|\\'(?:\\\\.|[^\\\\\\'])*\\'|\"(?:\\\\.|[^\\\\\"])*\"',\n            re.DOTALL | re.MULTILINE\n        )\n        temp=[]\n        for x in re.sub(pattern, replacer, source).split('\\n'):\n            if x.strip()!=\"\":\n                temp.append(x)\n        return '\\n'.join(temp)\n\ndef tree_to_token_index(root_node):\n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':\n        return [(root_node.start_point,root_node.end_point)]\n    else:\n        code_tokens=[]\n        for child in root_node.children:\n            code_tokens+=tree_to_token_index(child)\n        return code_tokens\n    \ndef tree_to_variable_index(root_node,index_to_code):","source_hash":"2ecb4ed83e010e05ab32e112e2cdae4eb0c050a5e1b8ecf0eb0533565c17c479","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.build","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.build#L1-L14","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.build","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/build.py","language":"python","start_line":1,"end_line":14,"context_start_line":1,"context_end_line":14,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom tree_sitter import Language, Parser\n\nLanguage.build_library(\n  # Store the library in the `build` directory\n  'my-languages.so',\n\n  # Include one or more languages\n  [ 'tree-sitter-python',\n  ]\n)\n","source_hash":"2bbff97589bc4dae551a425cfad854eb5cf1275bd35c236866208c13c8e52187","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.parameters","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.parameters#L1-L4","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.parameters","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/parameters.py","language":"python","start_line":1,"end_line":4,"context_start_line":1,"context_end_line":4,"code":"def g(h, i, /, j, *, k=100, **kwarg):\n    #       ^ operator\n    #             ^ operator\n    pass","source_hash":"71aeb9a20f6752310f9256cb0c5a7abf29a8e01aeb49f4857c6af0362ee0c072","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.parameters.g","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.parameters.g#L1-L4","kind":"function","name":"g","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/parameters.py","language":"python","start_line":1,"end_line":4,"context_start_line":1,"context_end_line":4,"code":"def g(h, i, /, j, *, k=100, **kwarg):\n    #       ^ operator\n    #             ^ operator\n    pass","source_hash":"71aeb9a20f6752310f9256cb0c5a7abf29a8e01aeb49f4857c6af0362ee0c072","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.keywords","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.keywords#L1-L30","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.keywords","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/keywords.py","language":"python","start_line":1,"end_line":30,"context_start_line":1,"context_end_line":30,"code":"if foo():\n# <- keyword\n    pass\n    # <- keyword\nelif bar():\n# <- keyword\n    pass\nelse:\n# <- keyword\n    foo\n\nreturn\n# ^ keyword\nraise e\n# ^ keyword\n\nfor i in foo():\n# <- keyword\n#   ^ variable\n#     ^ operator\n#        ^ function\n    continue\n    # <- keyword\n    break\n    # <- keyword\n\na and b or c\n# ^ operator\n#     ^ variable\n#       ^ operator","source_hash":"5209cfe62baefa489f21f00ca457880a0908c54e3613ab5f73fb0d568e3d9d69","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.pattern_matching","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.pattern_matching#L1-L54","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.test.highlight.pattern_matching","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/pattern_matching.py","language":"python","start_line":1,"end_line":54,"context_start_line":1,"context_end_line":54,"code":"match command.split():\n# ^ keyword\n    case [\"quit\"]:\n    # ^ keyword\n        print(\"Goodbye!\")\n        quit_game()\n    case [\"look\"]:\n    # ^ keyword\n        current_room.describe()\n    case [\"get\", obj]:\n    # ^ keyword\n        character.get(obj, current_room)\n    case [\"go\", direction]:\n    # ^ keyword\n        current_room = current_room.neighbor(direction)\n    # The rest of your commands go here\n\nmatch command.split():\n# ^ keyword\n    case [\"drop\", *objects]:\n    # ^ keyword\n        for obj in objects:\n            character.drop(obj, current_room)\n\nmatch command.split():\n# ^ keyword\n    case [\"quit\"]: ... # Code omitted for brevity\n    case [\"go\", direction]: pass\n    case [\"drop\", *objects]: pass\n    case _:\n        print(f\"Sorry, I couldn't understand {command!r}\")\n\nmatch command.split():\n# ^ keyword\n    case [\"north\"] | [\"go\", \"north\"]:\n    # ^ keyword\n        current_room = current_room.neighbor(\"north\")\n    case [\"get\", obj] | [\"pick\", \"up\", obj] | [\"pick\", obj, \"up\"]:\n    # ^ keyword\n        pass\n\nmatch = 2\n#   ^ variable\nmatch, a = 2, 3\n#   ^ variable\nmatch: int = secret\n#   ^ variable\nx, match: str = 2, \"hey, what's up?\"\n# <- variable\n#   ^ variable\n\nif match := re.fullmatch(r\"(-)?(\\d+:)?\\d?\\d:\\d\\d(\\.\\d*)?\", time, flags=re.ASCII):\n    # ^ variable\n    return match","source_hash":"9f5c10ed608893ffcbc0d85ea94ab67e50f903c25aee2982ad29f806f8298998","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.multiple-newlines","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.multiple-newlines#L1-L25","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.multiple-newlines","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/multiple-newlines.py","language":"python","start_line":1,"end_line":25,"context_start_line":1,"context_end_line":25,"code":"def hi():\n\n\n\n    print \"hi\"\n\n\ndef bye():\n    print \"bye\"\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n","source_hash":"43b9ea50f59013b0204eac91c14ebccd6e8507780923516f7f0d40146adcbb68","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python2-grammar-crlf","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python2-grammar-crlf#L1-L973","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python2-grammar-crlf","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python2-grammar-crlf.py","language":"python","start_line":1,"end_line":973,"context_start_line":1,"context_end_line":973,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.test_support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\\n        + 1\n        self.assertEquals(x, 2, 'backslash for line continuation')\n\n        # Backslash does not means continuation in comments :\\\n        x = 0\n        self.assertEquals(x, 0, 'backslash ending comment')\n\n    def testPlainIntegers(self):\n        self.assertEquals(0xff, 255)\n        self.assertEquals(0377, 255)\n        self.assertEquals(2147483647, 017777777777)\n        # \"0x\" is not a valid literal\n        self.assertRaises(SyntaxError, eval, \"0x\")\n        from sys import maxint\n        if maxint == 2147483647:\n            self.assertEquals(-2147483647-1, -020000000000)\n            # XXX -2147483648\n            self.assert_(037777777777 > 0)\n            self.assert_(0xffffffff > 0)\n            for s in '2147483648', '040000000000', '0x100000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        elif maxint == 9223372036854775807:\n            self.assertEquals(-9223372036854775807-1, -01000000000000000000000)\n            self.assert_(01777777777777777777777 > 0)\n            self.assert_(0xffffffffffffffff > 0)\n            for s in '9223372036854775808', '02000000000000000000000','0x10000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxint value %r' % maxint)\n\n    def testLongIntegers(self):\n        x = 0L\n        x = 0l\n        x = 0xffffffffffffffffL\n        x = 0xffffffffffffffffl\n        x = 077777777777777777L\n        x = 077777777777777777l\n        x = 123456789012345678901234567890L\n        x = 123456789012345678901234567890l\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### 'def' NAME parameters ':' suite\n        ### parameters: '(' [varargslist] ')'\n        ### varargslist: (fpdef ['=' test] ',')* ('*' NAME [',' ('**'|'*' '*') NAME]\n        ###            | ('**'|'*' '*') NAME)\n        ###            | fpdef ['=' test] (',' fpdef ['=' test])* [',']\n        ### fpdef: NAME | '(' fplist ')'\n        ### fplist: fpdef (',' fpdef)* [',']\n        ### arglist: (argument ',')* (argument | *' test [',' '**' test] | '**' test)\n        ### argument: [test '='] test   # Really [keyword '='] test\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        def f4(two, (compound, (argument, list))): pass\n        def f5((compound, first), two): pass\n        self.assertEquals(f2.func_code.co_varnames, ('one_argument',))\n        self.assertEquals(f3.func_code.co_varnames, ('two', 'arguments'))\n        if sys.platform.startswith('java'):\n            self.assertEquals(f4.func_code.co_varnames,\n                   ('two', '(compound, (argument, list))', 'compound', 'argument',\n                                'list',))\n            self.assertEquals(f5.func_code.co_varnames,\n                   ('(compound, first)', 'two', 'compound', 'first'))\n        else:\n            self.assertEquals(f4.func_code.co_varnames,\n                  ('two', '.1', 'compound', 'argument',  'list'))\n            self.assertEquals(f5.func_code.co_varnames,\n                  ('.0', 'two', 'compound', 'first'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n        def v3(a, (b, c), *rest): return a, b, c, rest\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        f4(1, (2, (3, 4)))\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n        v3(1,(2,3))\n        v3(1,(2,3),4)\n        v3(1,(2,3),4,5,6,7,8,9,0)\n\n        # ceval unpacks the formal arguments into the first argcount names;\n        # thus, the names nested inside tuples must appear after these names.\n        if sys.platform.startswith('java'):\n            self.assertEquals(v3.func_code.co_varnames, ('a', '(b, c)', 'rest', 'b', 'c'))\n        else:\n            self.assertEquals(v3.func_code.co_varnames, ('a', '.1', 'rest', 'b', 'c'))\n        self.assertEquals(v3(1, (2, 3), 4), (1, 2, 3, (4,)))\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n        def d31v((x)): pass\n        d31v(1)\n        def d32v((x,)): pass\n        d32v((1,))\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0L]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | print_stmt  | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt | exec_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testPrintStmt(self):\n        # 'print' (test ',')* [test]\n        import StringIO\n\n        # Can't test printing to real stdout without comparing output\n        # which is not available in unittest.\n        save_stdout = sys.stdout\n        sys.stdout = StringIO.StringIO()\n\n        print 1, 2, 3\n        print 1, 2, 3,\n        print\n        print 0 or 1, 0 or 1,\n        print 0 or 1\n\n        # 'print' '>>' test ','\n        print >> sys.stdout, 1, 2, 3\n        print >> sys.stdout, 1, 2, 3,\n        print >> sys.stdout\n        print >> sys.stdout, 0 or 1, 0 or 1,\n        print >> sys.stdout, 0 or 1\n\n        # test printing to an instance\n        class Gulp:\n            def write(self, msg): pass\n\n        gulp = Gulp()\n        print >> gulp, 1, 2, 3\n        print >> gulp, 1, 2, 3,\n        print >> gulp\n        print >> gulp, 0 or 1, 0 or 1,\n        print >> gulp, 0 or 1\n\n        # test print >> None\n        def driver():\n            oldstdout = sys.stdout\n            sys.stdout = Gulp()\n            try:\n                tellme(Gulp())\n                tellme()\n            finally:\n                sys.stdout = oldstdout\n\n        # we should see this once\n        def tellme(file=sys.stdout):\n            print >> file, 'hello world'\n\n        driver()\n\n        # we should not see this at all\n        def tellme(file=None):\n            print >> file, 'goodbye universe'\n\n        driver()\n\n        self.assertEqual(sys.stdout.getvalue(), '''\\\n1 2 3\n1 2 3\n1 1 1\n1 2 3\n1 2 3\n1 1 1\nhello world\n''')\n        sys.stdout = save_stdout\n\n        # syntax errors\n        check_syntax_error(self, 'print ,')\n        check_syntax_error(self, 'print >> x,')\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo <> 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError, 'just testing'\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testExec(self):\n        # 'exec' expr ['in' expr [',' expr]]\n        z = None\n        del z\n        exec 'z=1+1\\n'\n        if z != 2: self.fail('exec \\'z=1+1\\'\\\\n')\n        del z\n        exec 'z=1+1'\n        if z != 2: self.fail('exec \\'z=1+1\\'')\n        z = None\n        del z\n        import types\n        if hasattr(types, \"UnicodeType\"):\n            exec r\"\"\"if 1:\n            exec u'z=1+1\\n'\n            if z != 2: self.fail('exec u\\'z=1+1\\'\\\\n')\n            del z\n            exec u'z=1+1'\n            if z != 2: self.fail('exec u\\'z=1+1\\'')\"\"\"\n        g = {}\n        exec 'z = 1' in g\n        if g.has_key('__builtins__'): del g['__builtins__']\n        if g != {'z': 1}: self.fail('exec \\'z = 1\\' in g')\n        g = {}\n        l = {}\n\n        import warnings\n        warnings.filterwarnings(\"ignore\", \"global statement\", module=\"<string>\")\n        exec 'global a; a = 1; b = 2' in g, l\n        if g.has_key('__builtins__'): del g['__builtins__']\n        if l.has_key('__builtins__'): del l['__builtins__']\n        if (g, l) != ({'a':1}, {'b':2}):\n            self.fail('exec ... in g (%s), l (%s)' %(g,l))\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError, e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr [('as' | ',') expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError, msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError), msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'<>'|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 <> 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 <> 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n# ... truncated ...","source_hash":"42e9014edfae60326c9e26568e1ce04ca266616410b81ac5bcf09ab1424649bd","truncated":true}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.crlf-line-endings","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.crlf-line-endings#L1-L6","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.crlf-line-endings","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/crlf-line-endings.py","language":"python","start_line":1,"end_line":6,"context_start_line":1,"context_end_line":6,"code":"print a\n\nif b:    \n    if c:\n        d\n    e","source_hash":"cce5aa47eb68b83f709a86766ddfd75a7141d3d9708beeadc1a465acd8441796","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.trailing-whitespace","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.trailing-whitespace#L1-L6","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.trailing-whitespace","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/trailing-whitespace.py","language":"python","start_line":1,"end_line":6,"context_start_line":1,"context_end_line":6,"code":"print a    \n\nif b:    \n    if c:    \n        d\n    e     ","source_hash":"7d63af3a7bf0ad5f851005a540d9c0ee6bb92c07f90969361047ab0632a09eb3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.simple-statements-without-trailing-newline","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.simple-statements-without-trailing-newline#L1-L1","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.simple-statements-without-trailing-newline","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/simple-statements-without-trailing-newline.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":1,"code":"pass; print \"hi\"","source_hash":"35b1cefdabf4ddfe171073aed03d903af42713ac1aab854e94e4c3ac09ee42dc","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.tabs","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.tabs#L1-L32","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.tabs","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/tabs.py","language":"python","start_line":1,"end_line":32,"context_start_line":1,"context_end_line":32,"code":"def set_password(args):\n\tpassword = args.password\n\twhile not password  :\n\t\tpassword1 = getpass(\"\" if args.quiet else \"Provide password: \")\n\t\tpassword_repeat = getpass(\"\" if args.quiet else \"Repeat password:  \")\n\t\tif password1 != password_repeat:\n\t\t\tprint(\"Passwords do not match, try again\")\n\t\telif len(password1) < 4:\n\t\t\tprint(\"Please provide at least 4 characters\")\n\t\telse:\n\t\t\tpassword = password1\n\n\tpassword_hash = passwd(password)\n\tcfg = BaseJSONConfigManager(config_dir=jupyter_config_dir())\n\tcfg.update('jupyter_notebook_config', {\n\t\t'NotebookApp': {\n\t\t\t'password': password_hash,\n\t\t}\n\t})\n\tif not args.quiet:\n\t\tprint(\"password stored in config dir: %s\" % jupyter_config_dir())\n\ndef main(argv):\n\tparser = argparse.ArgumentParser(argv[0])\n\tsubparsers = parser.add_subparsers()\n\tparser_password = subparsers.add_parser('password', help='sets a password for your notebook server')\n\tparser_password.add_argument(\"password\", help=\"password to set, if not given, a password will be queried for (NOTE: this may not be safe)\",\n\t\t\tnargs=\"?\")\n\tparser_password.add_argument(\"--quiet\", help=\"suppress messages\", action=\"store_true\")\n\tparser_password.set_defaults(function=set_password)\n\targs = parser.parse_args(argv[1:])\n\targs.function(args)","source_hash":"6f3c435bec254bbaff69f0835685f018db23d21e6cae8cb77aaaaff030a08075","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.tabs.set_password","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.tabs.set_password#L1-L21","kind":"function","name":"set_password","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/tabs.py","language":"python","start_line":1,"end_line":21,"context_start_line":1,"context_end_line":32,"code":"def set_password(args):\n\tpassword = args.password\n\twhile not password  :\n\t\tpassword1 = getpass(\"\" if args.quiet else \"Provide password: \")\n\t\tpassword_repeat = getpass(\"\" if args.quiet else \"Repeat password:  \")\n\t\tif password1 != password_repeat:\n\t\t\tprint(\"Passwords do not match, try again\")\n\t\telif len(password1) < 4:\n\t\t\tprint(\"Please provide at least 4 characters\")\n\t\telse:\n\t\t\tpassword = password1\n\n\tpassword_hash = passwd(password)\n\tcfg = BaseJSONConfigManager(config_dir=jupyter_config_dir())\n\tcfg.update('jupyter_notebook_config', {\n\t\t'NotebookApp': {\n\t\t\t'password': password_hash,\n\t\t}\n\t})\n\tif not args.quiet:\n\t\tprint(\"password stored in config dir: %s\" % jupyter_config_dir())\n\ndef main(argv):\n\tparser = argparse.ArgumentParser(argv[0])\n\tsubparsers = parser.add_subparsers()\n\tparser_password = subparsers.add_parser('password', help='sets a password for your notebook server')\n\tparser_password.add_argument(\"password\", help=\"password to set, if not given, a password will be queried for (NOTE: this may not be safe)\",\n\t\t\tnargs=\"?\")\n\tparser_password.add_argument(\"--quiet\", help=\"suppress messages\", action=\"store_true\")\n\tparser_password.set_defaults(function=set_password)\n\targs = parser.parse_args(argv[1:])\n\targs.function(args)","source_hash":"6f3c435bec254bbaff69f0835685f018db23d21e6cae8cb77aaaaff030a08075","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.tabs.main","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.tabs.main#L23-L32","kind":"function","name":"main","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/tabs.py","language":"python","start_line":23,"end_line":32,"context_start_line":3,"context_end_line":32,"code":"\twhile not password  :\n\t\tpassword1 = getpass(\"\" if args.quiet else \"Provide password: \")\n\t\tpassword_repeat = getpass(\"\" if args.quiet else \"Repeat password:  \")\n\t\tif password1 != password_repeat:\n\t\t\tprint(\"Passwords do not match, try again\")\n\t\telif len(password1) < 4:\n\t\t\tprint(\"Please provide at least 4 characters\")\n\t\telse:\n\t\t\tpassword = password1\n\n\tpassword_hash = passwd(password)\n\tcfg = BaseJSONConfigManager(config_dir=jupyter_config_dir())\n\tcfg.update('jupyter_notebook_config', {\n\t\t'NotebookApp': {\n\t\t\t'password': password_hash,\n\t\t}\n\t})\n\tif not args.quiet:\n\t\tprint(\"password stored in config dir: %s\" % jupyter_config_dir())\n\ndef main(argv):\n\tparser = argparse.ArgumentParser(argv[0])\n\tsubparsers = parser.add_subparsers()\n\tparser_password = subparsers.add_parser('password', help='sets a password for your notebook server')\n\tparser_password.add_argument(\"password\", help=\"password to set, if not given, a password will be queried for (NOTE: this may not be safe)\",\n\t\t\tnargs=\"?\")\n\tparser_password.add_argument(\"--quiet\", help=\"suppress messages\", action=\"store_true\")\n\tparser_password.set_defaults(function=set_password)\n\targs = parser.parse_args(argv[1:])\n\targs.function(args)","source_hash":"6f3c435bec254bbaff69f0835685f018db23d21e6cae8cb77aaaaff030a08075","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.compound-statement-without-trailing-newline","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.compound-statement-without-trailing-newline#L1-L4","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.compound-statement-without-trailing-newline","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/compound-statement-without-trailing-newline.py","language":"python","start_line":1,"end_line":4,"context_start_line":1,"context_end_line":4,"code":"\nclass Foo:\n  def bar():\n    print \"hi\"","source_hash":"8296cca55337701b6da541e9a622df9928f784620326d45bc62f32675cc185e8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar#L1-L945","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":1,"end_line":945,"context_start_line":1,"context_end_line":945,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\\n        + 1\n        self.assertEquals(x, 2, 'backslash for line continuation')\n\n        # Backslash does not means continuation in comments :\\\n        x = 0\n        self.assertEquals(x, 0, 'backslash ending comment')\n\n    def testPlainIntegers(self):\n        self.assertEquals(type(000), type(0))\n        self.assertEquals(0xff, 255)\n        self.assertEquals(0o377, 255)\n        self.assertEquals(2147483647, 0o17777777777)\n        self.assertEquals(0b1001, 9)\n        # \"0x\" is not a valid literal\n        self.assertRaises(SyntaxError, eval, \"0x\")\n        from sys import maxsize\n        if maxsize == 2147483647:\n            self.assertEquals(-2147483647-1, -0o20000000000)\n            # XXX -2147483648\n            self.assert_(0o37777777777 > 0)\n            self.assert_(0xffffffff > 0)\n            self.assert_(0b1111111111111111111111111111111 > 0)\n            for s in ('2147483648', '0o40000000000', '0x100000000',\n                      '0b10000000000000000000000000000000'):\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        elif maxsize == 9223372036854775807:\n            self.assertEquals(-9223372036854775807-1, -0o1000000000000000000000)\n            self.assert_(0o1777777777777777777777 > 0)\n            self.assert_(0xffffffffffffffff > 0)\n            self.assert_(0b11111111111111111111111111111111111111111111111111111111111111 > 0)\n            for s in '9223372036854775808', '0o2000000000000000000000', \\\n                     '0x10000000000000000', \\\n                     '0b100000000000000000000000000000000000000000000000000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxsize value %r' % maxsize)\n\n    def testLongIntegers(self):\n        x = 0\n        x = 0xffffffffffffffff\n        x = 0Xffffffffffffffff\n        x = 0o77777777777777777\n        x = 0O77777777777777777\n        x = 123456789012345678901234567890\n        x = 0b100000000000000000000000000000000000000000000000000000000000000000000\n        x = 0B111111111111111111111111111111111111111111111111111111111111111111111\n\n    def testUnderscoresInNumbers(self):\n        # Integers\n        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67\n        x = 0O_45_67\n\n        # Floats\n        x = 3_1.4\n        x = 03_1.4\n        x = 3_1.\n        x = .3_1\n        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4\n        x = 3_1e+4_1\n        x = 3_1E-4_1\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})\n        def f(*x:str): pass\n        self.assertEquals(f.__annotations__, {'x': str})\n        def f(**x:float): pass\n        self.assertEquals(f.__annotations__, {'x': float})\n        def f(x, y:1+2): pass\n        self.assertEquals(f.__annotations__, {'y': 3})\n        def f(a, b:1, c:2, d): pass\n        self.assertEquals(f.__annotations__, {'b': 1, 'c': 2})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6): pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6, h:7, i=8, j:9=10,\n              **k:11) -> 12: pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6, 'h': 7, 'j': 9,\n                           'k': 11, 'return': 12})\n        # Check for SF Bug #1697248 - mixing decorators and a return annotation\n        def null(x): return x\n        @null\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n\n        # test closures with a variety of oparg's\n        closure = 1\n        def f(): return closure\n        def f(x=1): return closure\n        def f(*, k=1): return closure\n        def f() -> int: return closure\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError as msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError) as msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]\n# ... truncated ...","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":true}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.TokenTests","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.TokenTests#L17-L114","kind":"class","name":"TokenTests","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":17,"end_line":114,"context_start_line":1,"context_end_line":134,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\\n        + 1\n        self.assertEquals(x, 2, 'backslash for line continuation')\n\n        # Backslash does not means continuation in comments :\\\n        x = 0\n        self.assertEquals(x, 0, 'backslash ending comment')\n\n    def testPlainIntegers(self):\n        self.assertEquals(type(000), type(0))\n        self.assertEquals(0xff, 255)\n        self.assertEquals(0o377, 255)\n        self.assertEquals(2147483647, 0o17777777777)\n        self.assertEquals(0b1001, 9)\n        # \"0x\" is not a valid literal\n        self.assertRaises(SyntaxError, eval, \"0x\")\n        from sys import maxsize\n        if maxsize == 2147483647:\n            self.assertEquals(-2147483647-1, -0o20000000000)\n            # XXX -2147483648\n            self.assert_(0o37777777777 > 0)\n            self.assert_(0xffffffff > 0)\n            self.assert_(0b1111111111111111111111111111111 > 0)\n            for s in ('2147483648', '0o40000000000', '0x100000000',\n                      '0b10000000000000000000000000000000'):\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        elif maxsize == 9223372036854775807:\n            self.assertEquals(-9223372036854775807-1, -0o1000000000000000000000)\n            self.assert_(0o1777777777777777777777 > 0)\n            self.assert_(0xffffffffffffffff > 0)\n            self.assert_(0b11111111111111111111111111111111111111111111111111111111111111 > 0)\n            for s in '9223372036854775808', '0o2000000000000000000000', \\\n                     '0x10000000000000000', \\\n                     '0b100000000000000000000000000000000000000000000000000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxsize value %r' % maxsize)\n\n    def testLongIntegers(self):\n        x = 0\n        x = 0xffffffffffffffff\n        x = 0Xffffffffffffffff\n        x = 0o77777777777777777\n        x = 0O77777777777777777\n        x = 123456789012345678901234567890\n        x = 0b100000000000000000000000000000000000000000000000000000000000000000000\n        x = 0B111111111111111111111111111111111111111111111111111111111111111111111\n\n    def testUnderscoresInNumbers(self):\n        # Integers\n        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67\n        x = 0O_45_67\n\n        # Floats\n        x = 3_1.4\n        x = 03_1.4\n        x = 3_1.\n        x = .3_1\n        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4\n        x = 3_1e+4_1\n        x = 3_1E-4_1\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.GrammarTests","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.GrammarTests#L116-L938","kind":"class","name":"GrammarTests","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":116,"end_line":938,"context_start_line":96,"context_end_line":945,"code":"\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})\n        def f(*x:str): pass\n        self.assertEquals(f.__annotations__, {'x': str})\n        def f(**x:float): pass\n        self.assertEquals(f.__annotations__, {'x': float})\n        def f(x, y:1+2): pass\n        self.assertEquals(f.__annotations__, {'y': 3})\n        def f(a, b:1, c:2, d): pass\n        self.assertEquals(f.__annotations__, {'b': 1, 'c': 2})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6): pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6, h:7, i=8, j:9=10,\n              **k:11) -> 12: pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6, 'h': 7, 'j': 9,\n                           'k': 11, 'return': 12})\n        # Check for SF Bug #1697248 - mixing decorators and a return annotation\n        def null(x): return x\n        @null\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n\n        # test closures with a variety of oparg's\n        closure = 1\n        def f(): return closure\n        def f(x=1): return closure\n        def f(*, k=1): return closure\n        def f() -> int: return closure\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError as msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError) as msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]\n        x = time.time()\n        x = sys.modules['time'].time()\n        a = '01234'\n        c = a[0]\n        c = a[-1]\n        s = a[0:5]\n        s = a[:5]\n        s = a[0:]\n        s = a[:]\n        s = a[-5:]\n        s = a[:-1]\n        s = a[-4:-3]\n        # A rough test of SF bug 1333982.  http://python.org/sf/1333982\n        # The testing here is fairly incomplete.\n        # Test cases should include: commas with 1 and 2 colons\n        d = {}\n        d[1] = 1\n        d[1,] = 2\n        d[1,2] = 3\n        d[1,2,3] = 4\n        L = list(d)\n        L.sort(key=lambda x: x if isinstance(x, tuple) else ())\n        self.assertEquals(str(L), '[1, (1,), (1, 2), (1, 2, 3)]')\n\n    def testAtoms(self):\n        ### atom: '(' [testlist] ')' | '[' [testlist] ']' | '{' [dictsetmaker] '}' | NAME | NUMBER | STRING\n        ### dictsetmaker: (test ':' test (',' test ':' test)* [',']) | (test (',' test)* [','])\n\n        x = (1)\n        x = (1 or 2 or 3)\n        x = (1 or 2 or 3, 2, 3)\n\n        x = []\n        x = [1]\n        x = [1 or 2 or 3]\n        x = [1 or 2 or 3, 2, 3]\n        x = []\n\n        x = {}\n        x = {'one': 1}\n        x = {'one': 1,}\n        x = {'one' or 'two': 1 or 2}\n        x = {'one': 1, 'two': 2}\n        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Apple'), (3, 'Banana'), (3, 'Coconut'),\n                          (4, 'Apple'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Apple'), (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(i, s) for i in nums for s in [f for f in s\n# ... truncated ...","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":true}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_main","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_main#L941-L942","kind":"function","name":"test_main","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":941,"end_line":942,"context_start_line":921,"context_end_line":945,"code":"jumps over\nthe 'lazy' dog.\n'''\n        self.assertEquals(x, y)\n        y = \"\\n\\\nThe \\\"quick\\\"\\n\\\nbrown fox\\n\\\njumps over\\n\\\nthe 'lazy' dog.\\n\\\n\"\n        self.assertEquals(x, y)\n        y = '\\n\\\nThe \\\"quick\\\"\\n\\\nbrown fox\\n\\\njumps over\\n\\\nthe \\'lazy\\' dog.\\n\\\n'\n        self.assertEquals(x, y)\n\n\ndef test_main():\n    run_unittest(TokenTests, GrammarTests)\n\nif __name__ == '__main__':\n    test_main()","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testBackslash","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testBackslash#L19-L27","kind":"function","name":"testBackslash","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":19,"end_line":27,"context_start_line":1,"context_end_line":47,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\\n        + 1\n        self.assertEquals(x, 2, 'backslash for line continuation')\n\n        # Backslash does not means continuation in comments :\\\n        x = 0\n        self.assertEquals(x, 0, 'backslash ending comment')\n\n    def testPlainIntegers(self):\n        self.assertEquals(type(000), type(0))\n        self.assertEquals(0xff, 255)\n        self.assertEquals(0o377, 255)\n        self.assertEquals(2147483647, 0o17777777777)\n        self.assertEquals(0b1001, 9)\n        # \"0x\" is not a valid literal\n        self.assertRaises(SyntaxError, eval, \"0x\")\n        from sys import maxsize\n        if maxsize == 2147483647:\n            self.assertEquals(-2147483647-1, -0o20000000000)\n            # XXX -2147483648\n            self.assert_(0o37777777777 > 0)\n            self.assert_(0xffffffff > 0)\n            self.assert_(0b1111111111111111111111111111111 > 0)\n            for s in ('2147483648', '0o40000000000', '0x100000000',\n                      '0b10000000000000000000000000000000'):\n                try:\n                    x = eval(s)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testPlainIntegers","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testPlainIntegers#L29-L63","kind":"function","name":"testPlainIntegers","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":29,"end_line":63,"context_start_line":9,"context_end_line":83,"code":"# regrtest.py.\n\nfrom test.support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\\n        + 1\n        self.assertEquals(x, 2, 'backslash for line continuation')\n\n        # Backslash does not means continuation in comments :\\\n        x = 0\n        self.assertEquals(x, 0, 'backslash ending comment')\n\n    def testPlainIntegers(self):\n        self.assertEquals(type(000), type(0))\n        self.assertEquals(0xff, 255)\n        self.assertEquals(0o377, 255)\n        self.assertEquals(2147483647, 0o17777777777)\n        self.assertEquals(0b1001, 9)\n        # \"0x\" is not a valid literal\n        self.assertRaises(SyntaxError, eval, \"0x\")\n        from sys import maxsize\n        if maxsize == 2147483647:\n            self.assertEquals(-2147483647-1, -0o20000000000)\n            # XXX -2147483648\n            self.assert_(0o37777777777 > 0)\n            self.assert_(0xffffffff > 0)\n            self.assert_(0b1111111111111111111111111111111 > 0)\n            for s in ('2147483648', '0o40000000000', '0x100000000',\n                      '0b10000000000000000000000000000000'):\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        elif maxsize == 9223372036854775807:\n            self.assertEquals(-9223372036854775807-1, -0o1000000000000000000000)\n            self.assert_(0o1777777777777777777777 > 0)\n            self.assert_(0xffffffffffffffff > 0)\n            self.assert_(0b11111111111111111111111111111111111111111111111111111111111111 > 0)\n            for s in '9223372036854775808', '0o2000000000000000000000', \\\n                     '0x10000000000000000', \\\n                     '0b100000000000000000000000000000000000000000000000000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxsize value %r' % maxsize)\n\n    def testLongIntegers(self):\n        x = 0\n        x = 0xffffffffffffffff\n        x = 0Xffffffffffffffff\n        x = 0o77777777777777777\n        x = 0O77777777777777777\n        x = 123456789012345678901234567890\n        x = 0b100000000000000000000000000000000000000000000000000000000000000000000\n        x = 0B111111111111111111111111111111111111111111111111111111111111111111111\n\n    def testUnderscoresInNumbers(self):\n        # Integers\n        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testLongIntegers","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testLongIntegers#L65-L73","kind":"function","name":"testLongIntegers","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":65,"end_line":73,"context_start_line":45,"context_end_line":93,"code":"                      '0b10000000000000000000000000000000'):\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        elif maxsize == 9223372036854775807:\n            self.assertEquals(-9223372036854775807-1, -0o1000000000000000000000)\n            self.assert_(0o1777777777777777777777 > 0)\n            self.assert_(0xffffffffffffffff > 0)\n            self.assert_(0b11111111111111111111111111111111111111111111111111111111111111 > 0)\n            for s in '9223372036854775808', '0o2000000000000000000000', \\\n                     '0x10000000000000000', \\\n                     '0b100000000000000000000000000000000000000000000000000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxsize value %r' % maxsize)\n\n    def testLongIntegers(self):\n        x = 0\n        x = 0xffffffffffffffff\n        x = 0Xffffffffffffffff\n        x = 0o77777777777777777\n        x = 0O77777777777777777\n        x = 123456789012345678901234567890\n        x = 0b100000000000000000000000000000000000000000000000000000000000000000000\n        x = 0B111111111111111111111111111111111111111111111111111111111111111111111\n\n    def testUnderscoresInNumbers(self):\n        # Integers\n        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67\n        x = 0O_45_67\n\n        # Floats\n        x = 3_1.4\n        x = 03_1.4\n        x = 3_1.\n        x = .3_1\n        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testUnderscoresInNumbers","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testUnderscoresInNumbers#L75-L95","kind":"function","name":"testUnderscoresInNumbers","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":75,"end_line":95,"context_start_line":55,"context_end_line":115,"code":"            for s in '9223372036854775808', '0o2000000000000000000000', \\\n                     '0x10000000000000000', \\\n                     '0b100000000000000000000000000000000000000000000000000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxsize value %r' % maxsize)\n\n    def testLongIntegers(self):\n        x = 0\n        x = 0xffffffffffffffff\n        x = 0Xffffffffffffffff\n        x = 0o77777777777777777\n        x = 0O77777777777777777\n        x = 123456789012345678901234567890\n        x = 0b100000000000000000000000000000000000000000000000000000000000000000000\n        x = 0B111111111111111111111111111111111111111111111111111111111111111111111\n\n    def testUnderscoresInNumbers(self):\n        # Integers\n        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67\n        x = 0O_45_67\n\n        # Floats\n        x = 3_1.4\n        x = 03_1.4\n        x = 3_1.\n        x = .3_1\n        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4\n        x = 3_1e+4_1\n        x = 3_1E-4_1\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testFloats","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testFloats#L97-L109","kind":"function","name":"testFloats","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":97,"end_line":109,"context_start_line":77,"context_end_line":129,"code":"        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67\n        x = 0O_45_67\n\n        # Floats\n        x = 3_1.4\n        x = 03_1.4\n        x = 3_1.\n        x = .3_1\n        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4\n        x = 3_1e+4_1\n        x = 3_1E-4_1\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testEllipsis","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testEllipsis#L111-L114","kind":"function","name":"testEllipsis","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":111,"end_line":114,"context_start_line":91,"context_end_line":134,"code":"        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4\n        x = 3_1e+4_1\n        x = 3_1E-4_1\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testEvalInput","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testEvalInput#L127-L129","kind":"function","name":"testEvalInput","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":127,"end_line":129,"context_start_line":107,"context_end_line":149,"code":"        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testFuncdef","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testFuncdef#L131-L313","kind":"function","name":"testFuncdef","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":131,"end_line":313,"context_start_line":111,"context_end_line":333,"code":"    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})\n        def f(*x:str): pass\n        self.assertEquals(f.__annotations__, {'x': str})\n        def f(**x:float): pass\n        self.assertEquals(f.__annotations__, {'x': float})\n        def f(x, y:1+2): pass\n        self.assertEquals(f.__annotations__, {'y': 3})\n        def f(a, b:1, c:2, d): pass\n        self.assertEquals(f.__annotations__, {'b': 1, 'c': 2})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6): pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6, h:7, i=8, j:9=10,\n              **k:11) -> 12: pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6, 'h': 7, 'j': 9,\n                           'k': 11, 'return': 12})\n        # Check for SF Bug #1697248 - mixing decorators and a return annotation\n        def null(x): return x\n        @null\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n\n        # test closures with a variety of oparg's\n        closure = 1\n        def f(): return closure\n        def f(x=1): return closure\n        def f(*, k=1): return closure\n        def f() -> int: return closure\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testLambdef","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testLambdef#L315-L331","kind":"function","name":"testLambdef","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":315,"end_line":331,"context_start_line":295,"context_end_line":351,"code":"        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6, 'h': 7, 'j': 9,\n                           'k': 11, 'return': 12})\n        # Check for SF Bug #1697248 - mixing decorators and a return annotation\n        def null(x): return x\n        @null\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n\n        # test closures with a variety of oparg's\n        closure = 1\n        def f(): return closure\n        def f(x=1): return closure\n        def f(*, k=1): return closure\n        def f() -> int: return closure\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testSimpleStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testSimpleStmt#L337-L343","kind":"function","name":"testSimpleStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":337,"end_line":343,"context_start_line":317,"context_end_line":363,"code":"        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testExprStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testExprStmt#L348-L359","kind":"function","name":"testExprStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":348,"end_line":359,"context_start_line":328,"context_end_line":379,"code":"        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testDelStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testDelStmt#L361-L368","kind":"function","name":"testDelStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":361,"end_line":368,"context_start_line":341,"context_end_line":388,"code":"            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testPassStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testPassStmt#L370-L372","kind":"function","name":"testPassStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":370,"end_line":372,"context_start_line":350,"context_end_line":392,"code":"        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testBreakStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testBreakStmt#L377-L379","kind":"function","name":"testBreakStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":377,"end_line":379,"context_start_line":357,"context_end_line":399,"code":"\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testContinueStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testContinueStmt#L381-L405","kind":"function","name":"testContinueStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":381,"end_line":405,"context_start_line":361,"context_end_line":425,"code":"    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_break_continue_loop","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_break_continue_loop#L407-L431","kind":"function","name":"test_break_continue_loop","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":407,"end_line":431,"context_start_line":387,"context_end_line":451,"code":"        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testReturn","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testReturn#L433-L439","kind":"function","name":"testReturn","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":433,"end_line":439,"context_start_line":413,"context_end_line":459,"code":"        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testYield","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testYield#L441-L442","kind":"function","name":"testYield","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":441,"end_line":442,"context_start_line":421,"context_end_line":462,"code":"                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testRaise","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testRaise#L444-L449","kind":"function","name":"testRaise","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":444,"end_line":449,"context_start_line":424,"context_end_line":469,"code":"                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testImport","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testImport#L451-L462","kind":"function","name":"testImport","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":451,"end_line":462,"context_start_line":431,"context_end_line":482,"code":"        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testGlobal","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testGlobal#L464-L468","kind":"function","name":"testGlobal","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":464,"end_line":468,"context_start_line":444,"context_end_line":488,"code":"    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testNonlocal","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testNonlocal#L470-L476","kind":"function","name":"testNonlocal","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":470,"end_line":476,"context_start_line":450,"context_end_line":496,"code":"\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testAssert","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testAssert#L478-L490","kind":"function","name":"testAssert","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":478,"end_line":490,"context_start_line":458,"context_end_line":510,"code":"        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testIf","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testIf#L495-L506","kind":"function","name":"testIf","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":495,"end_line":506,"context_start_line":475,"context_end_line":526,"code":"            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testWhile","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testWhile#L508-L521","kind":"function","name":"testWhile","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":508,"end_line":521,"context_start_line":488,"context_end_line":541,"code":"        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testFor","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testFor#L523-L548","kind":"function","name":"testFor","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":523,"end_line":548,"context_start_line":503,"context_end_line":568,"code":"        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError as msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testTry","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testTry#L550-L571","kind":"function","name":"testTry","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":550,"end_line":571,"context_start_line":530,"context_end_line":591,"code":"                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError as msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError) as msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testSuite","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testSuite#L573-L585","kind":"function","name":"testSuite","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":573,"end_line":585,"context_start_line":553,"context_end_line":605,"code":"        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError as msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError) as msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testTest","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testTest#L588-L597","kind":"function","name":"testTest","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":588,"end_line":597,"context_start_line":568,"context_end_line":617,"code":"        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError) as msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testComparison","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testComparison#L599-L614","kind":"function","name":"testComparison","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":599,"end_line":614,"context_start_line":579,"context_end_line":634,"code":"            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testBinaryMaskOps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testBinaryMaskOps#L616-L619","kind":"function","name":"testBinaryMaskOps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":616,"end_line":619,"context_start_line":596,"context_end_line":639,"code":"        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testShiftOps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testShiftOps#L621-L624","kind":"function","name":"testShiftOps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":621,"end_line":624,"context_start_line":601,"context_end_line":644,"code":"        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testAdditiveOps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testAdditiveOps#L626-L630","kind":"function","name":"testAdditiveOps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":626,"end_line":630,"context_start_line":606,"context_end_line":650,"code":"        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testMultiplicativeOps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testMultiplicativeOps#L632-L636","kind":"function","name":"testMultiplicativeOps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":632,"end_line":636,"context_start_line":612,"context_end_line":656,"code":"        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]\n        x = time.time()\n        x = sys.modules['time'].time()\n        a = '01234'\n        c = a[0]\n        c = a[-1]\n        s = a[0:5]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testUnaryOps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testUnaryOps#L638-L643","kind":"function","name":"testUnaryOps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":638,"end_line":643,"context_start_line":618,"context_end_line":663,"code":"        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]\n        x = time.time()\n        x = sys.modules['time'].time()\n        a = '01234'\n        c = a[0]\n        c = a[-1]\n        s = a[0:5]\n        s = a[:5]\n        s = a[0:]\n        s = a[:]\n        s = a[-5:]\n        s = a[:-1]\n        s = a[-4:-3]\n        # A rough test of SF bug 1333982.  http://python.org/sf/1333982","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testSelectors","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testSelectors#L645-L673","kind":"function","name":"testSelectors","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":645,"end_line":673,"context_start_line":625,"context_end_line":693,"code":"\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]\n        x = time.time()\n        x = sys.modules['time'].time()\n        a = '01234'\n        c = a[0]\n        c = a[-1]\n        s = a[0:5]\n        s = a[:5]\n        s = a[0:]\n        s = a[:]\n        s = a[-5:]\n        s = a[:-1]\n        s = a[-4:-3]\n        # A rough test of SF bug 1333982.  http://python.org/sf/1333982\n        # The testing here is fairly incomplete.\n        # Test cases should include: commas with 1 and 2 colons\n        d = {}\n        d[1] = 1\n        d[1,] = 2\n        d[1,2] = 3\n        d[1,2,3] = 4\n        L = list(d)\n        L.sort(key=lambda x: x if isinstance(x, tuple) else ())\n        self.assertEquals(str(L), '[1, (1,), (1, 2), (1, 2, 3)]')\n\n    def testAtoms(self):\n        ### atom: '(' [testlist] ')' | '[' [testlist] ']' | '{' [dictsetmaker] '}' | NAME | NUMBER | STRING\n        ### dictsetmaker: (test ':' test (',' test ':' test)* [',']) | (test (',' test)* [','])\n\n        x = (1)\n        x = (1 or 2 or 3)\n        x = (1 or 2 or 3, 2, 3)\n\n        x = []\n        x = [1]\n        x = [1 or 2 or 3]\n        x = [1 or 2 or 3, 2, 3]\n        x = []\n\n        x = {}\n        x = {'one': 1}\n        x = {'one': 1,}\n        x = {'one' or 'two': 1 or 2}\n        x = {'one': 1, 'two': 2}","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testAtoms","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testAtoms#L675-L704","kind":"function","name":"testAtoms","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":675,"end_line":704,"context_start_line":655,"context_end_line":724,"code":"        c = a[-1]\n        s = a[0:5]\n        s = a[:5]\n        s = a[0:]\n        s = a[:]\n        s = a[-5:]\n        s = a[:-1]\n        s = a[-4:-3]\n        # A rough test of SF bug 1333982.  http://python.org/sf/1333982\n        # The testing here is fairly incomplete.\n        # Test cases should include: commas with 1 and 2 colons\n        d = {}\n        d[1] = 1\n        d[1,] = 2\n        d[1,2] = 3\n        d[1,2,3] = 4\n        L = list(d)\n        L.sort(key=lambda x: x if isinstance(x, tuple) else ())\n        self.assertEquals(str(L), '[1, (1,), (1, 2), (1, 2, 3)]')\n\n    def testAtoms(self):\n        ### atom: '(' [testlist] ')' | '[' [testlist] ']' | '{' [dictsetmaker] '}' | NAME | NUMBER | STRING\n        ### dictsetmaker: (test ':' test (',' test ':' test)* [',']) | (test (',' test)* [','])\n\n        x = (1)\n        x = (1 or 2 or 3)\n        x = (1 or 2 or 3, 2, 3)\n\n        x = []\n        x = [1]\n        x = [1 or 2 or 3]\n        x = [1 or 2 or 3, 2, 3]\n        x = []\n\n        x = {}\n        x = {'one': 1}\n        x = {'one': 1,}\n        x = {'one' or 'two': 1 or 2}\n        x = {'one': 1, 'two': 2}\n        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testClassdef","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testClassdef#L710-L727","kind":"function","name":"testClassdef","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":710,"end_line":727,"context_start_line":690,"context_end_line":747,"code":"        x = {'one': 1}\n        x = {'one': 1,}\n        x = {'one' or 'two': 1 or 2}\n        x = {'one': 1, 'two': 2}\n        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testDictcomps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testDictcomps#L729-L734","kind":"function","name":"testDictcomps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":729,"end_line":734,"context_start_line":709,"context_end_line":754,"code":"\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Apple'), (3, 'Banana'), (3, 'Coconut'),\n                          (4, 'Apple'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Apple'), (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(i, s) for i in nums for s in [f for f in strs if \"n\" in f]],\n                         [(1, 'Banana'), (1, 'Coconut'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Banana'), (3, 'Coconut'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Banana'), (5, 'Coconut')])","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testListcomps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testListcomps#L736-L797","kind":"function","name":"testListcomps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":736,"end_line":797,"context_start_line":716,"context_end_line":817,"code":"        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Apple'), (3, 'Banana'), (3, 'Coconut'),\n                          (4, 'Apple'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Apple'), (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(i, s) for i in nums for s in [f for f in strs if \"n\" in f]],\n                         [(1, 'Banana'), (1, 'Coconut'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Banana'), (3, 'Coconut'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(lambda a:[a**i for i in range(a+1)])(j) for j in range(5)],\n                         [[1], [1, 1], [1, 2, 4], [1, 3, 9, 27], [1, 4, 16, 64, 256]])\n\n        def test_in_func(l):\n            return [0 < x < 3 for x in l if x > 2]\n\n        self.assertEqual(test_in_func(nums), [False, False, False])\n\n        def test_nested_front():\n            self.assertEqual([[y for y in [x, x + 1]] for x in [1,3,5]],\n                             [[1, 2], [3, 4], [5, 6]])\n\n        test_nested_front()\n\n        check_syntax_error(self, \"[i, s for i in nums for s in strs]\")\n        check_syntax_error(self, \"[x if y]\")\n\n        suppliers = [\n          (1, \"Boeing\"),\n          (2, \"Ford\"),\n          (3, \"Macdonalds\")\n        ]\n\n        parts = [\n          (10, \"Airliner\"),\n          (20, \"Engine\"),\n          (30, \"Cheeseburger\")\n        ]\n\n        suppart = [\n          (1, 10), (1, 20), (2, 20), (3, 30)\n        ]\n\n        x = [\n          (sname, pname)\n            for (sno, sname) in suppliers\n              for (pno, pname) in parts\n                for (sp_sno, sp_pno) in suppart\n                  if sno == sp_sno and pno == sp_pno\n        ]\n\n        self.assertEqual(x, [('Boeing', 'Airliner'), ('Boeing', 'Engine'), ('Ford', 'Engine'),\n                             ('Macdonalds', 'Cheeseburger')])\n\n    def testGenexps(self):\n        # generator expression tests\n        g = ([x for x in range(10)] for x in range(1))\n        self.assertEqual(next(g), [x for x in range(10)])\n        try:\n            next(g)\n            self.fail('should produce StopIteration exception')\n        except StopIteration:\n            pass\n\n        a = 1\n        try:\n            g = (a for d in a)\n            next(g)\n            self.fail('should produce TypeError')\n        except TypeError:\n            pass\n\n        self.assertEqual(list((x, y) for x in 'abcd' for y in 'abcd'), [(x, y) for x in 'abcd' for y in 'abcd'])","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testGenexps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testGenexps#L799-L832","kind":"function","name":"testGenexps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":799,"end_line":832,"context_start_line":779,"context_end_line":852,"code":"          (10, \"Airliner\"),\n          (20, \"Engine\"),\n          (30, \"Cheeseburger\")\n        ]\n\n        suppart = [\n          (1, 10), (1, 20), (2, 20), (3, 30)\n        ]\n\n        x = [\n          (sname, pname)\n            for (sno, sname) in suppliers\n              for (pno, pname) in parts\n                for (sp_sno, sp_pno) in suppart\n                  if sno == sp_sno and pno == sp_pno\n        ]\n\n        self.assertEqual(x, [('Boeing', 'Airliner'), ('Boeing', 'Engine'), ('Ford', 'Engine'),\n                             ('Macdonalds', 'Cheeseburger')])\n\n    def testGenexps(self):\n        # generator expression tests\n        g = ([x for x in range(10)] for x in range(1))\n        self.assertEqual(next(g), [x for x in range(10)])\n        try:\n            next(g)\n            self.fail('should produce StopIteration exception')\n        except StopIteration:\n            pass\n\n        a = 1\n        try:\n            g = (a for d in a)\n            next(g)\n            self.fail('should produce TypeError')\n        except TypeError:\n            pass\n\n        self.assertEqual(list((x, y) for x in 'abcd' for y in 'abcd'), [(x, y) for x in 'abcd' for y in 'abcd'])\n        self.assertEqual(list((x, y) for x in 'ab' for y in 'xy'), [(x, y) for x in 'ab' for y in 'xy'])\n\n        a = [x for x in range(10)]\n        b = (x for x in (y for y in a))\n        self.assertEqual(sum(b), sum([x for x in range(10)]))\n\n        self.assertEqual(sum(x**2 for x in range(10)), sum([x**2 for x in range(10)]))\n        self.assertEqual(sum(x*x for x in range(10) if x%2), sum([x*x for x in range(10) if x%2]))\n        self.assertEqual(sum(x for x in (y for y in range(10))), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10)))), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in [y for y in (z for z in range(10))]), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10) if True)) if True), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10) if True) if False) if True), 0)\n        check_syntax_error(self, \"foo(x for x in range(10), 100)\")\n        check_syntax_error(self, \"foo(100, x for x in range(10))\")\n\n    def testComprehensionSpecials(self):\n        # test for outmost iterable precomputation\n        x = 10; g = (i for i in range(x)); x = 5\n        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testComprehensionSpecials","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testComprehensionSpecials#L834-L851","kind":"function","name":"testComprehensionSpecials","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":834,"end_line":851,"context_start_line":814,"context_end_line":871,"code":"        except TypeError:\n            pass\n\n        self.assertEqual(list((x, y) for x in 'abcd' for y in 'abcd'), [(x, y) for x in 'abcd' for y in 'abcd'])\n        self.assertEqual(list((x, y) for x in 'ab' for y in 'xy'), [(x, y) for x in 'ab' for y in 'xy'])\n\n        a = [x for x in range(10)]\n        b = (x for x in (y for y in a))\n        self.assertEqual(sum(b), sum([x for x in range(10)]))\n\n        self.assertEqual(sum(x**2 for x in range(10)), sum([x**2 for x in range(10)]))\n        self.assertEqual(sum(x*x for x in range(10) if x%2), sum([x*x for x in range(10) if x%2]))\n        self.assertEqual(sum(x for x in (y for y in range(10))), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10)))), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in [y for y in (z for z in range(10))]), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10) if True)) if True), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10) if True) if False) if True), 0)\n        check_syntax_error(self, \"foo(x for x in range(10), 100)\")\n        check_syntax_error(self, \"foo(100, x for x in range(10))\")\n\n    def testComprehensionSpecials(self):\n        # test for outmost iterable precomputation\n        x = 10; g = (i for i in range(x)); x = 5\n        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n\n    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_with_statement","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_with_statement#L853-L871","kind":"function","name":"test_with_statement","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":853,"end_line":871,"context_start_line":833,"context_end_line":891,"code":"\n    def testComprehensionSpecials(self):\n        # test for outmost iterable precomputation\n        x = 10; g = (i for i in range(x)); x = 5\n        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n\n    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"\n            print(x)\n            return ret\n\n        # the next line is not allowed anymore\n        #self.assertEqual([ x() for x in lambda: True, lambda: False if x() ], [True])\n        self.assertEqual([ x() for x in (lambda: True, lambda: False) if x() ], [True])\n        self.assertEqual([ x(False) for x in (lambda x: False if x else True, lambda x: True if x else False) if x(False) ], [True])\n        self.assertEqual((5 if 1 else _checkeval(\"check 1\", 0)), 5)\n        self.assertEqual((_checkeval(\"check 2\", 0) if 0 else 5), 5)\n        self.assertEqual((5 and 6 if 0 else 1), 1)\n        self.assertEqual(((5 and 6) if 0 else 1), 1)\n        self.assertEqual((5 and (6 if 1 else 1)), 6)\n        self.assertEqual((0 or _checkeval(\"check 3\", 2) if 0 else 3), 3)\n        self.assertEqual((1 or _checkeval(\"check 4\", 2) if 1 else _checkeval(\"check 5\", 3)), 1)\n        self.assertEqual((0 or 5 if 1 else _checkeval(\"check 6\", 3)), 5)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testIfElseExpr","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testIfElseExpr#L873-L898","kind":"function","name":"testIfElseExpr","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":873,"end_line":898,"context_start_line":853,"context_end_line":918,"code":"    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"\n            print(x)\n            return ret\n\n        # the next line is not allowed anymore\n        #self.assertEqual([ x() for x in lambda: True, lambda: False if x() ], [True])\n        self.assertEqual([ x() for x in (lambda: True, lambda: False) if x() ], [True])\n        self.assertEqual([ x(False) for x in (lambda x: False if x else True, lambda x: True if x else False) if x(False) ], [True])\n        self.assertEqual((5 if 1 else _checkeval(\"check 1\", 0)), 5)\n        self.assertEqual((_checkeval(\"check 2\", 0) if 0 else 5), 5)\n        self.assertEqual((5 and 6 if 0 else 1), 1)\n        self.assertEqual(((5 and 6) if 0 else 1), 1)\n        self.assertEqual((5 and (6 if 1 else 1)), 6)\n        self.assertEqual((0 or _checkeval(\"check 3\", 2) if 0 else 3), 3)\n        self.assertEqual((1 or _checkeval(\"check 4\", 2) if 1 else _checkeval(\"check 5\", 3)), 1)\n        self.assertEqual((0 or 5 if 1 else _checkeval(\"check 6\", 3)), 5)\n        self.assertEqual((not 5 if 1 else 1), False)\n        self.assertEqual((not 5 if 0 else 1), 1)\n        self.assertEqual((6 + 1 if 1 else 2), 7)\n        self.assertEqual((6 - 1 if 1 else 2), 5)\n        self.assertEqual((6 * 2 if 1 else 4), 12)\n        self.assertEqual((6 / 2 if 1 else 3), 3)\n        self.assertEqual((6 < 4 if 0 else 2), 2)\n\n    def testStringLiterals(self):\n        x = ''; y = \"\"; self.assert_(len(x) == 0 and x == y)\n        x = '\\''; y = \"'\"; self.assert_(len(x) == 1 and x == y and ord(x) == 39)\n        x = '\"'; y = \"\\\"\"; self.assert_(len(x) == 1 and x == y and ord(x) == 34)\n        x = \"doesn't \\\"shrink\\\" does it\"\n        y = 'doesn\\'t \"shrink\" does it'\n        self.assert_(len(x) == 24 and x == y)\n        x = \"does \\\"shrink\\\" doesn't it\"\n        y = 'does \"shrink\" doesn\\'t it'\n        self.assert_(len(x) == 24 and x == y)\n        x = f\"\"\"\nThe \"quick\"\nbrown fo{ok()}x\njumps over\nthe 'lazy' dog.\n\"\"\"\n        y = '\\nThe \"quick\"\\nbrown fox\\njumps over\\nthe \\'lazy\\' dog.\\n'\n        self.assertEquals(x, y)\n        y = '''","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testStringLiterals","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.testStringLiterals#L900-L938","kind":"function","name":"testStringLiterals","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":900,"end_line":938,"context_start_line":880,"context_end_line":945,"code":"        # the next line is not allowed anymore\n        #self.assertEqual([ x() for x in lambda: True, lambda: False if x() ], [True])\n        self.assertEqual([ x() for x in (lambda: True, lambda: False) if x() ], [True])\n        self.assertEqual([ x(False) for x in (lambda x: False if x else True, lambda x: True if x else False) if x(False) ], [True])\n        self.assertEqual((5 if 1 else _checkeval(\"check 1\", 0)), 5)\n        self.assertEqual((_checkeval(\"check 2\", 0) if 0 else 5), 5)\n        self.assertEqual((5 and 6 if 0 else 1), 1)\n        self.assertEqual(((5 and 6) if 0 else 1), 1)\n        self.assertEqual((5 and (6 if 1 else 1)), 6)\n        self.assertEqual((0 or _checkeval(\"check 3\", 2) if 0 else 3), 3)\n        self.assertEqual((1 or _checkeval(\"check 4\", 2) if 1 else _checkeval(\"check 5\", 3)), 1)\n        self.assertEqual((0 or 5 if 1 else _checkeval(\"check 6\", 3)), 5)\n        self.assertEqual((not 5 if 1 else 1), False)\n        self.assertEqual((not 5 if 0 else 1), 1)\n        self.assertEqual((6 + 1 if 1 else 2), 7)\n        self.assertEqual((6 - 1 if 1 else 2), 5)\n        self.assertEqual((6 * 2 if 1 else 4), 12)\n        self.assertEqual((6 / 2 if 1 else 3), 3)\n        self.assertEqual((6 < 4 if 0 else 2), 2)\n\n    def testStringLiterals(self):\n        x = ''; y = \"\"; self.assert_(len(x) == 0 and x == y)\n        x = '\\''; y = \"'\"; self.assert_(len(x) == 1 and x == y and ord(x) == 39)\n        x = '\"'; y = \"\\\"\"; self.assert_(len(x) == 1 and x == y and ord(x) == 34)\n        x = \"doesn't \\\"shrink\\\" does it\"\n        y = 'doesn\\'t \"shrink\" does it'\n        self.assert_(len(x) == 24 and x == y)\n        x = \"does \\\"shrink\\\" doesn't it\"\n        y = 'does \"shrink\" doesn\\'t it'\n        self.assert_(len(x) == 24 and x == y)\n        x = f\"\"\"\nThe \"quick\"\nbrown fo{ok()}x\njumps over\nthe 'lazy' dog.\n\"\"\"\n        y = '\\nThe \"quick\"\\nbrown fox\\njumps over\\nthe \\'lazy\\' dog.\\n'\n        self.assertEquals(x, y)\n        y = '''\nThe \"quick\"\nbrown fox\njumps over\nthe 'lazy' dog.\n'''\n        self.assertEquals(x, y)\n        y = \"\\n\\\nThe \\\"quick\\\"\\n\\\nbrown fox\\n\\\njumps over\\n\\\nthe 'lazy' dog.\\n\\\n\"\n        self.assertEquals(x, y)\n        y = '\\n\\\nThe \\\"quick\\\"\\n\\\nbrown fox\\n\\\njumps over\\n\\\nthe \\'lazy\\' dog.\\n\\\n'\n        self.assertEquals(x, y)\n\n\ndef test_main():\n    run_unittest(TokenTests, GrammarTests)\n\nif __name__ == '__main__':\n    test_main()","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.f1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.f1#L144-L144","kind":"function","name":"f1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":144,"end_line":144,"context_start_line":124,"context_end_line":164,"code":"    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.f2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.f2#L148-L148","kind":"function","name":"f2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":148,"end_line":148,"context_start_line":128,"context_end_line":168,"code":"        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.f3","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.f3#L149-L149","kind":"function","name":"f3","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":149,"end_line":149,"context_start_line":129,"context_end_line":169,"code":"        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.a1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.a1#L152-L152","kind":"function","name":"a1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":152,"end_line":152,"context_start_line":132,"context_end_line":172,"code":"        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.a2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.a2#L153-L153","kind":"function","name":"a2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":153,"end_line":153,"context_start_line":133,"context_end_line":173,"code":"        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.v0","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.v0#L154-L154","kind":"function","name":"v0","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":154,"end_line":154,"context_start_line":134,"context_end_line":174,"code":"        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.v1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.v1#L155-L155","kind":"function","name":"v1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":155,"end_line":155,"context_start_line":135,"context_end_line":175,"code":"        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.v2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.v2#L156-L156","kind":"function","name":"v2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":156,"end_line":156,"context_start_line":136,"context_end_line":176,"code":"        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d01","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d01#L178-L178","kind":"function","name":"d01","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":178,"end_line":178,"context_start_line":158,"context_end_line":198,"code":"        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d11","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d11#L183-L183","kind":"function","name":"d11","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":183,"end_line":183,"context_start_line":163,"context_end_line":203,"code":"        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d21","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d21#L187-L187","kind":"function","name":"d21","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":187,"end_line":187,"context_start_line":167,"context_end_line":207,"code":"        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d02","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d02#L194-L194","kind":"function","name":"d02","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":194,"end_line":194,"context_start_line":174,"context_end_line":214,"code":"        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d12","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d12#L202-L202","kind":"function","name":"d12","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":202,"end_line":202,"context_start_line":182,"context_end_line":222,"code":"        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d22","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d22#L206-L206","kind":"function","name":"d22","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":206,"end_line":206,"context_start_line":186,"context_end_line":226,"code":"        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d01v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d01v#L210-L210","kind":"function","name":"d01v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":210,"end_line":210,"context_start_line":190,"context_end_line":230,"code":"        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d11v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d11v#L217-L217","kind":"function","name":"d11v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":217,"end_line":217,"context_start_line":197,"context_end_line":237,"code":"        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d21v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d21v#L221-L221","kind":"function","name":"d21v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":221,"end_line":221,"context_start_line":201,"context_end_line":241,"code":"        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d02v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d02v#L227-L227","kind":"function","name":"d02v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":227,"end_line":227,"context_start_line":207,"context_end_line":247,"code":"        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d12v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d12v#L234-L234","kind":"function","name":"d12v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":234,"end_line":234,"context_start_line":214,"context_end_line":254,"code":"        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d22v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.d22v#L242-L242","kind":"function","name":"d22v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":242,"end_line":242,"context_start_line":222,"context_end_line":262,"code":"        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.pos0key1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.pos0key1#L259-L259","kind":"function","name":"pos0key1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":259,"end_line":259,"context_start_line":239,"context_end_line":279,"code":"        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.pos2key2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.pos2key2#L261-L261","kind":"function","name":"pos2key2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":261,"end_line":261,"context_start_line":241,"context_end_line":281,"code":"        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.pos2key2dict","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.pos2key2dict#L265-L265","kind":"function","name":"pos2key2dict","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":265,"end_line":265,"context_start_line":245,"context_end_line":285,"code":"        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})\n        def f(*x:str): pass\n        self.assertEquals(f.__annotations__, {'x': str})\n        def f(**x:float): pass\n        self.assertEquals(f.__annotations__, {'x': float})","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.f","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.f#L474-L476","kind":"function","name":"f","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":474,"end_line":476,"context_start_line":454,"context_end_line":496,"code":"        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.null","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.null#L299-L299","kind":"function","name":"null","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":299,"end_line":299,"context_start_line":279,"context_end_line":319,"code":"        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})\n        def f(*x:str): pass\n        self.assertEquals(f.__annotations__, {'x': str})\n        def f(**x:float): pass\n        self.assertEquals(f.__annotations__, {'x': float})\n        def f(x, y:1+2): pass\n        self.assertEquals(f.__annotations__, {'y': 3})\n        def f(a, b:1, c:2, d): pass\n        self.assertEquals(f.__annotations__, {'b': 1, 'c': 2})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6): pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6, h:7, i=8, j:9=10,\n              **k:11) -> 12: pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6, 'h': 7, 'j': 9,\n                           'k': 11, 'return': 12})\n        # Check for SF Bug #1697248 - mixing decorators and a return annotation\n        def null(x): return x\n        @null\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n\n        # test closures with a variety of oparg's\n        closure = 1\n        def f(): return closure\n        def f(x=1): return closure\n        def f(*, k=1): return closure\n        def f() -> int: return closure\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.foo","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.foo#L340-L342","kind":"function","name":"foo","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":340,"end_line":342,"context_start_line":320,"context_end_line":362,"code":"        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_inner","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_inner#L417-L430","kind":"function","name":"test_inner","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":417,"end_line":430,"context_start_line":397,"context_end_line":450,"code":"        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.g1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.g1#L435-L435","kind":"function","name":"g1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":435,"end_line":435,"context_start_line":415,"context_end_line":455,"code":"        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.g2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.g2#L436-L436","kind":"function","name":"g2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":436,"end_line":436,"context_start_line":416,"context_end_line":456,"code":"\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.Squares","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.Squares#L528-L539","kind":"class","name":"Squares","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":528,"end_line":539,"context_start_line":508,"context_end_line":559,"code":"    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.B","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.B#L712-L712","kind":"class","name":"B","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":712,"end_line":712,"context_start_line":692,"context_end_line":732,"code":"        x = {'one' or 'two': 1 or 2}\n        x = {'one': 1, 'two': 2}\n        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.B2","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.B2#L713-L713","kind":"class","name":"B2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":713,"end_line":713,"context_start_line":693,"context_end_line":733,"code":"        x = {'one': 1, 'two': 2}\n        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.C1","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.C1#L714-L714","kind":"class","name":"C1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":714,"end_line":714,"context_start_line":694,"context_end_line":734,"code":"        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.C2","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.C2#L715-L715","kind":"class","name":"C2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":715,"end_line":715,"context_start_line":695,"context_end_line":735,"code":"        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.D","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.D#L716-L716","kind":"class","name":"D","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":716,"end_line":716,"context_start_line":696,"context_end_line":736,"code":"\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.C","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.C#L717-L720","kind":"class","name":"C","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":717,"end_line":720,"context_start_line":697,"context_end_line":740,"code":"        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.class_decorator","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.class_decorator#L725-L725","kind":"function","name":"class_decorator","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":725,"end_line":725,"context_start_line":705,"context_end_line":745,"code":"\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.G","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.G#L727-L727","kind":"class","name":"G","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":727,"end_line":727,"context_start_line":707,"context_end_line":747,"code":"    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_in_func","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_in_func#L758-L759","kind":"function","name":"test_in_func","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":758,"end_line":759,"context_start_line":738,"context_end_line":779,"code":"        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Apple'), (3, 'Banana'), (3, 'Coconut'),\n                          (4, 'Apple'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Apple'), (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(i, s) for i in nums for s in [f for f in strs if \"n\" in f]],\n                         [(1, 'Banana'), (1, 'Coconut'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Banana'), (3, 'Coconut'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(lambda a:[a**i for i in range(a+1)])(j) for j in range(5)],\n                         [[1], [1, 1], [1, 2, 4], [1, 3, 9, 27], [1, 4, 16, 64, 256]])\n\n        def test_in_func(l):\n            return [0 < x < 3 for x in l if x > 2]\n\n        self.assertEqual(test_in_func(nums), [False, False, False])\n\n        def test_nested_front():\n            self.assertEqual([[y for y in [x, x + 1]] for x in [1,3,5]],\n                             [[1, 2], [3, 4], [5, 6]])\n\n        test_nested_front()\n\n        check_syntax_error(self, \"[i, s for i in nums for s in strs]\")\n        check_syntax_error(self, \"[x if y]\")\n\n        suppliers = [\n          (1, \"Boeing\"),\n          (2, \"Ford\"),\n          (3, \"Macdonalds\")\n        ]\n\n        parts = [\n          (10, \"Airliner\"),","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_nested_front","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.test_nested_front#L763-L765","kind":"function","name":"test_nested_front","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":763,"end_line":765,"context_start_line":743,"context_end_line":785,"code":"        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Apple'), (3, 'Banana'), (3, 'Coconut'),\n                          (4, 'Apple'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Apple'), (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(i, s) for i in nums for s in [f for f in strs if \"n\" in f]],\n                         [(1, 'Banana'), (1, 'Coconut'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Banana'), (3, 'Coconut'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(lambda a:[a**i for i in range(a+1)])(j) for j in range(5)],\n                         [[1], [1, 1], [1, 2, 4], [1, 3, 9, 27], [1, 4, 16, 64, 256]])\n\n        def test_in_func(l):\n            return [0 < x < 3 for x in l if x > 2]\n\n        self.assertEqual(test_in_func(nums), [False, False, False])\n\n        def test_nested_front():\n            self.assertEqual([[y for y in [x, x + 1]] for x in [1,3,5]],\n                             [[1, 2], [3, 4], [5, 6]])\n\n        test_nested_front()\n\n        check_syntax_error(self, \"[i, s for i in nums for s in strs]\")\n        check_syntax_error(self, \"[x if y]\")\n\n        suppliers = [\n          (1, \"Boeing\"),\n          (2, \"Ford\"),\n          (3, \"Macdonalds\")\n        ]\n\n        parts = [\n          (10, \"Airliner\"),\n          (20, \"Engine\"),\n          (30, \"Cheeseburger\")\n        ]\n\n        suppart = [\n          (1, 10), (1, 20), (2, 20), (3, 30)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.manager","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.manager#L854-L858","kind":"class","name":"manager","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":854,"end_line":858,"context_start_line":834,"context_end_line":878,"code":"    def testComprehensionSpecials(self):\n        # test for outmost iterable precomputation\n        x = 10; g = (i for i in range(x)); x = 5\n        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n\n    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"\n            print(x)\n            return ret","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar._checkeval","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar._checkeval#L875-L878","kind":"function","name":"_checkeval","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":875,"end_line":878,"context_start_line":855,"context_end_line":898,"code":"            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"\n            print(x)\n            return ret\n\n        # the next line is not allowed anymore\n        #self.assertEqual([ x() for x in lambda: True, lambda: False if x() ], [True])\n        self.assertEqual([ x() for x in (lambda: True, lambda: False) if x() ], [True])\n        self.assertEqual([ x(False) for x in (lambda x: False if x else True, lambda x: True if x else False) if x(False) ], [True])\n        self.assertEqual((5 if 1 else _checkeval(\"check 1\", 0)), 5)\n        self.assertEqual((_checkeval(\"check 2\", 0) if 0 else 5), 5)\n        self.assertEqual((5 and 6 if 0 else 1), 1)\n        self.assertEqual(((5 and 6) if 0 else 1), 1)\n        self.assertEqual((5 and (6 if 1 else 1)), 6)\n        self.assertEqual((0 or _checkeval(\"check 3\", 2) if 0 else 3), 3)\n        self.assertEqual((1 or _checkeval(\"check 4\", 2) if 1 else _checkeval(\"check 5\", 3)), 1)\n        self.assertEqual((0 or 5 if 1 else _checkeval(\"check 6\", 3)), 5)\n        self.assertEqual((not 5 if 1 else 1), False)\n        self.assertEqual((not 5 if 0 else 1), 1)\n        self.assertEqual((6 + 1 if 1 else 2), 7)\n        self.assertEqual((6 - 1 if 1 else 2), 5)\n        self.assertEqual((6 * 2 if 1 else 4), 12)\n        self.assertEqual((6 / 2 if 1 else 3), 3)\n        self.assertEqual((6 < 4 if 0 else 2), 2)","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.__init__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.__init__#L529-L531","kind":"function","name":"__init__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":529,"end_line":531,"context_start_line":509,"context_end_line":551,"code":"        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.__len__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.__len__#L532-L532","kind":"function","name":"__len__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":532,"end_line":532,"context_start_line":512,"context_end_line":552,"code":"        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.__getitem__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.__getitem__#L533-L539","kind":"function","name":"__getitem__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":533,"end_line":539,"context_start_line":513,"context_end_line":559,"code":"\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.meth1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.meth1#L718-L718","kind":"function","name":"meth1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":718,"end_line":718,"context_start_line":698,"context_end_line":738,"code":"        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.meth2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.meth2#L719-L719","kind":"function","name":"meth2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":719,"end_line":719,"context_start_line":699,"context_end_line":739,"code":"        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.meth3","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.meth3#L720-L720","kind":"function","name":"meth3","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":720,"end_line":720,"context_start_line":700,"context_end_line":740,"code":"        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.__enter__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.__enter__#L855-L856","kind":"function","name":"__enter__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":855,"end_line":856,"context_start_line":835,"context_end_line":876,"code":"        # test for outmost iterable precomputation\n        x = 10; g = (i for i in range(x)); x = 5\n        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n\n    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.__exit__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar.__exit__#L857-L858","kind":"function","name":"__exit__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":857,"end_line":858,"context_start_line":837,"context_end_line":878,"code":"        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n\n    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"\n            print(x)\n            return ret","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.mixed-spaces-tabs","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.mixed-spaces-tabs#L1-L4","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.mixed-spaces-tabs","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/mixed-spaces-tabs.py","language":"python","start_line":1,"end_line":4,"context_start_line":1,"context_end_line":4,"code":"def main():\n\tprint \"hello\"\n\t# 1 tab = 8 spaces in Python 2\n        return","source_hash":"2fe77b01e6e989bc120dd4985bbe06beb15ab84e8fb81350fc53b480f5f1cd9b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf#L1-L945","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":1,"end_line":945,"context_start_line":1,"context_end_line":945,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\\n        + 1\n        self.assertEquals(x, 2, 'backslash for line continuation')\n\n        # Backslash does not means continuation in comments :\\\n        x = 0\n        self.assertEquals(x, 0, 'backslash ending comment')\n\n    def testPlainIntegers(self):\n        self.assertEquals(type(000), type(0))\n        self.assertEquals(0xff, 255)\n        self.assertEquals(0o377, 255)\n        self.assertEquals(2147483647, 0o17777777777)\n        self.assertEquals(0b1001, 9)\n        # \"0x\" is not a valid literal\n        self.assertRaises(SyntaxError, eval, \"0x\")\n        from sys import maxsize\n        if maxsize == 2147483647:\n            self.assertEquals(-2147483647-1, -0o20000000000)\n            # XXX -2147483648\n            self.assert_(0o37777777777 > 0)\n            self.assert_(0xffffffff > 0)\n            self.assert_(0b1111111111111111111111111111111 > 0)\n            for s in ('2147483648', '0o40000000000', '0x100000000',\n                      '0b10000000000000000000000000000000'):\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        elif maxsize == 9223372036854775807:\n            self.assertEquals(-9223372036854775807-1, -0o1000000000000000000000)\n            self.assert_(0o1777777777777777777777 > 0)\n            self.assert_(0xffffffffffffffff > 0)\n            self.assert_(0b11111111111111111111111111111111111111111111111111111111111111 > 0)\n            for s in '9223372036854775808', '0o2000000000000000000000', \\\n                     '0x10000000000000000', \\\n                     '0b100000000000000000000000000000000000000000000000000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxsize value %r' % maxsize)\n\n    def testLongIntegers(self):\n        x = 0\n        x = 0xffffffffffffffff\n        x = 0Xffffffffffffffff\n        x = 0o77777777777777777\n        x = 0O77777777777777777\n        x = 123456789012345678901234567890\n        x = 0b100000000000000000000000000000000000000000000000000000000000000000000\n        x = 0B111111111111111111111111111111111111111111111111111111111111111111111\n\n    def testUnderscoresInNumbers(self):\n        # Integers\n        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67\n        x = 0O_45_67\n\n        # Floats\n        x = 3_1.4\n        x = 03_1.4\n        x = 3_1.\n        x = .3_1\n        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4\n        x = 3_1e+4_1\n        x = 3_1E-4_1\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})\n        def f(*x:str): pass\n        self.assertEquals(f.__annotations__, {'x': str})\n        def f(**x:float): pass\n        self.assertEquals(f.__annotations__, {'x': float})\n        def f(x, y:1+2): pass\n        self.assertEquals(f.__annotations__, {'y': 3})\n        def f(a, b:1, c:2, d): pass\n        self.assertEquals(f.__annotations__, {'b': 1, 'c': 2})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6): pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6, h:7, i=8, j:9=10,\n              **k:11) -> 12: pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6, 'h': 7, 'j': 9,\n                           'k': 11, 'return': 12})\n        # Check for SF Bug #1697248 - mixing decorators and a return annotation\n        def null(x): return x\n        @null\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n\n        # test closures with a variety of oparg's\n        closure = 1\n        def f(): return closure\n        def f(x=1): return closure\n        def f(*, k=1): return closure\n        def f() -> int: return closure\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError as msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError) as msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]\n# ... truncated ...","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":true}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.TokenTests","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.TokenTests#L17-L114","kind":"class","name":"TokenTests","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":17,"end_line":114,"context_start_line":1,"context_end_line":134,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\\n        + 1\n        self.assertEquals(x, 2, 'backslash for line continuation')\n\n        # Backslash does not means continuation in comments :\\\n        x = 0\n        self.assertEquals(x, 0, 'backslash ending comment')\n\n    def testPlainIntegers(self):\n        self.assertEquals(type(000), type(0))\n        self.assertEquals(0xff, 255)\n        self.assertEquals(0o377, 255)\n        self.assertEquals(2147483647, 0o17777777777)\n        self.assertEquals(0b1001, 9)\n        # \"0x\" is not a valid literal\n        self.assertRaises(SyntaxError, eval, \"0x\")\n        from sys import maxsize\n        if maxsize == 2147483647:\n            self.assertEquals(-2147483647-1, -0o20000000000)\n            # XXX -2147483648\n            self.assert_(0o37777777777 > 0)\n            self.assert_(0xffffffff > 0)\n            self.assert_(0b1111111111111111111111111111111 > 0)\n            for s in ('2147483648', '0o40000000000', '0x100000000',\n                      '0b10000000000000000000000000000000'):\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        elif maxsize == 9223372036854775807:\n            self.assertEquals(-9223372036854775807-1, -0o1000000000000000000000)\n            self.assert_(0o1777777777777777777777 > 0)\n            self.assert_(0xffffffffffffffff > 0)\n            self.assert_(0b11111111111111111111111111111111111111111111111111111111111111 > 0)\n            for s in '9223372036854775808', '0o2000000000000000000000', \\\n                     '0x10000000000000000', \\\n                     '0b100000000000000000000000000000000000000000000000000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxsize value %r' % maxsize)\n\n    def testLongIntegers(self):\n        x = 0\n        x = 0xffffffffffffffff\n        x = 0Xffffffffffffffff\n        x = 0o77777777777777777\n        x = 0O77777777777777777\n        x = 123456789012345678901234567890\n        x = 0b100000000000000000000000000000000000000000000000000000000000000000000\n        x = 0B111111111111111111111111111111111111111111111111111111111111111111111\n\n    def testUnderscoresInNumbers(self):\n        # Integers\n        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67\n        x = 0O_45_67\n\n        # Floats\n        x = 3_1.4\n        x = 03_1.4\n        x = 3_1.\n        x = .3_1\n        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4\n        x = 3_1e+4_1\n        x = 3_1E-4_1\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.GrammarTests","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.GrammarTests#L116-L938","kind":"class","name":"GrammarTests","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":116,"end_line":938,"context_start_line":96,"context_end_line":945,"code":"\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})\n        def f(*x:str): pass\n        self.assertEquals(f.__annotations__, {'x': str})\n        def f(**x:float): pass\n        self.assertEquals(f.__annotations__, {'x': float})\n        def f(x, y:1+2): pass\n        self.assertEquals(f.__annotations__, {'y': 3})\n        def f(a, b:1, c:2, d): pass\n        self.assertEquals(f.__annotations__, {'b': 1, 'c': 2})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6): pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6, h:7, i=8, j:9=10,\n              **k:11) -> 12: pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6, 'h': 7, 'j': 9,\n                           'k': 11, 'return': 12})\n        # Check for SF Bug #1697248 - mixing decorators and a return annotation\n        def null(x): return x\n        @null\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n\n        # test closures with a variety of oparg's\n        closure = 1\n        def f(): return closure\n        def f(x=1): return closure\n        def f(*, k=1): return closure\n        def f() -> int: return closure\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError as msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError) as msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]\n        x = time.time()\n        x = sys.modules['time'].time()\n        a = '01234'\n        c = a[0]\n        c = a[-1]\n        s = a[0:5]\n        s = a[:5]\n        s = a[0:]\n        s = a[:]\n        s = a[-5:]\n        s = a[:-1]\n        s = a[-4:-3]\n        # A rough test of SF bug 1333982.  http://python.org/sf/1333982\n        # The testing here is fairly incomplete.\n        # Test cases should include: commas with 1 and 2 colons\n        d = {}\n        d[1] = 1\n        d[1,] = 2\n        d[1,2] = 3\n        d[1,2,3] = 4\n        L = list(d)\n        L.sort(key=lambda x: x if isinstance(x, tuple) else ())\n        self.assertEquals(str(L), '[1, (1,), (1, 2), (1, 2, 3)]')\n\n    def testAtoms(self):\n        ### atom: '(' [testlist] ')' | '[' [testlist] ']' | '{' [dictsetmaker] '}' | NAME | NUMBER | STRING\n        ### dictsetmaker: (test ':' test (',' test ':' test)* [',']) | (test (',' test)* [','])\n\n        x = (1)\n        x = (1 or 2 or 3)\n        x = (1 or 2 or 3, 2, 3)\n\n        x = []\n        x = [1]\n        x = [1 or 2 or 3]\n        x = [1 or 2 or 3, 2, 3]\n        x = []\n\n        x = {}\n        x = {'one': 1}\n        x = {'one': 1,}\n        x = {'one' or 'two': 1 or 2}\n        x = {'one': 1, 'two': 2}\n        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Apple'), (3, 'Banana'), (3, 'Coconut'),\n                          (4, 'Apple'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Apple'), (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(i, s) for i in nums for s in [f for f in s\n# ... truncated ...","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":true}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_main","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_main#L941-L942","kind":"function","name":"test_main","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":941,"end_line":942,"context_start_line":921,"context_end_line":945,"code":"jumps over\nthe 'lazy' dog.\n'''\n        self.assertEquals(x, y)\n        y = \"\\n\\\nThe \\\"quick\\\"\\n\\\nbrown fox\\n\\\njumps over\\n\\\nthe 'lazy' dog.\\n\\\n\"\n        self.assertEquals(x, y)\n        y = '\\n\\\nThe \\\"quick\\\"\\n\\\nbrown fox\\n\\\njumps over\\n\\\nthe \\'lazy\\' dog.\\n\\\n'\n        self.assertEquals(x, y)\n\n\ndef test_main():\n    run_unittest(TokenTests, GrammarTests)\n\nif __name__ == '__main__':\n    test_main()","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testBackslash","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testBackslash#L19-L27","kind":"function","name":"testBackslash","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":19,"end_line":27,"context_start_line":1,"context_end_line":47,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\\n        + 1\n        self.assertEquals(x, 2, 'backslash for line continuation')\n\n        # Backslash does not means continuation in comments :\\\n        x = 0\n        self.assertEquals(x, 0, 'backslash ending comment')\n\n    def testPlainIntegers(self):\n        self.assertEquals(type(000), type(0))\n        self.assertEquals(0xff, 255)\n        self.assertEquals(0o377, 255)\n        self.assertEquals(2147483647, 0o17777777777)\n        self.assertEquals(0b1001, 9)\n        # \"0x\" is not a valid literal\n        self.assertRaises(SyntaxError, eval, \"0x\")\n        from sys import maxsize\n        if maxsize == 2147483647:\n            self.assertEquals(-2147483647-1, -0o20000000000)\n            # XXX -2147483648\n            self.assert_(0o37777777777 > 0)\n            self.assert_(0xffffffff > 0)\n            self.assert_(0b1111111111111111111111111111111 > 0)\n            for s in ('2147483648', '0o40000000000', '0x100000000',\n                      '0b10000000000000000000000000000000'):\n                try:\n                    x = eval(s)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testPlainIntegers","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testPlainIntegers#L29-L63","kind":"function","name":"testPlainIntegers","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":29,"end_line":63,"context_start_line":9,"context_end_line":83,"code":"# regrtest.py.\n\nfrom test.support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\\n        + 1\n        self.assertEquals(x, 2, 'backslash for line continuation')\n\n        # Backslash does not means continuation in comments :\\\n        x = 0\n        self.assertEquals(x, 0, 'backslash ending comment')\n\n    def testPlainIntegers(self):\n        self.assertEquals(type(000), type(0))\n        self.assertEquals(0xff, 255)\n        self.assertEquals(0o377, 255)\n        self.assertEquals(2147483647, 0o17777777777)\n        self.assertEquals(0b1001, 9)\n        # \"0x\" is not a valid literal\n        self.assertRaises(SyntaxError, eval, \"0x\")\n        from sys import maxsize\n        if maxsize == 2147483647:\n            self.assertEquals(-2147483647-1, -0o20000000000)\n            # XXX -2147483648\n            self.assert_(0o37777777777 > 0)\n            self.assert_(0xffffffff > 0)\n            self.assert_(0b1111111111111111111111111111111 > 0)\n            for s in ('2147483648', '0o40000000000', '0x100000000',\n                      '0b10000000000000000000000000000000'):\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        elif maxsize == 9223372036854775807:\n            self.assertEquals(-9223372036854775807-1, -0o1000000000000000000000)\n            self.assert_(0o1777777777777777777777 > 0)\n            self.assert_(0xffffffffffffffff > 0)\n            self.assert_(0b11111111111111111111111111111111111111111111111111111111111111 > 0)\n            for s in '9223372036854775808', '0o2000000000000000000000', \\\n                     '0x10000000000000000', \\\n                     '0b100000000000000000000000000000000000000000000000000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxsize value %r' % maxsize)\n\n    def testLongIntegers(self):\n        x = 0\n        x = 0xffffffffffffffff\n        x = 0Xffffffffffffffff\n        x = 0o77777777777777777\n        x = 0O77777777777777777\n        x = 123456789012345678901234567890\n        x = 0b100000000000000000000000000000000000000000000000000000000000000000000\n        x = 0B111111111111111111111111111111111111111111111111111111111111111111111\n\n    def testUnderscoresInNumbers(self):\n        # Integers\n        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testLongIntegers","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testLongIntegers#L65-L73","kind":"function","name":"testLongIntegers","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":65,"end_line":73,"context_start_line":45,"context_end_line":93,"code":"                      '0b10000000000000000000000000000000'):\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        elif maxsize == 9223372036854775807:\n            self.assertEquals(-9223372036854775807-1, -0o1000000000000000000000)\n            self.assert_(0o1777777777777777777777 > 0)\n            self.assert_(0xffffffffffffffff > 0)\n            self.assert_(0b11111111111111111111111111111111111111111111111111111111111111 > 0)\n            for s in '9223372036854775808', '0o2000000000000000000000', \\\n                     '0x10000000000000000', \\\n                     '0b100000000000000000000000000000000000000000000000000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxsize value %r' % maxsize)\n\n    def testLongIntegers(self):\n        x = 0\n        x = 0xffffffffffffffff\n        x = 0Xffffffffffffffff\n        x = 0o77777777777777777\n        x = 0O77777777777777777\n        x = 123456789012345678901234567890\n        x = 0b100000000000000000000000000000000000000000000000000000000000000000000\n        x = 0B111111111111111111111111111111111111111111111111111111111111111111111\n\n    def testUnderscoresInNumbers(self):\n        # Integers\n        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67\n        x = 0O_45_67\n\n        # Floats\n        x = 3_1.4\n        x = 03_1.4\n        x = 3_1.\n        x = .3_1\n        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testUnderscoresInNumbers","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testUnderscoresInNumbers#L75-L95","kind":"function","name":"testUnderscoresInNumbers","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":75,"end_line":95,"context_start_line":55,"context_end_line":115,"code":"            for s in '9223372036854775808', '0o2000000000000000000000', \\\n                     '0x10000000000000000', \\\n                     '0b100000000000000000000000000000000000000000000000000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxsize value %r' % maxsize)\n\n    def testLongIntegers(self):\n        x = 0\n        x = 0xffffffffffffffff\n        x = 0Xffffffffffffffff\n        x = 0o77777777777777777\n        x = 0O77777777777777777\n        x = 123456789012345678901234567890\n        x = 0b100000000000000000000000000000000000000000000000000000000000000000000\n        x = 0B111111111111111111111111111111111111111111111111111111111111111111111\n\n    def testUnderscoresInNumbers(self):\n        # Integers\n        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67\n        x = 0O_45_67\n\n        # Floats\n        x = 3_1.4\n        x = 03_1.4\n        x = 3_1.\n        x = .3_1\n        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4\n        x = 3_1e+4_1\n        x = 3_1E-4_1\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testFloats","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testFloats#L97-L109","kind":"function","name":"testFloats","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":97,"end_line":109,"context_start_line":77,"context_end_line":129,"code":"        x = 1_0\n        x = 123_456_7_89\n        x = 0xabc_123_4_5\n        x = 0X_abc_123\n        x = 0B11_01\n        x = 0b_11_01\n        x = 0o45_67\n        x = 0O_45_67\n\n        # Floats\n        x = 3_1.4\n        x = 03_1.4\n        x = 3_1.\n        x = .3_1\n        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4\n        x = 3_1e+4_1\n        x = 3_1E-4_1\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testEllipsis","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testEllipsis#L111-L114","kind":"function","name":"testEllipsis","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":111,"end_line":114,"context_start_line":91,"context_end_line":134,"code":"        x = 3.1_4\n        x = 0_3.1_4\n        x = 3e1_4\n        x = 3_1e+4_1\n        x = 3_1E-4_1\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testEvalInput","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testEvalInput#L127-L129","kind":"function","name":"testEvalInput","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":127,"end_line":129,"context_start_line":107,"context_end_line":149,"code":"        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\n    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testFuncdef","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testFuncdef#L131-L313","kind":"function","name":"testFuncdef","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":131,"end_line":313,"context_start_line":111,"context_end_line":333,"code":"    def testEllipsis(self):\n        x = ...\n        self.assert_(x is Ellipsis)\n        self.assertRaises(SyntaxError, eval, \".. .\")\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})\n        def f(*x:str): pass\n        self.assertEquals(f.__annotations__, {'x': str})\n        def f(**x:float): pass\n        self.assertEquals(f.__annotations__, {'x': float})\n        def f(x, y:1+2): pass\n        self.assertEquals(f.__annotations__, {'y': 3})\n        def f(a, b:1, c:2, d): pass\n        self.assertEquals(f.__annotations__, {'b': 1, 'c': 2})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6): pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6, h:7, i=8, j:9=10,\n              **k:11) -> 12: pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6, 'h': 7, 'j': 9,\n                           'k': 11, 'return': 12})\n        # Check for SF Bug #1697248 - mixing decorators and a return annotation\n        def null(x): return x\n        @null\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n\n        # test closures with a variety of oparg's\n        closure = 1\n        def f(): return closure\n        def f(x=1): return closure\n        def f(*, k=1): return closure\n        def f() -> int: return closure\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testLambdef","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testLambdef#L315-L331","kind":"function","name":"testLambdef","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":315,"end_line":331,"context_start_line":295,"context_end_line":351,"code":"        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6, 'h': 7, 'j': 9,\n                           'k': 11, 'return': 12})\n        # Check for SF Bug #1697248 - mixing decorators and a return annotation\n        def null(x): return x\n        @null\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n\n        # test closures with a variety of oparg's\n        closure = 1\n        def f(): return closure\n        def f(x=1): return closure\n        def f(*, k=1): return closure\n        def f() -> int: return closure\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testSimpleStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testSimpleStmt#L337-L343","kind":"function","name":"testSimpleStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":337,"end_line":343,"context_start_line":317,"context_end_line":363,"code":"        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testExprStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testExprStmt#L348-L359","kind":"function","name":"testExprStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":348,"end_line":359,"context_start_line":328,"context_end_line":379,"code":"        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testDelStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testDelStmt#L361-L368","kind":"function","name":"testDelStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":361,"end_line":368,"context_start_line":341,"context_end_line":388,"code":"            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testPassStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testPassStmt#L370-L372","kind":"function","name":"testPassStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":370,"end_line":372,"context_start_line":350,"context_end_line":392,"code":"        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testBreakStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testBreakStmt#L377-L379","kind":"function","name":"testBreakStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":377,"end_line":379,"context_start_line":357,"context_end_line":399,"code":"\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testContinueStmt","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testContinueStmt#L381-L405","kind":"function","name":"testContinueStmt","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":381,"end_line":405,"context_start_line":361,"context_end_line":425,"code":"    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_break_continue_loop","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_break_continue_loop#L407-L431","kind":"function","name":"test_break_continue_loop","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":407,"end_line":431,"context_start_line":387,"context_end_line":451,"code":"        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testReturn","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testReturn#L433-L439","kind":"function","name":"testReturn","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":433,"end_line":439,"context_start_line":413,"context_end_line":459,"code":"        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testYield","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testYield#L441-L442","kind":"function","name":"testYield","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":441,"end_line":442,"context_start_line":421,"context_end_line":462,"code":"                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testRaise","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testRaise#L444-L449","kind":"function","name":"testRaise","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":444,"end_line":449,"context_start_line":424,"context_end_line":469,"code":"                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testImport","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testImport#L451-L462","kind":"function","name":"testImport","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":451,"end_line":462,"context_start_line":431,"context_end_line":482,"code":"        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testGlobal","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testGlobal#L464-L468","kind":"function","name":"testGlobal","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":464,"end_line":468,"context_start_line":444,"context_end_line":488,"code":"    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testNonlocal","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testNonlocal#L470-L476","kind":"function","name":"testNonlocal","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":470,"end_line":476,"context_start_line":450,"context_end_line":496,"code":"\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testAssert","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testAssert#L478-L490","kind":"function","name":"testAssert","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":478,"end_line":490,"context_start_line":458,"context_end_line":510,"code":"        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testIf","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testIf#L495-L506","kind":"function","name":"testIf","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":495,"end_line":506,"context_start_line":475,"context_end_line":526,"code":"            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testWhile","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testWhile#L508-L521","kind":"function","name":"testWhile","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":508,"end_line":521,"context_start_line":488,"context_end_line":541,"code":"        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testFor","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testFor#L523-L548","kind":"function","name":"testFor","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":523,"end_line":548,"context_start_line":503,"context_end_line":568,"code":"        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError as msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testTry","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testTry#L550-L571","kind":"function","name":"testTry","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":550,"end_line":571,"context_start_line":530,"context_end_line":591,"code":"                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError as msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError) as msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testSuite","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testSuite#L573-L585","kind":"function","name":"testSuite","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":573,"end_line":585,"context_start_line":553,"context_end_line":605,"code":"        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError as msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError) as msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testTest","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testTest#L588-L597","kind":"function","name":"testTest","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":588,"end_line":597,"context_start_line":568,"context_end_line":617,"code":"        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError) as msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testComparison","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testComparison#L599-L614","kind":"function","name":"testComparison","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":599,"end_line":614,"context_start_line":579,"context_end_line":634,"code":"            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testBinaryMaskOps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testBinaryMaskOps#L616-L619","kind":"function","name":"testBinaryMaskOps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":616,"end_line":619,"context_start_line":596,"context_end_line":639,"code":"        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testShiftOps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testShiftOps#L621-L624","kind":"function","name":"testShiftOps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":621,"end_line":624,"context_start_line":601,"context_end_line":644,"code":"        ### comp_op: '<'|'>'|'=='|'>='|'<='|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testAdditiveOps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testAdditiveOps#L626-L630","kind":"function","name":"testAdditiveOps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":626,"end_line":630,"context_start_line":606,"context_end_line":650,"code":"        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testMultiplicativeOps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testMultiplicativeOps#L632-L636","kind":"function","name":"testMultiplicativeOps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":632,"end_line":636,"context_start_line":612,"context_end_line":656,"code":"        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]\n        x = time.time()\n        x = sys.modules['time'].time()\n        a = '01234'\n        c = a[0]\n        c = a[-1]\n        s = a[0:5]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testUnaryOps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testUnaryOps#L638-L643","kind":"function","name":"testUnaryOps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":638,"end_line":643,"context_start_line":618,"context_end_line":663,"code":"        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]\n        x = time.time()\n        x = sys.modules['time'].time()\n        a = '01234'\n        c = a[0]\n        c = a[-1]\n        s = a[0:5]\n        s = a[:5]\n        s = a[0:]\n        s = a[:]\n        s = a[-5:]\n        s = a[:-1]\n        s = a[-4:-3]\n        # A rough test of SF bug 1333982.  http://python.org/sf/1333982","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testSelectors","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testSelectors#L645-L673","kind":"function","name":"testSelectors","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":645,"end_line":673,"context_start_line":625,"context_end_line":693,"code":"\n    def testAdditiveOps(self):\n        x = 1\n        x = 1 + 1\n        x = 1 - 1 - 1\n        x = 1 - 1 + 1 - 1 + 1\n\n    def testMultiplicativeOps(self):\n        x = 1 * 1\n        x = 1 / 1\n        x = 1 % 1\n        x = 1 / 1 * 1 % 1\n\n    def testUnaryOps(self):\n        x = +1\n        x = -1\n        x = ~1\n        x = ~1 ^ 1 & 1 | 1 & 1 ^ -1\n        x = -1*1/1 + 1*1 - ---1*1\n\n    def testSelectors(self):\n        ### trailer: '(' [testlist] ')' | '[' subscript ']' | '.' NAME\n        ### subscript: expr | [expr] ':' [expr]\n\n        import sys, time\n        c = sys.path[0]\n        x = time.time()\n        x = sys.modules['time'].time()\n        a = '01234'\n        c = a[0]\n        c = a[-1]\n        s = a[0:5]\n        s = a[:5]\n        s = a[0:]\n        s = a[:]\n        s = a[-5:]\n        s = a[:-1]\n        s = a[-4:-3]\n        # A rough test of SF bug 1333982.  http://python.org/sf/1333982\n        # The testing here is fairly incomplete.\n        # Test cases should include: commas with 1 and 2 colons\n        d = {}\n        d[1] = 1\n        d[1,] = 2\n        d[1,2] = 3\n        d[1,2,3] = 4\n        L = list(d)\n        L.sort(key=lambda x: x if isinstance(x, tuple) else ())\n        self.assertEquals(str(L), '[1, (1,), (1, 2), (1, 2, 3)]')\n\n    def testAtoms(self):\n        ### atom: '(' [testlist] ')' | '[' [testlist] ']' | '{' [dictsetmaker] '}' | NAME | NUMBER | STRING\n        ### dictsetmaker: (test ':' test (',' test ':' test)* [',']) | (test (',' test)* [','])\n\n        x = (1)\n        x = (1 or 2 or 3)\n        x = (1 or 2 or 3, 2, 3)\n\n        x = []\n        x = [1]\n        x = [1 or 2 or 3]\n        x = [1 or 2 or 3, 2, 3]\n        x = []\n\n        x = {}\n        x = {'one': 1}\n        x = {'one': 1,}\n        x = {'one' or 'two': 1 or 2}\n        x = {'one': 1, 'two': 2}","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testAtoms","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testAtoms#L675-L704","kind":"function","name":"testAtoms","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":675,"end_line":704,"context_start_line":655,"context_end_line":724,"code":"        c = a[-1]\n        s = a[0:5]\n        s = a[:5]\n        s = a[0:]\n        s = a[:]\n        s = a[-5:]\n        s = a[:-1]\n        s = a[-4:-3]\n        # A rough test of SF bug 1333982.  http://python.org/sf/1333982\n        # The testing here is fairly incomplete.\n        # Test cases should include: commas with 1 and 2 colons\n        d = {}\n        d[1] = 1\n        d[1,] = 2\n        d[1,2] = 3\n        d[1,2,3] = 4\n        L = list(d)\n        L.sort(key=lambda x: x if isinstance(x, tuple) else ())\n        self.assertEquals(str(L), '[1, (1,), (1, 2), (1, 2, 3)]')\n\n    def testAtoms(self):\n        ### atom: '(' [testlist] ')' | '[' [testlist] ']' | '{' [dictsetmaker] '}' | NAME | NUMBER | STRING\n        ### dictsetmaker: (test ':' test (',' test ':' test)* [',']) | (test (',' test)* [','])\n\n        x = (1)\n        x = (1 or 2 or 3)\n        x = (1 or 2 or 3, 2, 3)\n\n        x = []\n        x = [1]\n        x = [1 or 2 or 3]\n        x = [1 or 2 or 3, 2, 3]\n        x = []\n\n        x = {}\n        x = {'one': 1}\n        x = {'one': 1,}\n        x = {'one' or 'two': 1 or 2}\n        x = {'one': 1, 'two': 2}\n        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testClassdef","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testClassdef#L710-L727","kind":"function","name":"testClassdef","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":710,"end_line":727,"context_start_line":690,"context_end_line":747,"code":"        x = {'one': 1}\n        x = {'one': 1,}\n        x = {'one' or 'two': 1 or 2}\n        x = {'one': 1, 'two': 2}\n        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testDictcomps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testDictcomps#L729-L734","kind":"function","name":"testDictcomps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":729,"end_line":734,"context_start_line":709,"context_end_line":754,"code":"\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Apple'), (3, 'Banana'), (3, 'Coconut'),\n                          (4, 'Apple'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Apple'), (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(i, s) for i in nums for s in [f for f in strs if \"n\" in f]],\n                         [(1, 'Banana'), (1, 'Coconut'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Banana'), (3, 'Coconut'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Banana'), (5, 'Coconut')])","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testListcomps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testListcomps#L736-L797","kind":"function","name":"testListcomps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":736,"end_line":797,"context_start_line":716,"context_end_line":817,"code":"        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Apple'), (3, 'Banana'), (3, 'Coconut'),\n                          (4, 'Apple'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Apple'), (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(i, s) for i in nums for s in [f for f in strs if \"n\" in f]],\n                         [(1, 'Banana'), (1, 'Coconut'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Banana'), (3, 'Coconut'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(lambda a:[a**i for i in range(a+1)])(j) for j in range(5)],\n                         [[1], [1, 1], [1, 2, 4], [1, 3, 9, 27], [1, 4, 16, 64, 256]])\n\n        def test_in_func(l):\n            return [0 < x < 3 for x in l if x > 2]\n\n        self.assertEqual(test_in_func(nums), [False, False, False])\n\n        def test_nested_front():\n            self.assertEqual([[y for y in [x, x + 1]] for x in [1,3,5]],\n                             [[1, 2], [3, 4], [5, 6]])\n\n        test_nested_front()\n\n        check_syntax_error(self, \"[i, s for i in nums for s in strs]\")\n        check_syntax_error(self, \"[x if y]\")\n\n        suppliers = [\n          (1, \"Boeing\"),\n          (2, \"Ford\"),\n          (3, \"Macdonalds\")\n        ]\n\n        parts = [\n          (10, \"Airliner\"),\n          (20, \"Engine\"),\n          (30, \"Cheeseburger\")\n        ]\n\n        suppart = [\n          (1, 10), (1, 20), (2, 20), (3, 30)\n        ]\n\n        x = [\n          (sname, pname)\n            for (sno, sname) in suppliers\n              for (pno, pname) in parts\n                for (sp_sno, sp_pno) in suppart\n                  if sno == sp_sno and pno == sp_pno\n        ]\n\n        self.assertEqual(x, [('Boeing', 'Airliner'), ('Boeing', 'Engine'), ('Ford', 'Engine'),\n                             ('Macdonalds', 'Cheeseburger')])\n\n    def testGenexps(self):\n        # generator expression tests\n        g = ([x for x in range(10)] for x in range(1))\n        self.assertEqual(next(g), [x for x in range(10)])\n        try:\n            next(g)\n            self.fail('should produce StopIteration exception')\n        except StopIteration:\n            pass\n\n        a = 1\n        try:\n            g = (a for d in a)\n            next(g)\n            self.fail('should produce TypeError')\n        except TypeError:\n            pass\n\n        self.assertEqual(list((x, y) for x in 'abcd' for y in 'abcd'), [(x, y) for x in 'abcd' for y in 'abcd'])","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testGenexps","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testGenexps#L799-L832","kind":"function","name":"testGenexps","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":799,"end_line":832,"context_start_line":779,"context_end_line":852,"code":"          (10, \"Airliner\"),\n          (20, \"Engine\"),\n          (30, \"Cheeseburger\")\n        ]\n\n        suppart = [\n          (1, 10), (1, 20), (2, 20), (3, 30)\n        ]\n\n        x = [\n          (sname, pname)\n            for (sno, sname) in suppliers\n              for (pno, pname) in parts\n                for (sp_sno, sp_pno) in suppart\n                  if sno == sp_sno and pno == sp_pno\n        ]\n\n        self.assertEqual(x, [('Boeing', 'Airliner'), ('Boeing', 'Engine'), ('Ford', 'Engine'),\n                             ('Macdonalds', 'Cheeseburger')])\n\n    def testGenexps(self):\n        # generator expression tests\n        g = ([x for x in range(10)] for x in range(1))\n        self.assertEqual(next(g), [x for x in range(10)])\n        try:\n            next(g)\n            self.fail('should produce StopIteration exception')\n        except StopIteration:\n            pass\n\n        a = 1\n        try:\n            g = (a for d in a)\n            next(g)\n            self.fail('should produce TypeError')\n        except TypeError:\n            pass\n\n        self.assertEqual(list((x, y) for x in 'abcd' for y in 'abcd'), [(x, y) for x in 'abcd' for y in 'abcd'])\n        self.assertEqual(list((x, y) for x in 'ab' for y in 'xy'), [(x, y) for x in 'ab' for y in 'xy'])\n\n        a = [x for x in range(10)]\n        b = (x for x in (y for y in a))\n        self.assertEqual(sum(b), sum([x for x in range(10)]))\n\n        self.assertEqual(sum(x**2 for x in range(10)), sum([x**2 for x in range(10)]))\n        self.assertEqual(sum(x*x for x in range(10) if x%2), sum([x*x for x in range(10) if x%2]))\n        self.assertEqual(sum(x for x in (y for y in range(10))), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10)))), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in [y for y in (z for z in range(10))]), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10) if True)) if True), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10) if True) if False) if True), 0)\n        check_syntax_error(self, \"foo(x for x in range(10), 100)\")\n        check_syntax_error(self, \"foo(100, x for x in range(10))\")\n\n    def testComprehensionSpecials(self):\n        # test for outmost iterable precomputation\n        x = 10; g = (i for i in range(x)); x = 5\n        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testComprehensionSpecials","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testComprehensionSpecials#L834-L851","kind":"function","name":"testComprehensionSpecials","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":834,"end_line":851,"context_start_line":814,"context_end_line":871,"code":"        except TypeError:\n            pass\n\n        self.assertEqual(list((x, y) for x in 'abcd' for y in 'abcd'), [(x, y) for x in 'abcd' for y in 'abcd'])\n        self.assertEqual(list((x, y) for x in 'ab' for y in 'xy'), [(x, y) for x in 'ab' for y in 'xy'])\n\n        a = [x for x in range(10)]\n        b = (x for x in (y for y in a))\n        self.assertEqual(sum(b), sum([x for x in range(10)]))\n\n        self.assertEqual(sum(x**2 for x in range(10)), sum([x**2 for x in range(10)]))\n        self.assertEqual(sum(x*x for x in range(10) if x%2), sum([x*x for x in range(10) if x%2]))\n        self.assertEqual(sum(x for x in (y for y in range(10))), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10)))), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in [y for y in (z for z in range(10))]), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10) if True)) if True), sum([x for x in range(10)]))\n        self.assertEqual(sum(x for x in (y for y in (z for z in range(10) if True) if False) if True), 0)\n        check_syntax_error(self, \"foo(x for x in range(10), 100)\")\n        check_syntax_error(self, \"foo(100, x for x in range(10))\")\n\n    def testComprehensionSpecials(self):\n        # test for outmost iterable precomputation\n        x = 10; g = (i for i in range(x)); x = 5\n        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n\n    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_with_statement","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_with_statement#L853-L871","kind":"function","name":"test_with_statement","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":853,"end_line":871,"context_start_line":833,"context_end_line":891,"code":"\n    def testComprehensionSpecials(self):\n        # test for outmost iterable precomputation\n        x = 10; g = (i for i in range(x)); x = 5\n        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n\n    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"\n            print(x)\n            return ret\n\n        # the next line is not allowed anymore\n        #self.assertEqual([ x() for x in lambda: True, lambda: False if x() ], [True])\n        self.assertEqual([ x() for x in (lambda: True, lambda: False) if x() ], [True])\n        self.assertEqual([ x(False) for x in (lambda x: False if x else True, lambda x: True if x else False) if x(False) ], [True])\n        self.assertEqual((5 if 1 else _checkeval(\"check 1\", 0)), 5)\n        self.assertEqual((_checkeval(\"check 2\", 0) if 0 else 5), 5)\n        self.assertEqual((5 and 6 if 0 else 1), 1)\n        self.assertEqual(((5 and 6) if 0 else 1), 1)\n        self.assertEqual((5 and (6 if 1 else 1)), 6)\n        self.assertEqual((0 or _checkeval(\"check 3\", 2) if 0 else 3), 3)\n        self.assertEqual((1 or _checkeval(\"check 4\", 2) if 1 else _checkeval(\"check 5\", 3)), 1)\n        self.assertEqual((0 or 5 if 1 else _checkeval(\"check 6\", 3)), 5)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testIfElseExpr","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testIfElseExpr#L873-L898","kind":"function","name":"testIfElseExpr","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":873,"end_line":898,"context_start_line":853,"context_end_line":918,"code":"    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"\n            print(x)\n            return ret\n\n        # the next line is not allowed anymore\n        #self.assertEqual([ x() for x in lambda: True, lambda: False if x() ], [True])\n        self.assertEqual([ x() for x in (lambda: True, lambda: False) if x() ], [True])\n        self.assertEqual([ x(False) for x in (lambda x: False if x else True, lambda x: True if x else False) if x(False) ], [True])\n        self.assertEqual((5 if 1 else _checkeval(\"check 1\", 0)), 5)\n        self.assertEqual((_checkeval(\"check 2\", 0) if 0 else 5), 5)\n        self.assertEqual((5 and 6 if 0 else 1), 1)\n        self.assertEqual(((5 and 6) if 0 else 1), 1)\n        self.assertEqual((5 and (6 if 1 else 1)), 6)\n        self.assertEqual((0 or _checkeval(\"check 3\", 2) if 0 else 3), 3)\n        self.assertEqual((1 or _checkeval(\"check 4\", 2) if 1 else _checkeval(\"check 5\", 3)), 1)\n        self.assertEqual((0 or 5 if 1 else _checkeval(\"check 6\", 3)), 5)\n        self.assertEqual((not 5 if 1 else 1), False)\n        self.assertEqual((not 5 if 0 else 1), 1)\n        self.assertEqual((6 + 1 if 1 else 2), 7)\n        self.assertEqual((6 - 1 if 1 else 2), 5)\n        self.assertEqual((6 * 2 if 1 else 4), 12)\n        self.assertEqual((6 / 2 if 1 else 3), 3)\n        self.assertEqual((6 < 4 if 0 else 2), 2)\n\n    def testStringLiterals(self):\n        x = ''; y = \"\"; self.assert_(len(x) == 0 and x == y)\n        x = '\\''; y = \"'\"; self.assert_(len(x) == 1 and x == y and ord(x) == 39)\n        x = '\"'; y = \"\\\"\"; self.assert_(len(x) == 1 and x == y and ord(x) == 34)\n        x = \"doesn't \\\"shrink\\\" does it\"\n        y = 'doesn\\'t \"shrink\" does it'\n        self.assert_(len(x) == 24 and x == y)\n        x = \"does \\\"shrink\\\" doesn't it\"\n        y = 'does \"shrink\" doesn\\'t it'\n        self.assert_(len(x) == 24 and x == y)\n        x = \"\"\"\nThe \"quick\"\nbrown fox\njumps over\nthe 'lazy' dog.\n\"\"\"\n        y = '\\nThe \"quick\"\\nbrown fox\\njumps over\\nthe \\'lazy\\' dog.\\n'\n        self.assertEquals(x, y)\n        y = '''","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testStringLiterals","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.testStringLiterals#L900-L938","kind":"function","name":"testStringLiterals","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":900,"end_line":938,"context_start_line":880,"context_end_line":945,"code":"        # the next line is not allowed anymore\n        #self.assertEqual([ x() for x in lambda: True, lambda: False if x() ], [True])\n        self.assertEqual([ x() for x in (lambda: True, lambda: False) if x() ], [True])\n        self.assertEqual([ x(False) for x in (lambda x: False if x else True, lambda x: True if x else False) if x(False) ], [True])\n        self.assertEqual((5 if 1 else _checkeval(\"check 1\", 0)), 5)\n        self.assertEqual((_checkeval(\"check 2\", 0) if 0 else 5), 5)\n        self.assertEqual((5 and 6 if 0 else 1), 1)\n        self.assertEqual(((5 and 6) if 0 else 1), 1)\n        self.assertEqual((5 and (6 if 1 else 1)), 6)\n        self.assertEqual((0 or _checkeval(\"check 3\", 2) if 0 else 3), 3)\n        self.assertEqual((1 or _checkeval(\"check 4\", 2) if 1 else _checkeval(\"check 5\", 3)), 1)\n        self.assertEqual((0 or 5 if 1 else _checkeval(\"check 6\", 3)), 5)\n        self.assertEqual((not 5 if 1 else 1), False)\n        self.assertEqual((not 5 if 0 else 1), 1)\n        self.assertEqual((6 + 1 if 1 else 2), 7)\n        self.assertEqual((6 - 1 if 1 else 2), 5)\n        self.assertEqual((6 * 2 if 1 else 4), 12)\n        self.assertEqual((6 / 2 if 1 else 3), 3)\n        self.assertEqual((6 < 4 if 0 else 2), 2)\n\n    def testStringLiterals(self):\n        x = ''; y = \"\"; self.assert_(len(x) == 0 and x == y)\n        x = '\\''; y = \"'\"; self.assert_(len(x) == 1 and x == y and ord(x) == 39)\n        x = '\"'; y = \"\\\"\"; self.assert_(len(x) == 1 and x == y and ord(x) == 34)\n        x = \"doesn't \\\"shrink\\\" does it\"\n        y = 'doesn\\'t \"shrink\" does it'\n        self.assert_(len(x) == 24 and x == y)\n        x = \"does \\\"shrink\\\" doesn't it\"\n        y = 'does \"shrink\" doesn\\'t it'\n        self.assert_(len(x) == 24 and x == y)\n        x = \"\"\"\nThe \"quick\"\nbrown fox\njumps over\nthe 'lazy' dog.\n\"\"\"\n        y = '\\nThe \"quick\"\\nbrown fox\\njumps over\\nthe \\'lazy\\' dog.\\n'\n        self.assertEquals(x, y)\n        y = '''\nThe \"quick\"\nbrown fox\njumps over\nthe 'lazy' dog.\n'''\n        self.assertEquals(x, y)\n        y = \"\\n\\\nThe \\\"quick\\\"\\n\\\nbrown fox\\n\\\njumps over\\n\\\nthe 'lazy' dog.\\n\\\n\"\n        self.assertEquals(x, y)\n        y = '\\n\\\nThe \\\"quick\\\"\\n\\\nbrown fox\\n\\\njumps over\\n\\\nthe \\'lazy\\' dog.\\n\\\n'\n        self.assertEquals(x, y)\n\n\ndef test_main():\n    run_unittest(TokenTests, GrammarTests)\n\nif __name__ == '__main__':\n    test_main()","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.f1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.f1#L144-L144","kind":"function","name":"f1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":144,"end_line":144,"context_start_line":124,"context_end_line":164,"code":"    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.f2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.f2#L148-L148","kind":"function","name":"f2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":148,"end_line":148,"context_start_line":128,"context_end_line":168,"code":"        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.f3","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.f3#L149-L149","kind":"function","name":"f3","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":149,"end_line":149,"context_start_line":129,"context_end_line":169,"code":"        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.a1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.a1#L152-L152","kind":"function","name":"a1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":152,"end_line":152,"context_start_line":132,"context_end_line":172,"code":"        ### [decorators] 'def' NAME parameters ['->' test] ':' suite\n        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.a2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.a2#L153-L153","kind":"function","name":"a2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":153,"end_line":153,"context_start_line":133,"context_end_line":173,"code":"        ### decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.v0","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.v0#L154-L154","kind":"function","name":"v0","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":154,"end_line":154,"context_start_line":134,"context_end_line":174,"code":"        ### decorators: decorator+\n        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.v1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.v1#L155-L155","kind":"function","name":"v1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":155,"end_line":155,"context_start_line":135,"context_end_line":175,"code":"        ### parameters: '(' [typedargslist] ')'\n        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.v2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.v2#L156-L156","kind":"function","name":"v2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":156,"end_line":156,"context_start_line":136,"context_end_line":176,"code":"        ### typedargslist: ((tfpdef ['=' test] ',')*\n        ###                ('*' [tfpdef] (',' tfpdef ['=' test])* [',' '**' tfpdef] | '**' tfpdef)\n        ###                | tfpdef ['=' test] (',' tfpdef ['=' test])* [','])\n        ### tfpdef: NAME [':' test]\n        ### varargslist: ((vfpdef ['=' test] ',')*\n        ###              ('*' [vfpdef] (',' vfpdef ['=' test])*  [',' '**' vfpdef] | '**' vfpdef)\n        ###              | vfpdef ['=' test] (',' vfpdef ['=' test])* [','])\n        ### vfpdef: NAME\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        self.assertEquals(f2.__code__.co_varnames, ('one_argument',))\n        self.assertEquals(f3.__code__.co_varnames, ('two', 'arguments'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d01","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d01#L178-L178","kind":"function","name":"d01","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":178,"end_line":178,"context_start_line":158,"context_end_line":198,"code":"        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d11","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d11#L183-L183","kind":"function","name":"d11","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":183,"end_line":183,"context_start_line":163,"context_end_line":203,"code":"        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d21","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d21#L187-L187","kind":"function","name":"d21","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":187,"end_line":187,"context_start_line":167,"context_end_line":207,"code":"        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d02","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d02#L194-L194","kind":"function","name":"d02","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":194,"end_line":194,"context_start_line":174,"context_end_line":214,"code":"        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d12","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d12#L202-L202","kind":"function","name":"d12","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":202,"end_line":202,"context_start_line":182,"context_end_line":222,"code":"        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d22","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d22#L206-L206","kind":"function","name":"d22","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":206,"end_line":206,"context_start_line":186,"context_end_line":226,"code":"        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d01v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d01v#L210-L210","kind":"function","name":"d01v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":210,"end_line":210,"context_start_line":190,"context_end_line":230,"code":"        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d11v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d11v#L217-L217","kind":"function","name":"d11v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":217,"end_line":217,"context_start_line":197,"context_end_line":237,"code":"        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d21v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d21v#L221-L221","kind":"function","name":"d21v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":221,"end_line":221,"context_start_line":201,"context_end_line":241,"code":"        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d02v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d02v#L227-L227","kind":"function","name":"d02v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":227,"end_line":227,"context_start_line":207,"context_end_line":247,"code":"        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d12v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d12v#L234-L234","kind":"function","name":"d12v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":234,"end_line":234,"context_start_line":214,"context_end_line":254,"code":"        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d22v","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.d22v#L242-L242","kind":"function","name":"d22v","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":242,"end_line":242,"context_start_line":222,"context_end_line":262,"code":"        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.pos0key1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.pos0key1#L259-L259","kind":"function","name":"pos0key1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":259,"end_line":259,"context_start_line":239,"context_end_line":279,"code":"        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.pos2key2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.pos2key2#L261-L261","kind":"function","name":"pos2key2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":261,"end_line":261,"context_start_line":241,"context_end_line":281,"code":"        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.pos2key2dict","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.pos2key2dict#L265-L265","kind":"function","name":"pos2key2dict","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":265,"end_line":265,"context_start_line":245,"context_end_line":285,"code":"        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n\n        # keyword argument type tests\n        try:\n            str('x', **{b'foo':1 })\n        except TypeError:\n            pass\n        else:\n            self.fail('Bytes should not work as keyword argument names')\n        # keyword only argument tests\n        def pos0key1(*, key): return key\n        pos0key1(key=100)\n        def pos2key2(p1, p2, *, k1, k2=100): return p1,p2,k1,k2\n        pos2key2(1, 2, k1=100)\n        pos2key2(1, 2, k1=100, k2=200)\n        pos2key2(1, 2, k2=100, k1=200)\n        def pos2key2dict(p1, p2, *, k1=100, k2, **kwarg): return p1,p2,k1,k2,kwarg\n        pos2key2dict(1,2,k2=100,tokwarg1=100,tokwarg2=200)\n        pos2key2dict(1,2,tokwarg1=100,tokwarg2=200, k2=100)\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # argument annotation tests\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})\n        def f(*x:str): pass\n        self.assertEquals(f.__annotations__, {'x': str})\n        def f(**x:float): pass\n        self.assertEquals(f.__annotations__, {'x': float})","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.f","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.f#L474-L476","kind":"function","name":"f","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":474,"end_line":476,"context_start_line":454,"context_end_line":496,"code":"        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testNonlocal(self):\n        # 'nonlocal' NAME (',' NAME)*\n        x = 0\n        y = 0\n        def f():\n            nonlocal x\n            nonlocal x, y\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError as e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.null","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.null#L299-L299","kind":"function","name":"null","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":299,"end_line":299,"context_start_line":279,"context_end_line":319,"code":"        self.assertEquals(f.__annotations__, {'return': list})\n        def f(x:int): pass\n        self.assertEquals(f.__annotations__, {'x': int})\n        def f(*x:str): pass\n        self.assertEquals(f.__annotations__, {'x': str})\n        def f(**x:float): pass\n        self.assertEquals(f.__annotations__, {'x': float})\n        def f(x, y:1+2): pass\n        self.assertEquals(f.__annotations__, {'y': 3})\n        def f(a, b:1, c:2, d): pass\n        self.assertEquals(f.__annotations__, {'b': 1, 'c': 2})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6): pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6})\n        def f(a, b:1, c:2, d, e:3=4, f=5, *g:6, h:7, i=8, j:9=10,\n              **k:11) -> 12: pass\n        self.assertEquals(f.__annotations__,\n                          {'b': 1, 'c': 2, 'e': 3, 'g': 6, 'h': 7, 'j': 9,\n                           'k': 11, 'return': 12})\n        # Check for SF Bug #1697248 - mixing decorators and a return annotation\n        def null(x): return x\n        @null\n        def f(x) -> list: pass\n        self.assertEquals(f.__annotations__, {'return': list})\n\n        # test closures with a variety of oparg's\n        closure = 1\n        def f(): return closure\n        def f(x=1): return closure\n        def f(*, k=1): return closure\n        def f() -> int: return closure\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.foo","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.foo#L340-L342","kind":"function","name":"foo","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":340,"end_line":342,"context_start_line":320,"context_end_line":362,"code":"        l3 = lambda : [2 < x for x in [-1, 3, 0]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n        l6 = lambda x, y, *, k=20: x+y+k\n        self.assertEquals(l6(1,2), 1+2+20)\n        self.assertEquals(l6(1,2,k=10), 1+2+10)\n\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testDelStmt(self):\n        # 'del' exprlist","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_inner","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_inner#L417-L430","kind":"function","name":"test_inner","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":417,"end_line":430,"context_start_line":397,"context_end_line":450,"code":"        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.g1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.g1#L435-L435","kind":"function","name":"g1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":435,"end_line":435,"context_start_line":415,"context_end_line":455,"code":"        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.g2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.g2#L436-L436","kind":"function","name":"g2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":436,"end_line":436,"context_start_line":416,"context_end_line":456,"code":"\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo != 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError('just testing')\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.Squares","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.Squares#L528-L539","kind":"class","name":"Squares","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":528,"end_line":539,"context_start_line":508,"context_end_line":559,"code":"    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.B","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.B#L712-L712","kind":"class","name":"B","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":712,"end_line":712,"context_start_line":692,"context_end_line":732,"code":"        x = {'one' or 'two': 1 or 2}\n        x = {'one': 1, 'two': 2}\n        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.B2","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.B2#L713-L713","kind":"class","name":"B2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":713,"end_line":713,"context_start_line":693,"context_end_line":733,"code":"        x = {'one': 1, 'two': 2}\n        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.C1","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.C1#L714-L714","kind":"class","name":"C1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":714,"end_line":714,"context_start_line":694,"context_end_line":734,"code":"        x = {'one': 1, 'two': 2,}\n        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.C2","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.C2#L715-L715","kind":"class","name":"C2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":715,"end_line":715,"context_start_line":695,"context_end_line":735,"code":"        x = {'one': 1, 'two': 2, 'three': 3, 'four': 4, 'five': 5, 'six': 6}\n\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.D","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.D#L716-L716","kind":"class","name":"D","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":716,"end_line":716,"context_start_line":696,"context_end_line":736,"code":"\n        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.C","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.C#L717-L720","kind":"class","name":"C","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":717,"end_line":720,"context_start_line":697,"context_end_line":740,"code":"        x = {'one'}\n        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.class_decorator","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.class_decorator#L725-L725","kind":"function","name":"class_decorator","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":725,"end_line":725,"context_start_line":705,"context_end_line":745,"code":"\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.G","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.G#L727-L727","kind":"class","name":"G","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":727,"end_line":727,"context_start_line":707,"context_end_line":747,"code":"    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_in_func","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_in_func#L758-L759","kind":"function","name":"test_in_func","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":758,"end_line":759,"context_start_line":738,"context_end_line":779,"code":"        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]\n\n        self.assertEqual([s.strip() for s in spcs], ['Apple', 'Banana', 'Coco  nut'])\n        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Apple'), (3, 'Banana'), (3, 'Coconut'),\n                          (4, 'Apple'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Apple'), (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(i, s) for i in nums for s in [f for f in strs if \"n\" in f]],\n                         [(1, 'Banana'), (1, 'Coconut'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Banana'), (3, 'Coconut'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(lambda a:[a**i for i in range(a+1)])(j) for j in range(5)],\n                         [[1], [1, 1], [1, 2, 4], [1, 3, 9, 27], [1, 4, 16, 64, 256]])\n\n        def test_in_func(l):\n            return [0 < x < 3 for x in l if x > 2]\n\n        self.assertEqual(test_in_func(nums), [False, False, False])\n\n        def test_nested_front():\n            self.assertEqual([[y for y in [x, x + 1]] for x in [1,3,5]],\n                             [[1, 2], [3, 4], [5, 6]])\n\n        test_nested_front()\n\n        check_syntax_error(self, \"[i, s for i in nums for s in strs]\")\n        check_syntax_error(self, \"[x if y]\")\n\n        suppliers = [\n          (1, \"Boeing\"),\n          (2, \"Ford\"),\n          (3, \"Macdonalds\")\n        ]\n\n        parts = [\n          (10, \"Airliner\"),","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_nested_front","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.test_nested_front#L763-L765","kind":"function","name":"test_nested_front","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":763,"end_line":765,"context_start_line":743,"context_end_line":785,"code":"        self.assertEqual([3 * x for x in nums], [3, 6, 9, 12, 15])\n        self.assertEqual([x for x in nums if x > 2], [3, 4, 5])\n        self.assertEqual([(i, s) for i in nums for s in strs],\n                         [(1, 'Apple'), (1, 'Banana'), (1, 'Coconut'),\n                          (2, 'Apple'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Apple'), (3, 'Banana'), (3, 'Coconut'),\n                          (4, 'Apple'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Apple'), (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(i, s) for i in nums for s in [f for f in strs if \"n\" in f]],\n                         [(1, 'Banana'), (1, 'Coconut'), (2, 'Banana'), (2, 'Coconut'),\n                          (3, 'Banana'), (3, 'Coconut'), (4, 'Banana'), (4, 'Coconut'),\n                          (5, 'Banana'), (5, 'Coconut')])\n        self.assertEqual([(lambda a:[a**i for i in range(a+1)])(j) for j in range(5)],\n                         [[1], [1, 1], [1, 2, 4], [1, 3, 9, 27], [1, 4, 16, 64, 256]])\n\n        def test_in_func(l):\n            return [0 < x < 3 for x in l if x > 2]\n\n        self.assertEqual(test_in_func(nums), [False, False, False])\n\n        def test_nested_front():\n            self.assertEqual([[y for y in [x, x + 1]] for x in [1,3,5]],\n                             [[1, 2], [3, 4], [5, 6]])\n\n        test_nested_front()\n\n        check_syntax_error(self, \"[i, s for i in nums for s in strs]\")\n        check_syntax_error(self, \"[x if y]\")\n\n        suppliers = [\n          (1, \"Boeing\"),\n          (2, \"Ford\"),\n          (3, \"Macdonalds\")\n        ]\n\n        parts = [\n          (10, \"Airliner\"),\n          (20, \"Engine\"),\n          (30, \"Cheeseburger\")\n        ]\n\n        suppart = [\n          (1, 10), (1, 20), (2, 20), (3, 30)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.manager","uri":"program://LLaMA-Adapter/class/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.manager#L854-L858","kind":"class","name":"manager","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":854,"end_line":858,"context_start_line":834,"context_end_line":878,"code":"    def testComprehensionSpecials(self):\n        # test for outmost iterable precomputation\n        x = 10; g = (i for i in range(x)); x = 5\n        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n\n    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"\n            print(x)\n            return ret","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf._checkeval","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf._checkeval#L875-L878","kind":"function","name":"_checkeval","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":875,"end_line":878,"context_start_line":855,"context_end_line":898,"code":"            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"\n            print(x)\n            return ret\n\n        # the next line is not allowed anymore\n        #self.assertEqual([ x() for x in lambda: True, lambda: False if x() ], [True])\n        self.assertEqual([ x() for x in (lambda: True, lambda: False) if x() ], [True])\n        self.assertEqual([ x(False) for x in (lambda x: False if x else True, lambda x: True if x else False) if x(False) ], [True])\n        self.assertEqual((5 if 1 else _checkeval(\"check 1\", 0)), 5)\n        self.assertEqual((_checkeval(\"check 2\", 0) if 0 else 5), 5)\n        self.assertEqual((5 and 6 if 0 else 1), 1)\n        self.assertEqual(((5 and 6) if 0 else 1), 1)\n        self.assertEqual((5 and (6 if 1 else 1)), 6)\n        self.assertEqual((0 or _checkeval(\"check 3\", 2) if 0 else 3), 3)\n        self.assertEqual((1 or _checkeval(\"check 4\", 2) if 1 else _checkeval(\"check 5\", 3)), 1)\n        self.assertEqual((0 or 5 if 1 else _checkeval(\"check 6\", 3)), 5)\n        self.assertEqual((not 5 if 1 else 1), False)\n        self.assertEqual((not 5 if 0 else 1), 1)\n        self.assertEqual((6 + 1 if 1 else 2), 7)\n        self.assertEqual((6 - 1 if 1 else 2), 5)\n        self.assertEqual((6 * 2 if 1 else 4), 12)\n        self.assertEqual((6 / 2 if 1 else 3), 3)\n        self.assertEqual((6 < 4 if 0 else 2), 2)","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.__init__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.__init__#L529-L531","kind":"function","name":"__init__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":529,"end_line":531,"context_start_line":509,"context_end_line":551,"code":"        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.__len__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.__len__#L532-L532","kind":"function","name":"__len__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":532,"end_line":532,"context_start_line":512,"context_end_line":552,"code":"        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.__getitem__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.__getitem__#L533-L539","kind":"function","name":"__getitem__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":533,"end_line":539,"context_start_line":513,"context_end_line":559,"code":"\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr ['as' expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.meth1","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.meth1#L718-L718","kind":"function","name":"meth1","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":718,"end_line":718,"context_start_line":698,"context_end_line":738,"code":"        x = {'one', 1,}\n        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.meth2","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.meth2#L719-L719","kind":"function","name":"meth2","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":719,"end_line":719,"context_start_line":699,"context_end_line":739,"code":"        x = {'one', 'two', 'three'}\n        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.meth3","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.meth3#L720-L720","kind":"function","name":"meth3","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":720,"end_line":720,"context_start_line":700,"context_end_line":740,"code":"        x = {2, 3, 4,}\n\n        x = x\n        x = 'x'\n        x = 123\n\n    ### exprlist: expr (',' expr)* [',']\n    ### testlist: test (',' test)* [',']\n    # These have been exercised enough above\n\n    def testClassdef(self):\n        # 'class' NAME ['(' [testlist] ')'] ':' suite\n        class B: pass\n        class B2(): pass\n        class C1(B): pass\n        class C2(B): pass\n        class D(C1, C2, B): pass\n        class C:\n            def meth1(self): pass\n            def meth2(self, arg): pass\n            def meth3(self, a1, a2): pass\n\n        # decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE\n        # decorators: decorator+\n        # decorated: decorators (classdef | funcdef)\n        def class_decorator(x): return x\n        @class_decorator\n        class G: pass\n\n    def testDictcomps(self):\n        # dictorsetmaker: ( (test ':' test (comp_for |\n        #                                   (',' test ':' test)* [','])) |\n        #                   (test (comp_for | (',' test)* [','])) )\n        nums = [1, 2, 3]\n        self.assertEqual({i:i+1 for i in nums}, {1: 2, 2: 3, 3: 4})\n\n    def testListcomps(self):\n        # list comprehension tests\n        nums = [1, 2, 3, 4, 5]\n        strs = [\"Apple\", \"Banana\", \"Coconut\"]\n        spcs = [\"  Apple\", \" Banana \", \"Coco  nut  \"]","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.__enter__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.__enter__#L855-L856","kind":"function","name":"__enter__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":855,"end_line":856,"context_start_line":835,"context_end_line":876,"code":"        # test for outmost iterable precomputation\n        x = 10; g = (i for i in range(x)); x = 5\n        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n\n    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.__exit__","uri":"program://LLaMA-Adapter/function/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python3-grammar-crlf.__exit__#L857-L858","kind":"function","name":"__exit__","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":857,"end_line":858,"context_start_line":837,"context_end_line":878,"code":"        self.assertEqual(len(list(g)), 10)\n\n        # This should hold, since we're only precomputing outmost iterable.\n        x = 10; t = False; g = ((i,j) for i in range(x) if t for j in range(x))\n        x = 5; t = True;\n        self.assertEqual([(i,j) for i in range(10) for j in range(5)], list(g))\n\n        # Grammar allows multiple adjacent 'if's in listcomps and genexps,\n        # even though it's silly. Make sure it works (ifelse broke this.)\n        self.assertEqual([ x for x in range(10) if x % 2 if x % 3 ], [1, 5, 7])\n        self.assertEqual(list(x for x in range(10) if x % 2 if x % 3), [1, 5, 7])\n\n        # verify unpacking single element tuples in listcomp/genexp.\n        self.assertEqual([x for x, in [(4,), (5,), (6,)]], [4, 5, 6])\n        self.assertEqual(list(x for x, in [(7,), (8,), (9,)]), [7, 8, 9])\n\n    def test_with_statement(self):\n        class manager(object):\n            def __enter__(self):\n                return (1, 2)\n            def __exit__(self, *args):\n                pass\n\n        with manager():\n            pass\n        with manager() as x:\n            pass\n        with manager() as (x, y):\n            pass\n        with manager(), manager():\n            pass\n        with manager() as x, manager() as y:\n            pass\n        with manager() as x, manager():\n            pass\n\n    def testIfElseExpr(self):\n        # Test ifelse expressions in various cases\n        def _checkeval(msg, ret):\n            \"helper to check that evaluation of expressions is done correctly\"\n            print(x)\n            return ret","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python2-grammar","uri":"program://LLaMA-Adapter/module/gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python2-grammar#L1-L975","kind":"module","name":"gorilla.gorilla-main.eval.eval-scripts.codebleu.parser.tree-sitter-python.examples.python2-grammar","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python2-grammar.py","language":"python","start_line":1,"end_line":975,"context_start_line":1,"context_end_line":975,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.test_support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\\n        + 1\n        self.assertEquals(x, 2, 'backslash for line continuation')\n\n        # Backslash does not means continuation in comments :\\\n        x = 0\n        self.assertEquals(x, 0, 'backslash ending comment')\n\n    def testPlainIntegers(self):\n        self.assertEquals(0xff, 255)\n        self.assertEquals(0377, 255)\n        self.assertEquals(2147483647, 017777777777)\n        # \"0x\" is not a valid literal\n        self.assertRaises(SyntaxError, eval, \"0x\")\n        from sys import maxint\n        if maxint == 2147483647:\n            self.assertEquals(-2147483647-1, -020000000000)\n            # XXX -2147483648\n            self.assert_(037777777777 > 0)\n            self.assert_(0xffffffff > 0)\n            for s in '2147483648', '040000000000', '0x100000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        elif maxint == 9223372036854775807:\n            self.assertEquals(-9223372036854775807-1, -01000000000000000000000)\n            self.assert_(01777777777777777777777 > 0)\n            self.assert_(0xffffffffffffffff > 0)\n            for s in '9223372036854775808', '02000000000000000000000', \\\n                     '0x10000000000000000':\n                try:\n                    x = eval(s)\n                except OverflowError:\n                    self.fail(\"OverflowError on huge integer literal %r\" % s)\n        else:\n            self.fail('Weird maxint value %r' % maxint)\n\n    def testLongIntegers(self):\n        x = 0L\n        x = 0l\n        x = 0xffffffffffffffffL\n        x = 0xffffffffffffffffl\n        x = 077777777777777777L\n        x = 077777777777777777l\n        x = 123456789012345678901234567890L\n        x = 123456789012345678901234567890l\n\n    def testFloats(self):\n        x = 3.14\n        x = 314.\n        x = 0.314\n        # XXX x = 000.314\n        x = .314\n        x = 3e14\n        x = 3E14\n        x = 3e-14\n        x = 3e+14\n        x = 3.e14\n        x = .3e14\n        x = 3.1e4\n\nclass GrammarTests(unittest.TestCase):\n\n    # single_input: NEWLINE | simple_stmt | compound_stmt NEWLINE\n    # XXX can't test in a script -- this rule is only used when interactive\n\n    # file_input: (NEWLINE | stmt)* ENDMARKER\n    # Being tested as this very moment this very module\n\n    # expr_input: testlist NEWLINE\n    # XXX Hard to test -- used only in calls to input()\n\n    def testEvalInput(self):\n        # testlist ENDMARKER\n        x = eval('1, 0 or 1')\n\n    def testFuncdef(self):\n        ### 'def' NAME parameters ':' suite\n        ### parameters: '(' [varargslist] ')'\n        ### varargslist: (fpdef ['=' test] ',')* ('*' NAME [',' ('**'|'*' '*') NAME]\n        ###            | ('**'|'*' '*') NAME)\n        ###            | fpdef ['=' test] (',' fpdef ['=' test])* [',']\n        ### fpdef: NAME | '(' fplist ')'\n        ### fplist: fpdef (',' fpdef)* [',']\n        ### arglist: (argument ',')* (argument | *' test [',' '**' test] | '**' test)\n        ### argument: [test '='] test   # Really [keyword '='] test\n        def f1(): pass\n        f1()\n        f1(*())\n        f1(*(), **{})\n        def f2(one_argument): pass\n        def f3(two, arguments): pass\n        def f4(two, (compound, (argument, list))): pass\n        def f5((compound, first), two): pass\n        self.assertEquals(f2.func_code.co_varnames, ('one_argument',))\n        self.assertEquals(f3.func_code.co_varnames, ('two', 'arguments'))\n        if sys.platform.startswith('java'):\n            self.assertEquals(f4.func_code.co_varnames,\n                   ('two', '(compound, (argument, list))', 'compound', 'argument',\n                                'list',))\n            self.assertEquals(f5.func_code.co_varnames,\n                   ('(compound, first)', 'two', 'compound', 'first'))\n        else:\n            self.assertEquals(f4.func_code.co_varnames,\n                  ('two', '.1', 'compound', 'argument',  'list'))\n            self.assertEquals(f5.func_code.co_varnames,\n                  ('.0', 'two', 'compound', 'first'))\n        def a1(one_arg,): pass\n        def a2(two, args,): pass\n        def v0(*rest): pass\n        def v1(a, *rest): pass\n        def v2(a, b, *rest): pass\n        def v3(a, (b, c), *rest): return a, b, c, rest\n\n        f1()\n        f2(1)\n        f2(1,)\n        f3(1, 2)\n        f3(1, 2,)\n        f4(1, (2, (3, 4)))\n        v0()\n        v0(1)\n        v0(1,)\n        v0(1,2)\n        v0(1,2,3,4,5,6,7,8,9,0)\n        v1(1)\n        v1(1,)\n        v1(1,2)\n        v1(1,2,3)\n        v1(1,2,3,4,5,6,7,8,9,0)\n        v2(1,2)\n        v2(1,2,3)\n        v2(1,2,3,4)\n        v2(1,2,3,4,5,6,7,8,9,0)\n        v3(1,(2,3))\n        v3(1,(2,3),4)\n        v3(1,(2,3),4,5,6,7,8,9,0)\n\n        # ceval unpacks the formal arguments into the first argcount names;\n        # thus, the names nested inside tuples must appear after these names.\n        if sys.platform.startswith('java'):\n            self.assertEquals(v3.func_code.co_varnames, ('a', '(b, c)', 'rest', 'b', 'c'))\n        else:\n            self.assertEquals(v3.func_code.co_varnames, ('a', '.1', 'rest', 'b', 'c'))\n        self.assertEquals(v3(1, (2, 3), 4), (1, 2, 3, (4,)))\n        def d01(a=1): pass\n        d01()\n        d01(1)\n        d01(*(1,))\n        d01(**{'a':2})\n        def d11(a, b=1): pass\n        d11(1)\n        d11(1, 2)\n        d11(1, **{'b':2})\n        def d21(a, b, c=1): pass\n        d21(1, 2)\n        d21(1, 2, 3)\n        d21(*(1, 2, 3))\n        d21(1, *(2, 3))\n        d21(1, 2, *(3,))\n        d21(1, 2, **{'c':3})\n        def d02(a=1, b=2): pass\n        d02()\n        d02(1)\n        d02(1, 2)\n        d02(*(1, 2))\n        d02(1, *(2,))\n        d02(1, **{'b':2})\n        d02(**{'a': 1, 'b': 2})\n        def d12(a, b=1, c=2): pass\n        d12(1)\n        d12(1, 2)\n        d12(1, 2, 3)\n        def d22(a, b, c=1, d=2): pass\n        d22(1, 2)\n        d22(1, 2, 3)\n        d22(1, 2, 3, 4)\n        def d01v(a=1, *rest): pass\n        d01v()\n        d01v(1)\n        d01v(1, 2)\n        d01v(*(1, 2, 3, 4))\n        d01v(*(1,))\n        d01v(**{'a':2})\n        def d11v(a, b=1, *rest): pass\n        d11v(1)\n        d11v(1, 2)\n        d11v(1, 2, 3)\n        def d21v(a, b, c=1, *rest): pass\n        d21v(1, 2)\n        d21v(1, 2, 3)\n        d21v(1, 2, 3, 4)\n        d21v(*(1, 2, 3, 4))\n        d21v(1, 2, **{'c': 3})\n        def d02v(a=1, b=2, *rest): pass\n        d02v()\n        d02v(1)\n        d02v(1, 2)\n        d02v(1, 2, 3)\n        d02v(1, *(2, 3, 4))\n        d02v(**{'a': 1, 'b': 2})\n        def d12v(a, b=1, c=2, *rest): pass\n        d12v(1)\n        d12v(1, 2)\n        d12v(1, 2, 3)\n        d12v(1, 2, 3, 4)\n        d12v(*(1, 2, 3, 4))\n        d12v(1, 2, *(3, 4, 5))\n        d12v(1, *(2,), **{'c': 3})\n        def d22v(a, b, c=1, d=2, *rest): pass\n        d22v(1, 2)\n        d22v(1, 2, 3)\n        d22v(1, 2, 3, 4)\n        d22v(1, 2, 3, 4, 5)\n        d22v(*(1, 2, 3, 4))\n        d22v(1, 2, *(3, 4, 5))\n        d22v(1, *(2, 3), **{'d': 4})\n        def d31v((x)): pass\n        d31v(1)\n        def d32v((x,)): pass\n        d32v((1,))\n\n        # keyword arguments after *arglist\n        def f(*args, **kwargs):\n            return args, kwargs\n        self.assertEquals(f(1, x=2, *[3, 4], y=5), ((1, 3, 4),\n                                                    {'x':2, 'y':5}))\n        self.assertRaises(SyntaxError, eval, \"f(1, *(2,3), 4)\")\n        self.assertRaises(SyntaxError, eval, \"f(1, x=2, *(3,4), x=5)\")\n\n        # Check ast errors in *args and *kwargs\n        check_syntax_error(self, \"f(*g(1=2))\")\n        check_syntax_error(self, \"f(**g(1=2))\")\n\n    def testLambdef(self):\n        ### lambdef: 'lambda' [varargslist] ':' test\n        l1 = lambda : 0\n        self.assertEquals(l1(), 0)\n        l2 = lambda : a[d] # XXX just testing the expression\n        l3 = lambda : [2 < x for x in [-1, 3, 0L]]\n        self.assertEquals(l3(), [0, 1, 0])\n        l4 = lambda x = lambda y = lambda z=1 : z : y() : x()\n        self.assertEquals(l4(), 1)\n        l5 = lambda x, y, z=2: x + y + z\n        self.assertEquals(l5(1, 2), 5)\n        self.assertEquals(l5(1, 2, 3), 6)\n        check_syntax_error(self, \"lambda x: x = 2\")\n        check_syntax_error(self, \"lambda (None,): None\")\n\n    ### stmt: simple_stmt | compound_stmt\n    # Tested below\n\n    def testSimpleStmt(self):\n        ### simple_stmt: small_stmt (';' small_stmt)* [';']\n        x = 1; pass; del x\n        def foo():\n            # verify statements that end with semi-colons\n            x = 1; pass; del x;\n        foo()\n\n    ### small_stmt: expr_stmt | print_stmt  | pass_stmt | del_stmt | flow_stmt | import_stmt | global_stmt | access_stmt | exec_stmt\n    # Tested below\n\n    def testExprStmt(self):\n        # (exprlist '=')* exprlist\n        1\n        1, 2, 3\n        x = 1\n        x = 1, 2, 3\n        x = y = z = 1, 2, 3\n        x, y, z = 1, 2, 3\n        abc = a, b, c = x, y, z = xyz = 1, 2, (3, 4)\n\n        check_syntax_error(self, \"x + 1 = 1\")\n        check_syntax_error(self, \"a + 1 = b + 2\")\n\n    def testPrintStmt(self):\n        # 'print' (test ',')* [test]\n        import StringIO\n\n        # Can't test printing to real stdout without comparing output\n        # which is not available in unittest.\n        save_stdout = sys.stdout\n        sys.stdout = StringIO.StringIO()\n\n        print 1, 2, 3\n        print 1, 2, 3,\n        print\n        print 0 or 1, 0 or 1,\n        print 0 or 1\n\n        # 'print' '>>' test ','\n        print >> sys.stdout, 1, 2, 3\n        print >> sys.stdout, 1, 2, 3,\n        print >> sys.stdout\n        print >> sys.stdout, 0 or 1, 0 or 1,\n        print >> sys.stdout, 0 or 1\n\n        # test printing to an instance\n        class Gulp:\n            def write(self, msg): pass\n\n        gulp = Gulp()\n        print >> gulp, 1, 2, 3\n        print >> gulp, 1, 2, 3,\n        print >> gulp\n        print >> gulp, 0 or 1, 0 or 1,\n        print >> gulp, 0 or 1\n\n        # test print >> None\n        def driver():\n            oldstdout = sys.stdout\n            sys.stdout = Gulp()\n            try:\n                tellme(Gulp())\n                tellme()\n            finally:\n                sys.stdout = oldstdout\n\n        # we should see this once\n        def tellme(file=sys.stdout):\n            print >> file, 'hello world'\n\n        driver()\n\n        # we should not see this at all\n        def tellme(file=None):\n            print >> file, 'goodbye universe'\n\n        driver()\n\n        self.assertEqual(sys.stdout.getvalue(), '''\\\n1 2 3\n1 2 3\n1 1 1\n1 2 3\n1 2 3\n1 1 1\nhello world\n''')\n        sys.stdout = save_stdout\n\n        # syntax errors\n        check_syntax_error(self, 'print ,')\n        check_syntax_error(self, 'print >> x,')\n\n    def testDelStmt(self):\n        # 'del' exprlist\n        abc = [1,2,3]\n        x, y, z = abc\n        xyz = x, y, z\n\n        del abc\n        del x, y, (z, xyz)\n\n    def testPassStmt(self):\n        # 'pass'\n        pass\n\n    # flow_stmt: break_stmt | continue_stmt | return_stmt | raise_stmt\n    # Tested below\n\n    def testBreakStmt(self):\n        # 'break'\n        while 1: break\n\n    def testContinueStmt(self):\n        # 'continue'\n        i = 1\n        while i: i = 0; continue\n\n        msg = \"\"\n        while not msg:\n            msg = \"ok\"\n            try:\n                continue\n                msg = \"continue failed to continue inside try\"\n            except:\n                msg = \"continue inside try called except block\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n        msg = \"\"\n        while not msg:\n            msg = \"finally block not called\"\n            try:\n                continue\n            finally:\n                msg = \"ok\"\n        if msg != \"ok\":\n            self.fail(msg)\n\n    def test_break_continue_loop(self):\n        # This test warrants an explanation. It is a test specifically for SF bugs\n        # #463359 and #462937. The bug is that a 'break' statement executed or\n        # exception raised inside a try/except inside a loop, *after* a continue\n        # statement has been executed in that loop, will cause the wrong number of\n        # arguments to be popped off the stack and the instruction pointer reset to\n        # a very small number (usually 0.) Because of this, the following test\n        # *must* written as a function, and the tracking vars *must* be function\n        # arguments with default values. Otherwise, the test will loop and loop.\n\n        def test_inner(extra_burning_oil = 1, count=0):\n            big_hippo = 2\n            while big_hippo:\n                count += 1\n                try:\n                    if extra_burning_oil and big_hippo == 1:\n                        extra_burning_oil -= 1\n                        break\n                    big_hippo -= 1\n                    continue\n                except:\n                    raise\n            if count > 2 or big_hippo <> 1:\n                self.fail(\"continue then break in try/except in loop broken!\")\n        test_inner()\n\n    def testReturn(self):\n        # 'return' [testlist]\n        def g1(): return\n        def g2(): return 1\n        g1()\n        x = g2()\n        check_syntax_error(self, \"class foo:return 1\")\n\n    def testYield(self):\n        check_syntax_error(self, \"class foo:yield 1\")\n\n    def testRaise(self):\n        # 'raise' test [',' test]\n        try: raise RuntimeError, 'just testing'\n        except RuntimeError: pass\n        try: raise KeyboardInterrupt\n        except KeyboardInterrupt: pass\n\n    def testImport(self):\n        # 'import' dotted_as_names\n        import sys\n        import time, sys\n        # 'from' dotted_name 'import' ('*' | '(' import_as_names ')' | import_as_names)\n        from time import time\n        from time import (time)\n        # not testable inside a function, but already done at top of the module\n        # from sys import *\n        from sys import path, argv\n        from sys import (path, argv)\n        from sys import (path, argv,)\n\n    def testGlobal(self):\n        # 'global' NAME (',' NAME)*\n        global a\n        global a, b\n        global one, two, three, four, five, six, seven, eight, nine, ten\n\n    def testExec(self):\n        # 'exec' expr ['in' expr [',' expr]]\n        z = None\n        del z\n        exec 'z=1+1\\n'\n        if z != 2: self.fail('exec \\'z=1+1\\'\\\\n')\n        del z\n        exec 'z=1+1'\n        if z != 2: self.fail('exec \\'z=1+1\\'')\n        z = None\n        del z\n        import types\n        if hasattr(types, \"UnicodeType\"):\n            exec r\"\"\"if 1:\n            exec u'z=1+1\\n'\n            if z != 2: self.fail('exec u\\'z=1+1\\'\\\\n')\n            del z\n            exec u'z=1+1'\n            if z != 2: self.fail('exec u\\'z=1+1\\'')\"\"\"\n        g = {}\n        exec 'z = 1' in g\n        if g.has_key('__builtins__'): del g['__builtins__']\n        if g != {'z': 1}: self.fail('exec \\'z = 1\\' in g')\n        g = {}\n        l = {}\n\n        import warnings\n        warnings.filterwarnings(\"ignore\", \"global statement\", module=\"<string>\")\n        exec 'global a; a = 1; b = 2' in g, l\n        if g.has_key('__builtins__'): del g['__builtins__']\n        if l.has_key('__builtins__'): del l['__builtins__']\n        if (g, l) != ({'a':1}, {'b':2}):\n            self.fail('exec ... in g (%s), l (%s)' %(g,l))\n\n    def testAssert(self):\n        # assert_stmt: 'assert' test [',' test]\n        assert 1\n        assert 1, 1\n        assert lambda x:x\n        assert 1, lambda x:x+1\n        try:\n            assert 0, \"msg\"\n        except AssertionError, e:\n            self.assertEquals(e.args[0], \"msg\")\n        else:\n            if __debug__:\n                self.fail(\"AssertionError not raised by assert 0\")\n\n    ### compound_stmt: if_stmt | while_stmt | for_stmt | try_stmt | funcdef | classdef\n    # Tested below\n\n    def testIf(self):\n        # 'if' test ':' suite ('elif' test ':' suite)* ['else' ':' suite]\n        if 1: pass\n        if 1: pass\n        else: pass\n        if 0: pass\n        elif 0: pass\n        if 0: pass\n        elif 0: pass\n        elif 0: pass\n        elif 0: pass\n        else: pass\n\n    def testWhile(self):\n        # 'while' test ':' suite ['else' ':' suite]\n        while 0: pass\n        while 0: pass\n        else: pass\n\n        # Issue1920: \"while 0\" is optimized away,\n        # ensure that the \"else\" clause is still present.\n        x = 0\n        while 0:\n            x = 1\n        else:\n            x = 2\n        self.assertEquals(x, 2)\n\n    def testFor(self):\n        # 'for' exprlist 'in' exprlist ':' suite ['else' ':' suite]\n        for i in 1, 2, 3: pass\n        for i, j, k in (): pass\n        else: pass\n        class Squares:\n            def __init__(self, max):\n                self.max = max\n                self.sofar = []\n            def __len__(self): return len(self.sofar)\n            def __getitem__(self, i):\n                if not 0 <= i < self.max: raise IndexError\n                n = len(self.sofar)\n                while n <= i:\n                    self.sofar.append(n*n)\n                    n = n+1\n                return self.sofar[i]\n        n = 0\n        for x in Squares(10): n = n+x\n        if n != 285:\n            self.fail('for over growing sequence')\n\n        result = []\n        for x, in [(1,), (2,), (3,)]:\n            result.append(x)\n        self.assertEqual(result, [1, 2, 3])\n\n    def testTry(self):\n        ### try_stmt: 'try' ':' suite (except_clause ':' suite)+ ['else' ':' suite]\n        ###         | 'try' ':' suite 'finally' ':' suite\n        ### except_clause: 'except' [expr [('as' | ',') expr]]\n        try:\n            1/0\n        except ZeroDivisionError:\n            pass\n        else:\n            pass\n        try: 1/0\n        except EOFError: pass\n        except TypeError as msg: pass\n        except RuntimeError, msg: pass\n        except: pass\n        else: pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError): pass\n        try: 1/0\n        except (EOFError, TypeError, ZeroDivisionError), msg: pass\n        try: pass\n        finally: pass\n\n    def testSuite(self):\n        # simple_stmt | NEWLINE INDENT NEWLINE* (stmt NEWLINE*)+ DEDENT\n        if 1: pass\n        if 1:\n            pass\n        if 1:\n            #\n            #\n            #\n            pass\n            pass\n            #\n            pass\n            #\n\n    def testTest(self):\n        ### and_test ('or' and_test)*\n        ### and_test: not_test ('and' not_test)*\n        ### not_test: 'not' not_test | comparison\n        if not 1: pass\n        if 1 and 1: pass\n        if 1 or 1: pass\n        if not not not 1: pass\n        if not 1 and 1 and 1: pass\n        if 1 and 1 or 1 and 1 and 1 or not 1 and 1: pass\n\n    def testComparison(self):\n        ### comparison: expr (comp_op expr)*\n        ### comp_op: '<'|'>'|'=='|'>='|'<='|'<>'|'!='|'in'|'not' 'in'|'is'|'is' 'not'\n        if 1: pass\n        x = (1 == 1)\n        if 1 == 1: pass\n        if 1 != 1: pass\n        if 1 <> 1: pass\n        if 1 < 1: pass\n        if 1 > 1: pass\n        if 1 <= 1: pass\n        if 1 >= 1: pass\n        if 1 is 1: pass\n        if 1 is not 1: pass\n        if 1 in (): pass\n        if 1 not in (): pass\n        if 1 < 1 > 1 == 1 >= 1 <= 1 <> 1 != 1 in 1 not in 1 is 1 is not 1: pass\n\n    def testBinaryMaskOps(self):\n        x = 1 & 1\n        x = 1 ^ 1\n        x = 1 | 1\n\n    def testShiftOps(self):\n        x = 1 << 1\n        x = 1 >> 1\n        x = 1 << 1 >> 1\n\n    def testAdditiveOps(s\n# ... truncated ...","source_hash":"5008d0357851f744321f0ec3e8c6568214d5dd88e251c9a861ad533b1557d92a","truncated":true}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.demo","uri":"program://LLaMA-Adapter/module/imagebind_LLM.demo#L1-L22","kind":"module","name":"imagebind_LLM.demo","path":"imagebind_LLM/demo.py","language":"python","start_line":1,"end_line":22,"context_start_line":1,"context_end_line":22,"code":"import ImageBind.data as data\nimport llama\n\n\nllama_dir = \"/path/to/LLaMA\"\n\nmodel = llama.load(\"7B\", llama_dir, knn=True)\nmodel.eval()\n\ninputs = {}\nimage = data.load_and_transform_vision_data([\"examples/girl.jpg\"], device='cuda')\ninputs['Image'] = [image, 1]\naudio = data.load_and_transform_audio_data(['examples/girl_bgm.wav'], device='cuda')\ninputs['Audio'] = [audio, 1]\n\nresults = model.generate(\n    inputs,\n    [llama.format_prompt(\"Guess the girl's mood based on the background music and explain the reason?\")],\n    max_gen_len=256\n)\nresult = results[0].strip()\nprint(result)","source_hash":"520a739d1e57efbfcb4b801c864db1ad9e97743e74ad4d5bd72c263157ecf4de","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.demo_3d","uri":"program://LLaMA-Adapter/module/imagebind_LLM.demo_3d#L1-L20","kind":"module","name":"imagebind_LLM.demo_3d","path":"imagebind_LLM/demo_3d.py","language":"python","start_line":1,"end_line":20,"context_start_line":1,"context_end_line":20,"code":"import ImageBind.data as data\nimport llama\n\n\nllama_dir = \"/path/to/LLaMA\"\n\nmodel = llama.load(\"7B\", llama_dir, knn=True)\nmodel.eval()\n\ninputs = {}\npoint = data.load_and_transform_point_cloud_data([\"examples/airplane.pt\"], device='cuda')\ninputs['Point'] = [point, 1]\n\nresults = model.generate(\n    inputs,\n    [llama.format_prompt(\"Describe the 3D object in detail.\")],\n    max_gen_len=256\n)\nresult = results[0].strip()\nprint(result)","source_hash":"5d4916dbfe01372c3c782cc75a980c7b5dc75c0af139a13d3dd22145241514a8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.main_finetune","uri":"program://LLaMA-Adapter/module/imagebind_LLM.main_finetune#L1-L205","kind":"module","name":"imagebind_LLM.main_finetune","path":"imagebind_LLM/main_finetune.py","language":"python","start_line":1,"end_line":205,"context_start_line":1,"context_end_line":205,"code":"import torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import FinetuneDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_finetune import train_one_epoch\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('imagebind-llm pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=64, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--epochs', default=400, type=int)\n    parser.add_argument('--accum_iter', default=1, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='7B_chinese', type=str,\n                        help='Type of LLaMA model') #\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='path to checkpoint from pretrain stage')\n    parser.add_argument('--max_words', default=512, type=int,\n                        help='max number of input words')\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.05,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=None, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--blr', type=float, default=1e-3, metavar='LR',\n                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')\n    parser.add_argument('--min_lr', type=float, default=0., metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_epochs', type=int, default=40, metavar='N',\n                        help='epochs to warmup LR')\n\n    # Dataset parameters\n    parser.add_argument('--data_config', default='configs/data/finetune/EN.yaml', type=str,\n                        help='dataset config path')\n    parser.add_argument('--num_workers', default=10, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n\n    parser.add_argument('--output_dir', default='./output',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')\n    model = LLaMA_adapter(llama_ckpt_dir, llama_tokenzier_path)\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    print(\"Trainable Params:\")\n    print([(key, val.shape) for key, val in model.named_parameters() if val.requires_grad])\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # training detail\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n    misc.load_model(model_without_ddp, args.pretrained_path)\n\n\n    dataset_train = FinetuneDataset(args.data_config, transform=transform_train,\n                                max_words=args.max_words, tokenizer_path=llama_tokenzier_path)\n    print(dataset_train)\n    num_tasks = misc.get_world_size()\n    global_rank = misc.get_rank()\n    sampler_train = torch.utils.data.DistributedSampler(\n        dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True\n    )\n    print(\"Sampler_train = %s\" % str(sampler_train))\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train, sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # SummaryWrite\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train,\n            optimizer, device, epoch, loss_scaler,\n            log_writer=log_writer,\n            args=args\n        )\n\n        if args.output_dir and (epoch % 5 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,\n                loss_scaler=loss_scaler, epoch=epoch)\n\n        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},\n                     'epoch': epoch,\n                     **{f'val_{k}': v for k, v in train_stats.items()}}\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"ff1670518e18fa13f63278df89d074094d85ceb7f336ff9bb7b123e2abb8e9aa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.main_finetune.get_args_parser","uri":"program://LLaMA-Adapter/function/imagebind_LLM.main_finetune.get_args_parser#L23-L85","kind":"function","name":"get_args_parser","path":"imagebind_LLM/main_finetune.py","language":"python","start_line":23,"end_line":85,"context_start_line":3,"context_end_line":105,"code":"from torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import FinetuneDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_finetune import train_one_epoch\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('imagebind-llm pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=64, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--epochs', default=400, type=int)\n    parser.add_argument('--accum_iter', default=1, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='7B_chinese', type=str,\n                        help='Type of LLaMA model') #\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--pretrained_path', default='/path/to/pretrained', type=str,\n                        help='path to checkpoint from pretrain stage')\n    parser.add_argument('--max_words', default=512, type=int,\n                        help='max number of input words')\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.05,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=None, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--blr', type=float, default=1e-3, metavar='LR',\n                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')\n    parser.add_argument('--min_lr', type=float, default=0., metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_epochs', type=int, default=40, metavar='N',\n                        help='epochs to warmup LR')\n\n    # Dataset parameters\n    parser.add_argument('--data_config', default='configs/data/finetune/EN.yaml', type=str,\n                        help='dataset config path')\n    parser.add_argument('--num_workers', default=10, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n\n    parser.add_argument('--output_dir', default='./output',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')","source_hash":"ff1670518e18fa13f63278df89d074094d85ceb7f336ff9bb7b123e2abb8e9aa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.main_finetune.main","uri":"program://LLaMA-Adapter/function/imagebind_LLM.main_finetune.main#L88-L197","kind":"function","name":"main","path":"imagebind_LLM/main_finetune.py","language":"python","start_line":88,"end_line":197,"context_start_line":68,"context_end_line":205,"code":"                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')\n    model = LLaMA_adapter(llama_ckpt_dir, llama_tokenzier_path)\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    print(\"Trainable Params:\")\n    print([(key, val.shape) for key, val in model.named_parameters() if val.requires_grad])\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # training detail\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n    misc.load_model(model_without_ddp, args.pretrained_path)\n\n\n    dataset_train = FinetuneDataset(args.data_config, transform=transform_train,\n                                max_words=args.max_words, tokenizer_path=llama_tokenzier_path)\n    print(dataset_train)\n    num_tasks = misc.get_world_size()\n    global_rank = misc.get_rank()\n    sampler_train = torch.utils.data.DistributedSampler(\n        dataset_train, num_replicas=num_tasks, rank=global_rank, shuffle=True\n    )\n    print(\"Sampler_train = %s\" % str(sampler_train))\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train, sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # SummaryWrite\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train,\n            optimizer, device, epoch, loss_scaler,\n            log_writer=log_writer,\n            args=args\n        )\n\n        if args.output_dir and (epoch % 5 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,\n                loss_scaler=loss_scaler, epoch=epoch)\n\n        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},\n                     'epoch': epoch,\n                     **{f'val_{k}': v for k, v in train_stats.items()}}\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"ff1670518e18fa13f63278df89d074094d85ceb7f336ff9bb7b123e2abb8e9aa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.main_pretrain","uri":"program://LLaMA-Adapter/module/imagebind_LLM.main_pretrain#L1-L202","kind":"module","name":"imagebind_LLM.main_pretrain","path":"imagebind_LLM/main_pretrain.py","language":"python","start_line":1,"end_line":202,"context_start_line":1,"context_end_line":202,"code":"import torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import PretrainDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_pretrain import train_one_epoch\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('imagebind-llm pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=64, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--epochs', default=400, type=int)\n    parser.add_argument('--accum_iter', default=1, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='7B_chinese', type=str,\n                        help='Type of LLaMA model') #\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--max_words', default=96, type=int,\n                        help='max number of input words')\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.05,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=None, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--blr', type=float, default=1e-3, metavar='LR',\n                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')\n    parser.add_argument('--min_lr', type=float, default=0., metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_epochs', type=int, default=40, metavar='N',\n                        help='epochs to warmup LR')\n\n    # Dataset parameters\n    parser.add_argument('--data_config', default='configs/data/pretrain/EN.yaml', type=str,\n                        help='dataset config path')\n    parser.add_argument('--num_workers', default=10, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n\n    parser.add_argument('--output_dir', default='./output',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--split_epoch', type=int, default=50)\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')\n    model = LLaMA_adapter(llama_ckpt_dir, llama_tokenzier_path, knn=False, phase=\"pretrain\")\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    print(\"Trainable Params:\")\n    print([(key, val.shape) for key, val in model.named_parameters() if val.requires_grad])\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # training detail\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n\n\n    dataset_train = PretrainDataset(args.data_config, transform=transform_train,\n                                max_words=args.max_words, tokenizer_path=llama_tokenzier_path)\n    print(dataset_train)\n    num_tasks = misc.get_world_size()\n    global_rank = misc.get_rank()\n    sampler_train = misc.DistributedSubEpochSampler(\n        dataset_train, num_replicas=num_tasks, rank=global_rank, split_epoch=args.split_epoch, shuffle=True\n    )\n    print(\"Sampler_train = %s\" % str(sampler_train))\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train, sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # SummaryWrite\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train,\n            optimizer, device, epoch, loss_scaler,\n            log_writer=log_writer,\n            args=args\n        )\n\n        if args.output_dir and (epoch % 2 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,\n                loss_scaler=loss_scaler, epoch=epoch)\n\n        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},\n                     'epoch': epoch}\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"4a687bac8852db4420d7421898270cd7629d02950989a8cdbd34fba84d19402c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.main_pretrain.get_args_parser","uri":"program://LLaMA-Adapter/function/imagebind_LLM.main_pretrain.get_args_parser#L23-L84","kind":"function","name":"get_args_parser","path":"imagebind_LLM/main_pretrain.py","language":"python","start_line":23,"end_line":84,"context_start_line":3,"context_end_line":104,"code":"from torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import PretrainDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_pretrain import train_one_epoch\n\n\ndef get_args_parser():\n    parser = argparse.ArgumentParser('imagebind-llm pre-training', add_help=False)\n    parser.add_argument('--batch_size', default=64, type=int,\n                        help='Batch size per GPU (effective batch size is batch_size * accum_iter * # gpus')\n    parser.add_argument('--epochs', default=400, type=int)\n    parser.add_argument('--accum_iter', default=1, type=int,\n                        help='Accumulate gradient iterations (for increasing the effective batch size under memory constraints)')\n\n    # Model parameters\n    parser.add_argument('--llama_type', default='7B_chinese', type=str,\n                        help='Type of LLaMA model') #\n    parser.add_argument('--llama_path', default='/path/to/llama', type=str,\n                        help='path to LLaMA pretrained checkpoint')\n    parser.add_argument('--max_words', default=96, type=int,\n                        help='max number of input words')\n\n    # Optimizer parameters\n    parser.add_argument('--weight_decay', type=float, default=0.05,\n                        help='weight decay (default: 0.05)')\n\n    parser.add_argument('--lr', type=float, default=None, metavar='LR',\n                        help='learning rate (absolute lr)')\n    parser.add_argument('--blr', type=float, default=1e-3, metavar='LR',\n                        help='base learning rate: absolute_lr = base_lr * total_batch_size / 256')\n    parser.add_argument('--min_lr', type=float, default=0., metavar='LR',\n                        help='lower lr bound for cyclic schedulers that hit 0')\n\n    parser.add_argument('--warmup_epochs', type=int, default=40, metavar='N',\n                        help='epochs to warmup LR')\n\n    # Dataset parameters\n    parser.add_argument('--data_config', default='configs/data/pretrain/EN.yaml', type=str,\n                        help='dataset config path')\n    parser.add_argument('--num_workers', default=10, type=int)\n    parser.add_argument('--pin_mem', action='store_true',\n                        help='Pin CPU memory in DataLoader for more efficient (sometimes) transfer to GPU.')\n    parser.add_argument('--no_pin_mem', action='store_false', dest='pin_mem')\n    parser.set_defaults(pin_mem=True)\n\n\n    parser.add_argument('--output_dir', default='./output',\n                        help='path where to save, empty for no saving')\n    parser.add_argument('--log_dir', default='./output',\n                        help='path where to tensorboard log')\n    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--split_epoch', type=int, default=50)\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')","source_hash":"4a687bac8852db4420d7421898270cd7629d02950989a8cdbd34fba84d19402c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.main_pretrain.main","uri":"program://LLaMA-Adapter/function/imagebind_LLM.main_pretrain.main#L87-L194","kind":"function","name":"main","path":"imagebind_LLM/main_pretrain.py","language":"python","start_line":87,"end_line":194,"context_start_line":67,"context_end_line":202,"code":"    parser.add_argument('--device', default='cuda',\n                        help='device to use for training / testing')\n    parser.add_argument('--seed', default=0, type=int)\n\n    parser.add_argument('--start_epoch', default=0, type=int, metavar='N',\n                        help='start epoch')\n\n    # distributed training parameters\n    parser.add_argument('--world_size', default=1, type=int,\n                        help='number of distributed processes')\n    parser.add_argument('--local_rank', default=-1, type=int)\n    parser.add_argument('--dist_on_itp', action='store_true')\n    parser.add_argument('--dist_url', default='env://',\n                        help='url used to set up distributed training')\n\n    parser.add_argument('--split_epoch', type=int, default=50)\n\n    return parser\n\n\ndef main(args):\n    misc.init_distributed_mode(args)\n\n    print('job dir: {}'.format(os.path.dirname(os.path.realpath(__file__))))\n    print(\"{}\".format(args).replace(', ', ',\\n'))\n\n    device = torch.device(args.device)\n\n    # fix the seed for reproducibility\n    seed = args.seed + misc.get_rank()\n    torch.manual_seed(seed)\n    np.random.seed(seed)\n    cudnn.benchmark = True\n\n    # define the model\n    llama_type = args.llama_type\n    llama_ckpt_dir = os.path.join(args.llama_path, llama_type)\n    llama_tokenzier_path = os.path.join(args.llama_path, 'tokenizer.model')\n    model = LLaMA_adapter(llama_ckpt_dir, llama_tokenzier_path, knn=False, phase=\"pretrain\")\n\n    model.to(device)\n\n    model_without_ddp = model\n    print(\"Model = %s\" % str(model_without_ddp))\n\n    print(\"Trainable Params:\")\n    print([(key, val.shape) for key, val in model.named_parameters() if val.requires_grad])\n\n    if args.distributed:\n        model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu], find_unused_parameters=True)\n        model_without_ddp = model.module\n\n    # training detail\n    eff_batch_size = args.batch_size * args.accum_iter * misc.get_world_size()\n\n    if args.lr is None:  # only base_lr is specified\n        args.lr = args.blr * eff_batch_size / 256\n\n    print(\"base lr: %.2e\" % (args.lr * 256 / eff_batch_size))\n    print(\"actual lr: %.2e\" % args.lr)\n\n    print(\"accumulate grad iterations: %d\" % args.accum_iter)\n    print(\"effective batch size: %d\" % eff_batch_size)\n\n    # following timm: set wd as 0 for bias and norm layers\n    param_groups = misc.add_weight_decay(model_without_ddp, args.weight_decay)\n    optimizer = torch.optim.AdamW(param_groups, lr=args.lr, betas=(0.9, 0.95))\n    print(optimizer)\n    loss_scaler = NativeScaler()\n\n\n\n    dataset_train = PretrainDataset(args.data_config, transform=transform_train,\n                                max_words=args.max_words, tokenizer_path=llama_tokenzier_path)\n    print(dataset_train)\n    num_tasks = misc.get_world_size()\n    global_rank = misc.get_rank()\n    sampler_train = misc.DistributedSubEpochSampler(\n        dataset_train, num_replicas=num_tasks, rank=global_rank, split_epoch=args.split_epoch, shuffle=True\n    )\n    print(\"Sampler_train = %s\" % str(sampler_train))\n\n    data_loader_train = torch.utils.data.DataLoader(\n        dataset_train, sampler=sampler_train,\n        batch_size=args.batch_size,\n        num_workers=args.num_workers,\n        pin_memory=args.pin_mem,\n        drop_last=True,\n    )\n\n    # SummaryWrite\n    if global_rank == 0 and args.log_dir is not None:\n        os.makedirs(args.log_dir, exist_ok=True)\n        log_writer = SummaryWriter(log_dir=args.log_dir)\n    else:\n        log_writer = None\n\n\n    print(f\"Start training for {args.epochs} epochs\")\n    start_time = time.time()\n    for epoch in range(args.start_epoch, args.epochs):\n        if args.distributed:\n            data_loader_train.sampler.set_epoch(epoch)\n\n        train_stats = train_one_epoch(\n            model, data_loader_train,\n            optimizer, device, epoch, loss_scaler,\n            log_writer=log_writer,\n            args=args\n        )\n\n        if args.output_dir and (epoch % 2 == 0 or epoch + 1 == args.epochs):\n            misc.save_model(\n                args=args, model=model, model_without_ddp=model_without_ddp, optimizer=optimizer,\n                loss_scaler=loss_scaler, epoch=epoch)\n\n        log_stats = {**{f'train_{k}': v for k, v in train_stats.items()},\n                     'epoch': epoch}\n\n        if args.output_dir and misc.is_main_process():\n            if log_writer is not None:\n                log_writer.flush()\n            with open(os.path.join(args.output_dir, \"log.txt\"), mode=\"a\", encoding=\"utf-8\") as f:\n                f.write(json.dumps(log_stats) + \"\\n\")\n\n    total_time = time.time() - start_time\n    total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n    print('Training time {}'.format(total_time_str))\n\n\nif __name__ == '__main__':\n    args = get_args_parser()\n    args = args.parse_args()\n    if args.output_dir:\n        Path(args.output_dir).mkdir(parents=True, exist_ok=True)\n    main(args)","source_hash":"4a687bac8852db4420d7421898270cd7629d02950989a8cdbd34fba84d19402c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.engine_finetune","uri":"program://LLaMA-Adapter/module/imagebind_LLM.engine_finetune#L1-L77","kind":"module","name":"imagebind_LLM.engine_finetune","path":"imagebind_LLM/engine_finetune.py","language":"python","start_line":1,"end_line":77,"context_start_line":1,"context_end_line":77,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, imgs) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        imgs = imgs.to(device, non_blocking=True)\n        with torch.cuda.amp.autocast():\n             c_loss, m_loss = model(examples, labels, imgs)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n        loss_scaler(loss, optimizer, parameters=model.parameters(),\n                    update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.engine_finetune.train_one_epoch","uri":"program://LLaMA-Adapter/function/imagebind_LLM.engine_finetune.train_one_epoch#L12-L77","kind":"function","name":"train_one_epoch","path":"imagebind_LLM/engine_finetune.py","language":"python","start_line":12,"end_line":77,"context_start_line":1,"context_end_line":77,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, imgs) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        imgs = imgs.to(device, non_blocking=True)\n        with torch.cuda.amp.autocast():\n             c_loss, m_loss = model(examples, labels, imgs)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n        loss_scaler(loss, optimizer, parameters=model.parameters(),\n                    update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.gradio_app","uri":"program://LLaMA-Adapter/module/imagebind_LLM.gradio_app#L1-L271","kind":"module","name":"imagebind_LLM.gradio_app","path":"imagebind_LLM/gradio_app.py","language":"python","start_line":1,"end_line":271,"context_start_line":1,"context_end_line":271,"code":"import os\nimport argparse\n\nimport gradio as gr\nimport plotly.graph_objects as go\nimport torch, numpy, random\nimport torch.cuda\n\nimport ImageBind.data as data\nfrom diffusers import StableUnCLIPImg2ImgPipeline\nfrom image_generate import image_generate\n\nimport llama\n\nparser = argparse.ArgumentParser()\nparser.add_argument(\n    \"--model\", default=\"7B\", type=str,\n    help=\"Name of or path to ImageBind-LLM pretrained checkpoint\",\n)\nparser.add_argument(\n    \"--llama_type\", default=\"7B_chinese\", type=str,\n    help=\"Type of llama original weight\",\n)\nparser.add_argument(\n    \"--llama_dir\", default=\"/path/to/llama\", type=str,\n    help=\"Path to LLaMA pretrained checkpoint\",\n)\nargs = parser.parse_args()\n\nmodel = llama.load(args.model, args.llama_dir, knn=True, llama_type=args.llama_type)\nmodel.eval()\n\npipe = StableUnCLIPImg2ImgPipeline.from_pretrained(\"stabilityai/stable-diffusion-2-1-unclip\", cache_dir=\"./ckpts\")\npipe = pipe.to(\"cuda\")\n\n\ndef multimodal_generate(\n        modality,\n        img_path,\n        img_weight,\n        text_path,\n        text_weight,\n        video_path,\n        video_weight,\n        audio_path,\n        audio_weight,\n        point_path,\n        point_weight,\n        prompt,\n        question_input,\n        cache_size,\n        cache_t,\n        cache_weight,\n        max_gen_len,\n        gen_t, top_p, output_type\n):\n    if len(modality) == 0:\n        raise gr.Error('Please select at least one modality!')\n\n    inputs = {}\n    if 'Image' in modality:\n        if img_path is None:\n            raise gr.Error('Please select an image')\n        if img_weight == 0:\n            raise gr.Error('Please set the weight')\n        image = data.load_and_transform_vision_data([img_path], device='cuda')\n        inputs['Image'] = [image, img_weight]\n    if 'Text' in modality:\n        if text_path == '':\n            raise gr.Error('Please input the text')\n        if text_weight == 0:\n            raise gr.Error('Please set the weight')\n        text = data.load_and_transform_text([text_path], device='cuda')\n        inputs['Text'] = [text, text_weight]\n    if 'Video' in modality:\n        if video_path is None:\n            raise gr.Error('Please select a video')\n        if video_weight == 0:\n            raise gr.Error('Please set the weight')\n        video = data.load_and_transform_video_data([video_path], device='cuda')\n        inputs['Video'] = [video, video_weight]\n    if 'Audio' in modality:\n        if audio_path is None:\n            raise gr.Error('Please select an audio')\n        if audio_weight == 0:\n            raise gr.Error('Please set the weight')\n        audio = data.load_and_transform_audio_data([audio_path], device='cuda')\n        inputs['Audio'] = [audio, audio_weight]\n    if 'Point Cloud' in modality:\n        if point_path is None:\n            raise gr.Error('Please select a point cloud')\n        if point_weight == 0:\n            raise gr.Error('Please set the weight')\n        point = data.load_and_transform_point_cloud_data([point_path], device='cuda')\n        inputs['Point'] = [point, point_weight]\n\n    image_prompt = prompt # image use original prompt\n\n    text_output = None\n    image_output = None\n    if output_type == \"Text\":\n        # text output\n        prompts = [llama.format_prompt(prompt, question_input)]\n\n        prompts = [model.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n        with torch.cuda.amp.autocast():\n            results = model.generate(inputs, prompts, max_gen_len=max_gen_len, temperature=gen_t, top_p=top_p,\n                                         cache_size=cache_size, cache_t=cache_t, cache_weight=cache_weight)\n        text_output = results[0].strip()\n        print(text_output)\n\n    else:\n        # image output\n        image_output = image_generate(inputs, model, pipe, image_prompt, cache_size, cache_t, cache_weight)\n\n    return text_output, image_output\n\ndef show_point_cloud(file):\n    point = torch.load(file.name).numpy()\n    fig = go.Figure(\n        data=[\n            go.Scatter3d(\n                x=point[:,0], y=point[:,1], z=point[:,2], \n                mode='markers',\n                marker=dict(\n                size=1.2,\n                color='gray'\n            )\n            )\n        ],\n        layout=dict(\n            scene=dict(\n                xaxis=dict(visible=False),\n                yaxis=dict(visible=False),\n                zaxis=dict(visible=False)\n            )\n        ),\n    )\n    return fig\n\n\ndef create_imagebind_llm_demo():\n    with gr.Blocks() as imagebind_llm_demo:\n        modality = gr.CheckboxGroup(choices=['Image', 'Text', 'Video', 'Audio', 'Point Cloud'], value='Image', interactive=True,\n                                    label='Input Modalities')\n        with gr.Row():\n            with gr.Column():\n                with gr.Row():\n                    with gr.Column() as image_input:\n                        img_path = gr.Image(label='Image Input', type='filepath')\n                        img_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                    with gr.Column() as text_input:\n                        text_path = gr.Textbox(label='Text Input', lines=9)\n                        text_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                with gr.Row():\n                    with gr.Column() as video_input:\n                        video_path = gr.Video(label='Video Input')\n                        video_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                    with gr.Column() as audio_input:\n                        audio_path = gr.Audio(label='Audio Input', type='filepath')\n                        audio_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                with gr.Row():\n                    with gr.Column(scale=1) as point_input:\n                        point_path = gr.File(label='Point Cloud Input', elem_id=\"pointpath\", elem_classes=\"\")\n                        output = gr.Plot()\n                        btn = gr.Button(value=\"Show Point Cloud\")\n                        btn.click(show_point_cloud, point_path, output)\n                        point_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n            with gr.Column():\n                with gr.Row():\n                    output_dropdown = gr.Dropdown(['Text', 'Image'], value='Text', label='Output type')\n                with gr.Row():\n                    prompt = gr.Textbox(lines=4, label=\"Question\")\n                with gr.Row():\n                    question_input = gr.Textbox(lines=4, label=\"Question Input (Optional)\")\n                with gr.Row():\n                    cache_size = gr.Slider(minimum=1, maximum=100, value=10, interactive=True, label=\"Cache Size\")\n                    cache_t = gr.Slider(minimum=0.0, maximum=100, value=20, interactive=True, label=\"Cache Temperature\")\n                    cache_weight = gr.Slider(minimum=0.0, maximum=1, value=0.5, interactive=True, label=\"Cache Weight\")\n                with gr.Row() as text_config_row:\n                    max_gen_len = gr.Slider(minimum=1, maximum=512, value=128, interactive=True, label=\"Max Length\")\n                    # with gr.Accordion(label='Advanced options', open=False):\n                    gen_t = gr.Slider(minimum=0, maximum=1, value=0.1, interactive=True, label=\"Temperature\")\n                    top_p = gr.Slider(minimum=0, maximum=1, value=0.75, interactive=True, label=\"Top p\")\n                with gr.Row():\n                    # clear_botton = gr.Button(\"Clear\")\n                    run_botton = gr.Button(\"Run\", variant='primary')\n\n                with gr.Row():\n                    gr.Markdown(\"Output\")\n                with gr.Row():\n                    text_output = gr.Textbox(lines=11, label='Text Out')\n                    image_output = gr.Image(label='Image Out', visible=False)\n\n    def modality_select(modality, img, text, video, audio, point):\n        modality = []\n        if img is not None:\n            modality.append('Image')\n        if len(text) > 0:\n            modality.append('Text')\n        if video is not None:\n            modality.append('Video')\n        if audio is not None:\n            modality.append('Audio')\n        if point is not None:\n            modality.append('Point Cloud')\n        return modality\n\n    def change_output_type(output_type):\n        if output_type == 'Text':\n            result = [gr.update(visible=False),\n            gr.update(visible=True),\n            gr.update(visible=True),\n            gr.update(label='Question'),\n            gr.update(visible=True)]\n        elif output_type == 'Image':\n            result = [gr.update(visible=True),\n            gr.update(visible=False),\n            gr.update(visible=False),\n            gr.update(label='Prompt'),\n            gr.update(visible=False)]\n\n        return result\n\n    output_dropdown.change(change_output_type, output_dropdown,\n                           [image_output, text_output, question_input, prompt, text_config_row])\n\n\n    img_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path], outputs=[modality])\n    text_path.blur(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path], outputs=[modality])\n    video_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path],\n                      outputs=[modality])\n    audio_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path],\n                      outputs=[modality])\n    point_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path],\n                      outputs=[modality])\n\n    inputs = [\n        modality,\n        img_path, img_weight,\n        text_path, text_weight,\n        video_path, video_weight,\n        audio_path, audio_weight,\n        point_path, point_weight,\n        prompt, question_input,\n        cache_size, cache_t, cache_weight,\n        max_gen_len, gen_t, top_p, output_dropdown\n    ]\n    outputs = [text_output, image_output]\n    run_botton.click(fn=multimodal_generate, inputs=inputs, outputs=outputs)\n\n    # gr.Examples(\n    #     examples=examples,\n    #     inputs=inputs,\n    #     outputs=outputs,\n    #     fn=multimodal_generate,\n    #     cache_examples=False)\n\n    return imagebind_llm_demo\n\n\ndescription = \"\"\"\n# ImageBind-LLM🚀\n\"\"\"\n\nwith gr.Blocks(theme=gr.themes.Default(), css=\"#pointpath {height: 10em} .label {height: 3em}\") as demo:\n    gr.Markdown(description)\n    create_imagebind_llm_demo()\n\n\ndemo.queue(api_open=True, concurrency_count=1).launch(share=True)","source_hash":"cd48ac1e298f556bfd6ff9dbfc3d5a25d76532dfee22a81cdf1f3c22b3843f39","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.gradio_app.multimodal_generate","uri":"program://LLaMA-Adapter/function/imagebind_LLM.gradio_app.multimodal_generate#L37-L116","kind":"function","name":"multimodal_generate","path":"imagebind_LLM/gradio_app.py","language":"python","start_line":37,"end_line":116,"context_start_line":17,"context_end_line":136,"code":"    \"--model\", default=\"7B\", type=str,\n    help=\"Name of or path to ImageBind-LLM pretrained checkpoint\",\n)\nparser.add_argument(\n    \"--llama_type\", default=\"7B_chinese\", type=str,\n    help=\"Type of llama original weight\",\n)\nparser.add_argument(\n    \"--llama_dir\", default=\"/path/to/llama\", type=str,\n    help=\"Path to LLaMA pretrained checkpoint\",\n)\nargs = parser.parse_args()\n\nmodel = llama.load(args.model, args.llama_dir, knn=True, llama_type=args.llama_type)\nmodel.eval()\n\npipe = StableUnCLIPImg2ImgPipeline.from_pretrained(\"stabilityai/stable-diffusion-2-1-unclip\", cache_dir=\"./ckpts\")\npipe = pipe.to(\"cuda\")\n\n\ndef multimodal_generate(\n        modality,\n        img_path,\n        img_weight,\n        text_path,\n        text_weight,\n        video_path,\n        video_weight,\n        audio_path,\n        audio_weight,\n        point_path,\n        point_weight,\n        prompt,\n        question_input,\n        cache_size,\n        cache_t,\n        cache_weight,\n        max_gen_len,\n        gen_t, top_p, output_type\n):\n    if len(modality) == 0:\n        raise gr.Error('Please select at least one modality!')\n\n    inputs = {}\n    if 'Image' in modality:\n        if img_path is None:\n            raise gr.Error('Please select an image')\n        if img_weight == 0:\n            raise gr.Error('Please set the weight')\n        image = data.load_and_transform_vision_data([img_path], device='cuda')\n        inputs['Image'] = [image, img_weight]\n    if 'Text' in modality:\n        if text_path == '':\n            raise gr.Error('Please input the text')\n        if text_weight == 0:\n            raise gr.Error('Please set the weight')\n        text = data.load_and_transform_text([text_path], device='cuda')\n        inputs['Text'] = [text, text_weight]\n    if 'Video' in modality:\n        if video_path is None:\n            raise gr.Error('Please select a video')\n        if video_weight == 0:\n            raise gr.Error('Please set the weight')\n        video = data.load_and_transform_video_data([video_path], device='cuda')\n        inputs['Video'] = [video, video_weight]\n    if 'Audio' in modality:\n        if audio_path is None:\n            raise gr.Error('Please select an audio')\n        if audio_weight == 0:\n            raise gr.Error('Please set the weight')\n        audio = data.load_and_transform_audio_data([audio_path], device='cuda')\n        inputs['Audio'] = [audio, audio_weight]\n    if 'Point Cloud' in modality:\n        if point_path is None:\n            raise gr.Error('Please select a point cloud')\n        if point_weight == 0:\n            raise gr.Error('Please set the weight')\n        point = data.load_and_transform_point_cloud_data([point_path], device='cuda')\n        inputs['Point'] = [point, point_weight]\n\n    image_prompt = prompt # image use original prompt\n\n    text_output = None\n    image_output = None\n    if output_type == \"Text\":\n        # text output\n        prompts = [llama.format_prompt(prompt, question_input)]\n\n        prompts = [model.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n        with torch.cuda.amp.autocast():\n            results = model.generate(inputs, prompts, max_gen_len=max_gen_len, temperature=gen_t, top_p=top_p,\n                                         cache_size=cache_size, cache_t=cache_t, cache_weight=cache_weight)\n        text_output = results[0].strip()\n        print(text_output)\n\n    else:\n        # image output\n        image_output = image_generate(inputs, model, pipe, image_prompt, cache_size, cache_t, cache_weight)\n\n    return text_output, image_output\n\ndef show_point_cloud(file):\n    point = torch.load(file.name).numpy()\n    fig = go.Figure(\n        data=[\n            go.Scatter3d(\n                x=point[:,0], y=point[:,1], z=point[:,2], \n                mode='markers',\n                marker=dict(\n                size=1.2,\n                color='gray'\n            )\n            )\n        ],\n        layout=dict(\n            scene=dict(\n                xaxis=dict(visible=False),\n                yaxis=dict(visible=False),\n                zaxis=dict(visible=False)\n            )","source_hash":"cd48ac1e298f556bfd6ff9dbfc3d5a25d76532dfee22a81cdf1f3c22b3843f39","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.gradio_app.show_point_cloud","uri":"program://LLaMA-Adapter/function/imagebind_LLM.gradio_app.show_point_cloud#L118-L139","kind":"function","name":"show_point_cloud","path":"imagebind_LLM/gradio_app.py","language":"python","start_line":118,"end_line":139,"context_start_line":98,"context_end_line":159,"code":"\n    text_output = None\n    image_output = None\n    if output_type == \"Text\":\n        # text output\n        prompts = [llama.format_prompt(prompt, question_input)]\n\n        prompts = [model.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n        with torch.cuda.amp.autocast():\n            results = model.generate(inputs, prompts, max_gen_len=max_gen_len, temperature=gen_t, top_p=top_p,\n                                         cache_size=cache_size, cache_t=cache_t, cache_weight=cache_weight)\n        text_output = results[0].strip()\n        print(text_output)\n\n    else:\n        # image output\n        image_output = image_generate(inputs, model, pipe, image_prompt, cache_size, cache_t, cache_weight)\n\n    return text_output, image_output\n\ndef show_point_cloud(file):\n    point = torch.load(file.name).numpy()\n    fig = go.Figure(\n        data=[\n            go.Scatter3d(\n                x=point[:,0], y=point[:,1], z=point[:,2], \n                mode='markers',\n                marker=dict(\n                size=1.2,\n                color='gray'\n            )\n            )\n        ],\n        layout=dict(\n            scene=dict(\n                xaxis=dict(visible=False),\n                yaxis=dict(visible=False),\n                zaxis=dict(visible=False)\n            )\n        ),\n    )\n    return fig\n\n\ndef create_imagebind_llm_demo():\n    with gr.Blocks() as imagebind_llm_demo:\n        modality = gr.CheckboxGroup(choices=['Image', 'Text', 'Video', 'Audio', 'Point Cloud'], value='Image', interactive=True,\n                                    label='Input Modalities')\n        with gr.Row():\n            with gr.Column():\n                with gr.Row():\n                    with gr.Column() as image_input:\n                        img_path = gr.Image(label='Image Input', type='filepath')\n                        img_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                    with gr.Column() as text_input:\n                        text_path = gr.Textbox(label='Text Input', lines=9)\n                        text_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                with gr.Row():\n                    with gr.Column() as video_input:\n                        video_path = gr.Video(label='Video Input')\n                        video_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                    with gr.Column() as audio_input:","source_hash":"cd48ac1e298f556bfd6ff9dbfc3d5a25d76532dfee22a81cdf1f3c22b3843f39","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.gradio_app.create_imagebind_llm_demo","uri":"program://LLaMA-Adapter/function/imagebind_LLM.gradio_app.create_imagebind_llm_demo#L142-L259","kind":"function","name":"create_imagebind_llm_demo","path":"imagebind_LLM/gradio_app.py","language":"python","start_line":142,"end_line":259,"context_start_line":122,"context_end_line":271,"code":"            go.Scatter3d(\n                x=point[:,0], y=point[:,1], z=point[:,2], \n                mode='markers',\n                marker=dict(\n                size=1.2,\n                color='gray'\n            )\n            )\n        ],\n        layout=dict(\n            scene=dict(\n                xaxis=dict(visible=False),\n                yaxis=dict(visible=False),\n                zaxis=dict(visible=False)\n            )\n        ),\n    )\n    return fig\n\n\ndef create_imagebind_llm_demo():\n    with gr.Blocks() as imagebind_llm_demo:\n        modality = gr.CheckboxGroup(choices=['Image', 'Text', 'Video', 'Audio', 'Point Cloud'], value='Image', interactive=True,\n                                    label='Input Modalities')\n        with gr.Row():\n            with gr.Column():\n                with gr.Row():\n                    with gr.Column() as image_input:\n                        img_path = gr.Image(label='Image Input', type='filepath')\n                        img_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                    with gr.Column() as text_input:\n                        text_path = gr.Textbox(label='Text Input', lines=9)\n                        text_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                with gr.Row():\n                    with gr.Column() as video_input:\n                        video_path = gr.Video(label='Video Input')\n                        video_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                    with gr.Column() as audio_input:\n                        audio_path = gr.Audio(label='Audio Input', type='filepath')\n                        audio_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n                with gr.Row():\n                    with gr.Column(scale=1) as point_input:\n                        point_path = gr.File(label='Point Cloud Input', elem_id=\"pointpath\", elem_classes=\"\")\n                        output = gr.Plot()\n                        btn = gr.Button(value=\"Show Point Cloud\")\n                        btn.click(show_point_cloud, point_path, output)\n                        point_weight = gr.Slider(minimum=0.0, maximum=1.0, value=1.0, interactive=True, label='Weight')\n            with gr.Column():\n                with gr.Row():\n                    output_dropdown = gr.Dropdown(['Text', 'Image'], value='Text', label='Output type')\n                with gr.Row():\n                    prompt = gr.Textbox(lines=4, label=\"Question\")\n                with gr.Row():\n                    question_input = gr.Textbox(lines=4, label=\"Question Input (Optional)\")\n                with gr.Row():\n                    cache_size = gr.Slider(minimum=1, maximum=100, value=10, interactive=True, label=\"Cache Size\")\n                    cache_t = gr.Slider(minimum=0.0, maximum=100, value=20, interactive=True, label=\"Cache Temperature\")\n                    cache_weight = gr.Slider(minimum=0.0, maximum=1, value=0.5, interactive=True, label=\"Cache Weight\")\n                with gr.Row() as text_config_row:\n                    max_gen_len = gr.Slider(minimum=1, maximum=512, value=128, interactive=True, label=\"Max Length\")\n                    # with gr.Accordion(label='Advanced options', open=False):\n                    gen_t = gr.Slider(minimum=0, maximum=1, value=0.1, interactive=True, label=\"Temperature\")\n                    top_p = gr.Slider(minimum=0, maximum=1, value=0.75, interactive=True, label=\"Top p\")\n                with gr.Row():\n                    # clear_botton = gr.Button(\"Clear\")\n                    run_botton = gr.Button(\"Run\", variant='primary')\n\n                with gr.Row():\n                    gr.Markdown(\"Output\")\n                with gr.Row():\n                    text_output = gr.Textbox(lines=11, label='Text Out')\n                    image_output = gr.Image(label='Image Out', visible=False)\n\n    def modality_select(modality, img, text, video, audio, point):\n        modality = []\n        if img is not None:\n            modality.append('Image')\n        if len(text) > 0:\n            modality.append('Text')\n        if video is not None:\n            modality.append('Video')\n        if audio is not None:\n            modality.append('Audio')\n        if point is not None:\n            modality.append('Point Cloud')\n        return modality\n\n    def change_output_type(output_type):\n        if output_type == 'Text':\n            result = [gr.update(visible=False),\n            gr.update(visible=True),\n            gr.update(visible=True),\n            gr.update(label='Question'),\n            gr.update(visible=True)]\n        elif output_type == 'Image':\n            result = [gr.update(visible=True),\n            gr.update(visible=False),\n            gr.update(visible=False),\n            gr.update(label='Prompt'),\n            gr.update(visible=False)]\n\n        return result\n\n    output_dropdown.change(change_output_type, output_dropdown,\n                           [image_output, text_output, question_input, prompt, text_config_row])\n\n\n    img_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path], outputs=[modality])\n    text_path.blur(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path], outputs=[modality])\n    video_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path],\n                      outputs=[modality])\n    audio_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path],\n                      outputs=[modality])\n    point_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path],\n                      outputs=[modality])\n\n    inputs = [\n        modality,\n        img_path, img_weight,\n        text_path, text_weight,\n        video_path, video_weight,\n        audio_path, audio_weight,\n        point_path, point_weight,\n        prompt, question_input,\n        cache_size, cache_t, cache_weight,\n        max_gen_len, gen_t, top_p, output_dropdown\n    ]\n    outputs = [text_output, image_output]\n    run_botton.click(fn=multimodal_generate, inputs=inputs, outputs=outputs)\n\n    # gr.Examples(\n    #     examples=examples,\n    #     inputs=inputs,\n    #     outputs=outputs,\n    #     fn=multimodal_generate,\n    #     cache_examples=False)\n\n    return imagebind_llm_demo\n\n\ndescription = \"\"\"\n# ImageBind-LLM🚀\n\"\"\"\n\nwith gr.Blocks(theme=gr.themes.Default(), css=\"#pointpath {height: 10em} .label {height: 3em}\") as demo:\n    gr.Markdown(description)\n    create_imagebind_llm_demo()\n\n\ndemo.queue(api_open=True, concurrency_count=1).launch(share=True)","source_hash":"cd48ac1e298f556bfd6ff9dbfc3d5a25d76532dfee22a81cdf1f3c22b3843f39","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.gradio_app.modality_select","uri":"program://LLaMA-Adapter/function/imagebind_LLM.gradio_app.modality_select#L195-L207","kind":"function","name":"modality_select","path":"imagebind_LLM/gradio_app.py","language":"python","start_line":195,"end_line":207,"context_start_line":175,"context_end_line":227,"code":"                    question_input = gr.Textbox(lines=4, label=\"Question Input (Optional)\")\n                with gr.Row():\n                    cache_size = gr.Slider(minimum=1, maximum=100, value=10, interactive=True, label=\"Cache Size\")\n                    cache_t = gr.Slider(minimum=0.0, maximum=100, value=20, interactive=True, label=\"Cache Temperature\")\n                    cache_weight = gr.Slider(minimum=0.0, maximum=1, value=0.5, interactive=True, label=\"Cache Weight\")\n                with gr.Row() as text_config_row:\n                    max_gen_len = gr.Slider(minimum=1, maximum=512, value=128, interactive=True, label=\"Max Length\")\n                    # with gr.Accordion(label='Advanced options', open=False):\n                    gen_t = gr.Slider(minimum=0, maximum=1, value=0.1, interactive=True, label=\"Temperature\")\n                    top_p = gr.Slider(minimum=0, maximum=1, value=0.75, interactive=True, label=\"Top p\")\n                with gr.Row():\n                    # clear_botton = gr.Button(\"Clear\")\n                    run_botton = gr.Button(\"Run\", variant='primary')\n\n                with gr.Row():\n                    gr.Markdown(\"Output\")\n                with gr.Row():\n                    text_output = gr.Textbox(lines=11, label='Text Out')\n                    image_output = gr.Image(label='Image Out', visible=False)\n\n    def modality_select(modality, img, text, video, audio, point):\n        modality = []\n        if img is not None:\n            modality.append('Image')\n        if len(text) > 0:\n            modality.append('Text')\n        if video is not None:\n            modality.append('Video')\n        if audio is not None:\n            modality.append('Audio')\n        if point is not None:\n            modality.append('Point Cloud')\n        return modality\n\n    def change_output_type(output_type):\n        if output_type == 'Text':\n            result = [gr.update(visible=False),\n            gr.update(visible=True),\n            gr.update(visible=True),\n            gr.update(label='Question'),\n            gr.update(visible=True)]\n        elif output_type == 'Image':\n            result = [gr.update(visible=True),\n            gr.update(visible=False),\n            gr.update(visible=False),\n            gr.update(label='Prompt'),\n            gr.update(visible=False)]\n\n        return result\n\n    output_dropdown.change(change_output_type, output_dropdown,\n                           [image_output, text_output, question_input, prompt, text_config_row])\n","source_hash":"cd48ac1e298f556bfd6ff9dbfc3d5a25d76532dfee22a81cdf1f3c22b3843f39","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.gradio_app.change_output_type","uri":"program://LLaMA-Adapter/function/imagebind_LLM.gradio_app.change_output_type#L209-L223","kind":"function","name":"change_output_type","path":"imagebind_LLM/gradio_app.py","language":"python","start_line":209,"end_line":223,"context_start_line":189,"context_end_line":243,"code":"                with gr.Row():\n                    gr.Markdown(\"Output\")\n                with gr.Row():\n                    text_output = gr.Textbox(lines=11, label='Text Out')\n                    image_output = gr.Image(label='Image Out', visible=False)\n\n    def modality_select(modality, img, text, video, audio, point):\n        modality = []\n        if img is not None:\n            modality.append('Image')\n        if len(text) > 0:\n            modality.append('Text')\n        if video is not None:\n            modality.append('Video')\n        if audio is not None:\n            modality.append('Audio')\n        if point is not None:\n            modality.append('Point Cloud')\n        return modality\n\n    def change_output_type(output_type):\n        if output_type == 'Text':\n            result = [gr.update(visible=False),\n            gr.update(visible=True),\n            gr.update(visible=True),\n            gr.update(label='Question'),\n            gr.update(visible=True)]\n        elif output_type == 'Image':\n            result = [gr.update(visible=True),\n            gr.update(visible=False),\n            gr.update(visible=False),\n            gr.update(label='Prompt'),\n            gr.update(visible=False)]\n\n        return result\n\n    output_dropdown.change(change_output_type, output_dropdown,\n                           [image_output, text_output, question_input, prompt, text_config_row])\n\n\n    img_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path], outputs=[modality])\n    text_path.blur(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path], outputs=[modality])\n    video_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path],\n                      outputs=[modality])\n    audio_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path],\n                      outputs=[modality])\n    point_path.change(modality_select, inputs=[modality, img_path, text_path, video_path, audio_path, point_path],\n                      outputs=[modality])\n\n    inputs = [\n        modality,\n        img_path, img_weight,\n        text_path, text_weight,\n        video_path, video_weight,\n        audio_path, audio_weight,","source_hash":"cd48ac1e298f556bfd6ff9dbfc3d5a25d76532dfee22a81cdf1f3c22b3843f39","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.image_generate","uri":"program://LLaMA-Adapter/module/imagebind_LLM.image_generate#L1-L51","kind":"module","name":"imagebind_LLM.image_generate","path":"imagebind_LLM/image_generate.py","language":"python","start_line":1,"end_line":51,"context_start_line":1,"context_end_line":51,"code":"import llama\nimport torch\nimport numpy as np\n\n\n@torch.inference_mode()\ndef image_generate(inputs, model: llama.LLaMA_adapter, pipe, prompt, cache_size, cache_t, cache_weight, knn=True, point_scale=5.):\n\n    embeddings = []\n    embeddings_weights = []\n\n    for input_type, (input, input_weight) in inputs.items():\n        if input_type in ['Image', 'Video']:\n            type = 'vision'\n        else:\n            type = input_type.lower()\n        embedding = model.image_bind({type : input}, prenorm=True)[1][type]\n        if type == 'point':\n            embedding = embedding / point_scale\n        embeddings.append(embedding)\n        embeddings_weights.append(input_weight)\n    embeddings_weights = [x/(sum(embeddings_weights)+1e-6) for x in embeddings_weights]\n    embedding = sum([embedding*embedding_weight for embedding, embedding_weight in zip(embeddings, embeddings_weights)])\n\n    if knn:\n        index = model.index\n\n        embedding_norm_scale = embedding.norm(dim=-1, keepdim=True)\n        embedding = embedding / embedding_norm_scale\n        embedding_ori = embedding\n\n        sims, indices = index.search(embedding.detach().cpu(), int(cache_size))\n        B = sims.shape[0]\n        prototypes = [index.reconstruct(x) for x in indices.reshape(-1, ).tolist()]\n        prototypes = np.vstack(prototypes).reshape(B, int(cache_size), -1)  # [N, top_k, 1024]\n        sims = torch.tensor(sims, device='cuda')\n        prototypes = torch.tensor(prototypes, device='cuda')\n\n        sims = (sims * cache_t).softmax(dim=-1)\n        embedding = sims @ prototypes\n        embedding = embedding / embedding.norm(dim=-1, keepdim=True)\n\n        embedding = (1-cache_weight) * embedding_ori + cache_weight * embedding\n        embedding = embedding / embedding.norm(dim=-1, keepdim=True)\n\n        embedding = embedding_norm_scale*embedding\n\n    embedding = torch.squeeze(embedding,0)\n    image = pipe(prompt=prompt, image_embeds=embedding).images[0]\n\n    return image","source_hash":"4545f651278425e7793206653469106f5cd06367171233b17e7ee097fa2b155e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.image_generate.image_generate","uri":"program://LLaMA-Adapter/function/imagebind_LLM.image_generate.image_generate#L7-L51","kind":"function","name":"image_generate","path":"imagebind_LLM/image_generate.py","language":"python","start_line":7,"end_line":51,"context_start_line":1,"context_end_line":51,"code":"import llama\nimport torch\nimport numpy as np\n\n\n@torch.inference_mode()\ndef image_generate(inputs, model: llama.LLaMA_adapter, pipe, prompt, cache_size, cache_t, cache_weight, knn=True, point_scale=5.):\n\n    embeddings = []\n    embeddings_weights = []\n\n    for input_type, (input, input_weight) in inputs.items():\n        if input_type in ['Image', 'Video']:\n            type = 'vision'\n        else:\n            type = input_type.lower()\n        embedding = model.image_bind({type : input}, prenorm=True)[1][type]\n        if type == 'point':\n            embedding = embedding / point_scale\n        embeddings.append(embedding)\n        embeddings_weights.append(input_weight)\n    embeddings_weights = [x/(sum(embeddings_weights)+1e-6) for x in embeddings_weights]\n    embedding = sum([embedding*embedding_weight for embedding, embedding_weight in zip(embeddings, embeddings_weights)])\n\n    if knn:\n        index = model.index\n\n        embedding_norm_scale = embedding.norm(dim=-1, keepdim=True)\n        embedding = embedding / embedding_norm_scale\n        embedding_ori = embedding\n\n        sims, indices = index.search(embedding.detach().cpu(), int(cache_size))\n        B = sims.shape[0]\n        prototypes = [index.reconstruct(x) for x in indices.reshape(-1, ).tolist()]\n        prototypes = np.vstack(prototypes).reshape(B, int(cache_size), -1)  # [N, top_k, 1024]\n        sims = torch.tensor(sims, device='cuda')\n        prototypes = torch.tensor(prototypes, device='cuda')\n\n        sims = (sims * cache_t).softmax(dim=-1)\n        embedding = sims @ prototypes\n        embedding = embedding / embedding.norm(dim=-1, keepdim=True)\n\n        embedding = (1-cache_weight) * embedding_ori + cache_weight * embedding\n        embedding = embedding / embedding.norm(dim=-1, keepdim=True)\n\n        embedding = embedding_norm_scale*embedding\n\n    embedding = torch.squeeze(embedding,0)\n    image = pipe(prompt=prompt, image_embeds=embedding).images[0]\n\n    return image","source_hash":"4545f651278425e7793206653469106f5cd06367171233b17e7ee097fa2b155e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.convert_ckpt","uri":"program://LLaMA-Adapter/module/imagebind_LLM.convert_ckpt#L1-L53","kind":"module","name":"imagebind_LLM.convert_ckpt","path":"imagebind_LLM/convert_ckpt.py","language":"python","start_line":1,"end_line":53,"context_start_line":1,"context_end_line":53,"code":"import torch\nfrom collections import OrderedDict\nimport argparse\nfrom pathlib import Path\n\nparser = argparse.ArgumentParser()\nparser.add_argument(\n    \"--ori\", required=True, type=str,\n    help=\"Name of or path to LLaMAAdapter pretrained checkpoint\",\n)\nparser.add_argument(\n    \"--target\", default=None,\n    help=\"target position for the ckpt\",\n)\nargs = parser.parse_args()\n\nori_ckpt_path = Path(args.ori)\ntarget_ckpt_path = ori_ckpt_path.with_stem(\"converted_\" + ori_ckpt_path.stem)\n\nckpt = torch.load(ori_ckpt_path, map_location='cpu')\n\nreplace_dict = {\n    'llma': 'llama'\n}\nrenamed_ckpt = {}\nfor key, val in ckpt['model'].items():\n    for replace_key, replace_val in replace_dict.items():\n        key = key.replace(replace_key, replace_val)\n    renamed_ckpt[key] = val\n\n\nnew_ckpt = {}\ndiscarded  = []\n\nfor key, val in renamed_ckpt.items():\n    if key.startswith('image_bind.'):\n        discarded.append(key)\n    elif key.startswith(\"llama.\") and \"bias\" not in key and \"gate\" not in key and \"lora\" not in key and \"norm\" not in key:\n        discarded.append(key)\n    else:\n        new_ckpt[key] = val\n\nto_remove = ['prefix_projector_norm.weight', 'prefix_projector_norm.bias']\nfor _ in to_remove:\n    if _ in new_ckpt:\n        del new_ckpt[_]\n\n\nprint(f\"discarded: {discarded}\")\nprint(f\"saved: {list(new_ckpt.keys())}\")\n\nnew_ckpt = {'model': new_ckpt}\ntorch.save(new_ckpt, target_ckpt_path)","source_hash":"9d4665a5d6453dfdc2e5f3622a37bfdcaaa771818947bff9644a614f03166c42","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.engine_pretrain","uri":"program://LLaMA-Adapter/module/imagebind_LLM.engine_pretrain#L1-L77","kind":"module","name":"imagebind_LLM.engine_pretrain","path":"imagebind_LLM/engine_pretrain.py","language":"python","start_line":1,"end_line":77,"context_start_line":1,"context_end_line":77,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, imgs) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        imgs = imgs.to(device, non_blocking=True)\n        with torch.cuda.amp.autocast():\n             c_loss, m_loss = model(examples, labels, imgs)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n        loss_scaler(loss, optimizer, parameters=model.parameters(),\n                    update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.engine_pretrain.train_one_epoch","uri":"program://LLaMA-Adapter/function/imagebind_LLM.engine_pretrain.train_one_epoch#L12-L77","kind":"function","name":"train_one_epoch","path":"imagebind_LLM/engine_pretrain.py","language":"python","start_line":12,"end_line":77,"context_start_line":1,"context_end_line":77,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n    header = 'Epoch: [{}]'.format(epoch)\n    print_freq = 10\n\n    accum_iter = args.accum_iter\n\n    optimizer.zero_grad()\n\n    if log_writer is not None:\n        print('log_dir: {}'.format(log_writer.log_dir))\n    for data_iter_step, (examples, labels, example_mask, imgs) in enumerate(metric_logger.log_every(data_loader, print_freq, header)):\n        # we use a per iteration (instead of per epoch) lr scheduler\n        if data_iter_step % accum_iter == 0:\n            lr_sched.adjust_learning_rate(optimizer, data_iter_step / len(data_loader) + epoch, args)\n\n        imgs = imgs.to(device, non_blocking=True)\n        with torch.cuda.amp.autocast():\n             c_loss, m_loss = model(examples, labels, imgs)\n        loss = c_loss  + m_loss * 0\n        loss_value = loss.item()\n        c_loss_value = c_loss.item()\n        m_loss_value = m_loss\n        if not math.isfinite(loss_value):\n            print(\"Loss is {}, stopping training\".format(loss_value))\n            sys.exit(1)\n\n        loss /= accum_iter\n        loss_scaler(loss, optimizer, parameters=model.parameters(),\n                    update_grad=(data_iter_step + 1) % accum_iter == 0)\n        if (data_iter_step + 1) % accum_iter == 0:\n            optimizer.zero_grad()\n\n        torch.cuda.synchronize()\n\n        metric_logger.update(closs=c_loss_value)\n        metric_logger.update(mloss=m_loss_value)\n\n        lr = optimizer.param_groups[0][\"lr\"]\n        metric_logger.update(lr=lr)\n\n        loss_value_reduce = misc.all_reduce_mean(loss_value)\n        c_loss_value_reduce = misc.all_reduce_mean(c_loss_value)\n        m_loss_value_reduce = misc.all_reduce_mean(m_loss_value)\n        if log_writer is not None and (data_iter_step + 1) % accum_iter == 0:\n            \"\"\" We use epoch_1000x as the x-axis in tensorboard.\n            This calibrates different curves when batch size changes.\n            \"\"\"\n            epoch_1000x = int((data_iter_step / len(data_loader) + epoch) * 1000)\n            log_writer.add_scalar('c_train_loss', c_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('m_train_loss', m_loss_value_reduce, epoch_1000x)\n            log_writer.add_scalar('lr', lr, epoch_1000x)\n\n\n    # gather the stats from all processes\n    metric_logger.synchronize_between_processes()\n    print(\"Averaged stats:\", metric_logger)\n    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc","uri":"program://LLaMA-Adapter/module/imagebind_LLM.util.misc#L1-L413","kind":"module","name":"imagebind_LLM.util.misc","path":"imagebind_LLM/util/misc.py","language":"python","start_line":1,"end_line":413,"context_start_line":1,"context_end_line":413,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\nimport urllib\nfrom tqdm import tqdm\n\nimport torch\nimport torch.utils.data\nimport torch.distributed as dist\nfrom torch._six import inf\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    print(\"GPU::\", args.gpu)\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(model_without_ddp, path):\n    if path.startswith('https'):\n        checkpoint = torch.hub.load_state_dict_from_url(\n            path, map_location='cpu', check_hash=True)\n    else:\n        checkpoint = torch.load(path, map_location='cpu')\n    new_checkpoint = {}\n    for key, value in checkpoint['model'].items():\n        key = key.replace(\"llma\", \"llama\")\n        new_checkpoint[key] = value\n    print(model_without_ddp.load_state_dict(new_checkpoint, strict=False))\n    print(\"Load checkpoint %s\" % path)\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        return download_target\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:\n                    break\n                output.write(buffer)\n                loop.update(len(buffer))\n\n\n    return download_target","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.SmoothedValue","uri":"program://LLaMA-Adapter/class/imagebind_LLM.util.misc.SmoothedValue#L27-L86","kind":"class","name":"SmoothedValue","path":"imagebind_LLM/util/misc.py","language":"python","start_line":27,"end_line":86,"context_start_line":7,"context_end_line":106,"code":"# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\nimport urllib\nfrom tqdm import tqdm\n\nimport torch\nimport torch.utils.data\nimport torch.distributed as dist\nfrom torch._six import inf\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.MetricLogger","uri":"program://LLaMA-Adapter/class/imagebind_LLM.util.misc.MetricLogger#L89-L170","kind":"class","name":"MetricLogger","path":"imagebind_LLM/util/misc.py","language":"python","start_line":89,"end_line":170,"context_start_line":69,"context_end_line":190,"code":"    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.setup_for_distributed","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.setup_for_distributed#L173-L187","kind":"function","name":"setup_for_distributed","path":"imagebind_LLM/util/misc.py","language":"python","start_line":173,"end_line":187,"context_start_line":153,"context_end_line":207,"code":"                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.is_dist_avail_and_initialized","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.is_dist_avail_and_initialized#L190-L195","kind":"function","name":"is_dist_avail_and_initialized","path":"imagebind_LLM/util/misc.py","language":"python","start_line":190,"end_line":195,"context_start_line":170,"context_end_line":215,"code":"            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.get_world_size","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.get_world_size#L198-L201","kind":"function","name":"get_world_size","path":"imagebind_LLM/util/misc.py","language":"python","start_line":198,"end_line":201,"context_start_line":178,"context_end_line":221,"code":"\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.get_rank","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.get_rank#L204-L207","kind":"function","name":"get_rank","path":"imagebind_LLM/util/misc.py","language":"python","start_line":204,"end_line":207,"context_start_line":184,"context_end_line":227,"code":"            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.is_main_process","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.is_main_process#L210-L211","kind":"function","name":"is_main_process","path":"imagebind_LLM/util/misc.py","language":"python","start_line":210,"end_line":211,"context_start_line":190,"context_end_line":231,"code":"def is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.save_on_master","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.save_on_master#L214-L216","kind":"function","name":"save_on_master","path":"imagebind_LLM/util/misc.py","language":"python","start_line":214,"end_line":216,"context_start_line":194,"context_end_line":236,"code":"        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.init_distributed_mode","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.init_distributed_mode#L219-L252","kind":"function","name":"init_distributed_mode","path":"imagebind_LLM/util/misc.py","language":"python","start_line":219,"end_line":252,"context_start_line":199,"context_end_line":272,"code":"    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode(args):\n    if args.dist_on_itp:\n        args.rank = int(os.environ['OMPI_COMM_WORLD_RANK'])\n        args.world_size = int(os.environ['OMPI_COMM_WORLD_SIZE'])\n        args.gpu = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK'])\n        args.dist_url = \"tcp://%s:%s\" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT'])\n        os.environ['LOCAL_RANK'] = str(args.gpu)\n        os.environ['RANK'] = str(args.rank)\n        os.environ['WORLD_SIZE'] = str(args.world_size)\n        # [\"RANK\", \"WORLD_SIZE\", \"MASTER_ADDR\", \"MASTER_PORT\", \"LOCAL_RANK\"]\n    elif 'RANK' in os.environ and 'WORLD_SIZE' in os.environ:\n        args.rank = int(os.environ[\"RANK\"])\n        args.world_size = int(os.environ['WORLD_SIZE'])\n        args.gpu = int(os.environ['LOCAL_RANK'])\n    elif 'SLURM_PROCID' in os.environ:\n        args.rank = int(os.environ['SLURM_PROCID'])\n        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    print(\"GPU::\", args.gpu)\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.NativeScalerWithGradNormCount","uri":"program://LLaMA-Adapter/class/imagebind_LLM.util.misc.NativeScalerWithGradNormCount#L255-L281","kind":"class","name":"NativeScalerWithGradNormCount","path":"imagebind_LLM/util/misc.py","language":"python","start_line":255,"end_line":281,"context_start_line":235,"context_end_line":301,"code":"        args.gpu = args.rank % torch.cuda.device_count()\n    else:\n        print('Not using distributed mode')\n        setup_for_distributed(is_master=True)  # hack\n        args.distributed = False\n        return\n\n    args.distributed = True\n\n    print(\"GPU::\", args.gpu)\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.get_grad_norm_","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.get_grad_norm_#L284-L296","kind":"function","name":"get_grad_norm_","path":"imagebind_LLM/util/misc.py","language":"python","start_line":284,"end_line":296,"context_start_line":264,"context_end_line":316,"code":"            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.save_model","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.save_model#L299-L316","kind":"function","name":"save_model","path":"imagebind_LLM/util/misc.py","language":"python","start_line":299,"end_line":316,"context_start_line":279,"context_end_line":336,"code":"\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(model_without_ddp, path):\n    if path.startswith('https'):\n        checkpoint = torch.hub.load_state_dict_from_url(\n            path, map_location='cpu', check_hash=True)\n    else:\n        checkpoint = torch.load(path, map_location='cpu')\n    new_checkpoint = {}\n    for key, value in checkpoint['model'].items():\n        key = key.replace(\"llma\", \"llama\")\n        new_checkpoint[key] = value\n    print(model_without_ddp.load_state_dict(new_checkpoint, strict=False))\n    print(\"Load checkpoint %s\" % path)\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.load_model","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.load_model#L319-L330","kind":"function","name":"load_model","path":"imagebind_LLM/util/misc.py","language":"python","start_line":319,"end_line":330,"context_start_line":299,"context_end_line":350,"code":"def save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / ('checkpoint-%s.pth' % epoch_name)]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                'model': model_without_ddp.state_dict(),\n                'optimizer': optimizer.state_dict(),\n                'epoch': epoch,\n                'scaler': loss_scaler.state_dict(),\n                'args': args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(model_without_ddp, path):\n    if path.startswith('https'):\n        checkpoint = torch.hub.load_state_dict_from_url(\n            path, map_location='cpu', check_hash=True)\n    else:\n        checkpoint = torch.load(path, map_location='cpu')\n    new_checkpoint = {}\n    for key, value in checkpoint['model'].items():\n        key = key.replace(\"llma\", \"llama\")\n        new_checkpoint[key] = value\n    print(model_without_ddp.load_state_dict(new_checkpoint, strict=False))\n    print(\"Load checkpoint %s\" % path)\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.all_reduce_mean","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.all_reduce_mean#L333-L341","kind":"function","name":"all_reduce_mean","path":"imagebind_LLM/util/misc.py","language":"python","start_line":333,"end_line":341,"context_start_line":313,"context_end_line":361,"code":"            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {'epoch': epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(model_without_ddp, path):\n    if path.startswith('https'):\n        checkpoint = torch.hub.load_state_dict_from_url(\n            path, map_location='cpu', check_hash=True)\n    else:\n        checkpoint = torch.load(path, map_location='cpu')\n    new_checkpoint = {}\n    for key, value in checkpoint['model'].items():\n        key = key.replace(\"llma\", \"llama\")\n        new_checkpoint[key] = value\n    print(model_without_ddp.load_state_dict(new_checkpoint, strict=False))\n    print(\"Load checkpoint %s\" % path)\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.add_weight_decay","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.add_weight_decay#L344-L356","kind":"function","name":"add_weight_decay","path":"imagebind_LLM/util/misc.py","language":"python","start_line":344,"end_line":356,"context_start_line":324,"context_end_line":376,"code":"        checkpoint = torch.load(path, map_location='cpu')\n    new_checkpoint = {}\n    for key, value in checkpoint['model'].items():\n        key = key.replace(\"llma\", \"llama\")\n        new_checkpoint[key] = value\n    print(model_without_ddp.load_state_dict(new_checkpoint, strict=False))\n    print(\"Load checkpoint %s\" % path)\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.DistributedSubEpochSampler","uri":"program://LLaMA-Adapter/class/imagebind_LLM.util.misc.DistributedSubEpochSampler#L359-L390","kind":"class","name":"DistributedSubEpochSampler","path":"imagebind_LLM/util/misc.py","language":"python","start_line":359,"end_line":390,"context_start_line":339,"context_end_line":410,"code":"        return x_reduce.item()\n    else:\n        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        return download_target\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:\n                    break\n                output.write(buffer)\n                loop.update(len(buffer))","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.download","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.download#L392-L413","kind":"function","name":"download","path":"imagebind_LLM/util/misc.py","language":"python","start_line":392,"end_line":413,"context_start_line":372,"context_end_line":413,"code":"        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        return download_target\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:\n                    break\n                output.write(buffer)\n                loop.update(len(buffer))\n\n\n    return download_target","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.__init__#L361-L369","kind":"function","name":"__init__","path":"imagebind_LLM/util/misc.py","language":"python","start_line":361,"end_line":369,"context_start_line":341,"context_end_line":389,"code":"        return x\n\n\ndef add_weight_decay(model, weight_decay=1e-5, skip_list=()):\n    decay = []\n    no_decay = []\n    for name, param in model.named_parameters():\n        if not param.requires_grad:\n            continue  # frozen weights\n        if len(param.shape) == 1 or name.endswith(\".bias\") or name in skip_list:\n            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.update","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.update#L94-L101","kind":"function","name":"update","path":"imagebind_LLM/util/misc.py","language":"python","start_line":94,"end_line":101,"context_start_line":74,"context_end_line":121,"code":"        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.synchronize_between_processes","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.synchronize_between_processes#L119-L121","kind":"function","name":"synchronize_between_processes","path":"imagebind_LLM/util/misc.py","language":"python","start_line":119,"end_line":121,"context_start_line":99,"context_end_line":141,"code":"                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.median","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.median#L59-L61","kind":"function","name":"median","path":"imagebind_LLM/util/misc.py","language":"python","start_line":59,"end_line":61,"context_start_line":39,"context_end_line":81,"code":"\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.avg","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.avg#L64-L66","kind":"function","name":"avg","path":"imagebind_LLM/util/misc.py","language":"python","start_line":64,"end_line":66,"context_start_line":44,"context_end_line":86,"code":"\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.global_avg","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.global_avg#L69-L70","kind":"function","name":"global_avg","path":"imagebind_LLM/util/misc.py","language":"python","start_line":69,"end_line":70,"context_start_line":49,"context_end_line":90,"code":"        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device='cuda')\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.max","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.max#L73-L74","kind":"function","name":"max","path":"imagebind_LLM/util/misc.py","language":"python","start_line":73,"end_line":74,"context_start_line":53,"context_end_line":94,"code":"        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.value","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.value#L77-L78","kind":"function","name":"value","path":"imagebind_LLM/util/misc.py","language":"python","start_line":77,"end_line":78,"context_start_line":57,"context_end_line":98,"code":"\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median,\n            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.__str__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.__str__#L111-L117","kind":"function","name":"__str__","path":"imagebind_LLM/util/misc.py","language":"python","start_line":111,"end_line":117,"context_start_line":91,"context_end_line":137,"code":"        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.__getattr__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.__getattr__#L103-L109","kind":"function","name":"__getattr__","path":"imagebind_LLM/util/misc.py","language":"python","start_line":103,"end_line":109,"context_start_line":83,"context_end_line":129,"code":"            avg=self.avg,\n            global_avg=self.global_avg,\n            max=self.max,\n            value=self.value)\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.add_meter","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.add_meter#L123-L124","kind":"function","name":"add_meter","path":"imagebind_LLM/util/misc.py","language":"python","start_line":123,"end_line":124,"context_start_line":103,"context_end_line":144,"code":"    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.log_every","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.log_every#L126-L170","kind":"function","name":"log_every","path":"imagebind_LLM/util/misc.py","language":"python","start_line":126,"end_line":170,"context_start_line":106,"context_end_line":190,"code":"        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(\n            type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\n                \"{}: {}\".format(name, str(meter))\n            )\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = ''\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt='{avg:.4f}')\n        data_time = SmoothedValue(fmt='{avg:.4f}')\n        space_fmt = ':' + str(len(str(len(iterable)))) + 'd'\n        log_msg = [\n            header,\n            '[{0' + space_fmt + '}/{1}]',\n            'eta: {eta}',\n            '{meters}',\n            'time: {time}',\n            'data: {data}'\n        ]\n        if torch.cuda.is_available():\n            log_msg.append('max mem: {memory:.0f}')\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time),\n                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.print","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.print#L179-L185","kind":"function","name":"print","path":"imagebind_LLM/util/misc.py","language":"python","start_line":179,"end_line":185,"context_start_line":159,"context_end_line":205,"code":"                        memory=torch.cuda.max_memory_allocated() / MB))\n                else:\n                    print(log_msg.format(\n                        i, len(iterable), eta=eta_string,\n                        meters=str(self),\n                        time=str(iter_time), data=str(data_time)))\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print('{} Total time: {} ({:.4f} s / it)'.format(\n            header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop('force', False)\n        force = force or (get_world_size() > 8)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print('[{}] '.format(now), end='')  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.__call__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.__call__#L261-L275","kind":"function","name":"__call__","path":"imagebind_LLM/util/misc.py","language":"python","start_line":261,"end_line":275,"context_start_line":241,"context_end_line":295,"code":"\n    args.distributed = True\n\n    print(\"GPU::\", args.gpu)\n    torch.cuda.set_device(args.gpu)\n    args.dist_backend = 'nccl'\n    print('| distributed init (rank {}): {}, gpu {}'.format(\n        args.rank, args.dist_url, args.gpu), flush=True)\n    torch.distributed.init_process_group(backend=args.dist_backend, init_method=args.dist_url,\n                                         world_size=args.world_size, rank=args.rank)\n    torch.distributed.barrier()\n    setup_for_distributed(args.rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.state_dict","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.state_dict#L277-L278","kind":"function","name":"state_dict","path":"imagebind_LLM/util/misc.py","language":"python","start_line":277,"end_line":278,"context_start_line":257,"context_end_line":298,"code":"\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.load_state_dict","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.load_state_dict#L280-L281","kind":"function","name":"load_state_dict","path":"imagebind_LLM/util/misc.py","language":"python","start_line":280,"end_line":281,"context_start_line":260,"context_end_line":301,"code":"\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type)\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.__len__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.__len__#L371-L372","kind":"function","name":"__len__","path":"imagebind_LLM/util/misc.py","language":"python","start_line":371,"end_line":372,"context_start_line":351,"context_end_line":392,"code":"            no_decay.append(param)\n        else:\n            decay.append(param)\n    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.__iter__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.__iter__#L374-L387","kind":"function","name":"__iter__","path":"imagebind_LLM/util/misc.py","language":"python","start_line":374,"end_line":387,"context_start_line":354,"context_end_line":407,"code":"    return [\n        {'params': no_decay, 'weight_decay': 0.},\n        {'params': decay, 'weight_decay': weight_decay}]\n\n\nclass DistributedSubEpochSampler(torch.utils.data.Sampler):\n\n    def __init__(self, dataset, num_replicas, rank, shuffle, split_epoch=1, seed=0):\n        self.dataset = dataset\n        self.num_replicas = num_replicas\n        self.rank = rank\n        self.shuffle = shuffle\n        self.split_epoch = split_epoch\n        self.seed = seed\n\n        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        return download_target\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.misc.set_epoch","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.misc.set_epoch#L389-L390","kind":"function","name":"set_epoch","path":"imagebind_LLM/util/misc.py","language":"python","start_line":389,"end_line":390,"context_start_line":369,"context_end_line":410,"code":"        self.num_samples = len(dataset) // (num_replicas * split_epoch)\n\n    def __len__(self):\n        return self.num_samples\n\n    def __iter__(self):\n        if self.shuffle:\n            # deterministically shuffle based on epoch and seed\n            g = torch.Generator()\n            g.manual_seed(self.seed + self.epoch // self.split_epoch)\n            indices = torch.randperm(len(self.dataset), generator=g).tolist()  # type: ignore[arg-type]\n        else:\n            indices = list(range(len(self.dataset)))  # type: ignore[arg-type]\n\n        indices = indices[self.rank * self.split_epoch + self.epoch % self.split_epoch::self.num_replicas * self.split_epoch]\n        assert len(indices) >= self.num_samples\n        indices = indices[:self.num_samples]\n\n        return iter(indices)\n\n    def set_epoch(self, epoch):\n        self.epoch = epoch\n\ndef download(url: str, root: str):\n    os.makedirs(root, exist_ok=True)\n    filename = os.path.basename(url)\n    download_target = os.path.join(root, filename)\n\n    if os.path.exists(download_target) and not os.path.isfile(download_target):\n        raise RuntimeError(f\"{download_target} exists and is not a regular file\")\n\n    if os.path.isfile(download_target):\n        return download_target\n\n    with urllib.request.urlopen(url) as source, open(download_target, \"wb\") as output:\n        with tqdm(total=int(source.info().get(\"Content-Length\")), ncols=80, unit='iB', unit_scale=True, unit_divisor=1024) as loop:\n            while True:\n                buffer = source.read(8192)\n                if not buffer:\n                    break\n                output.write(buffer)\n                loop.update(len(buffer))","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.lr_sched","uri":"program://LLaMA-Adapter/module/imagebind_LLM.util.lr_sched#L1-L21","kind":"module","name":"imagebind_LLM.util.lr_sched","path":"imagebind_LLM/util/lr_sched.py","language":"python","start_line":1,"end_line":21,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs \n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * \\\n            (1. + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs)))\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"4ab5d5633bda0be9173ec91570bb3050326d942582ded2267702b53c3ac87c2c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.util.lr_sched.adjust_learning_rate","uri":"program://LLaMA-Adapter/function/imagebind_LLM.util.lr_sched.adjust_learning_rate#L9-L21","kind":"function","name":"adjust_learning_rate","path":"imagebind_LLM/util/lr_sched.py","language":"python","start_line":9,"end_line":21,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs \n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * \\\n            (1. + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs)))\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"4ab5d5633bda0be9173ec91570bb3050326d942582ded2267702b53c3ac87c2c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.data.dataset","uri":"program://LLaMA-Adapter/module/imagebind_LLM.data.dataset#L1-L156","kind":"module","name":"imagebind_LLM.data.dataset","path":"imagebind_LLM/data/dataset.py","language":"python","start_line":1,"end_line":156,"context_start_line":1,"context_end_line":156,"code":"import torch\nimport yaml\nfrom torch.utils.data import Dataset\nfrom PIL import Image\nimport json\nimport llama.utils\nfrom llama import Tokenizer\nimport copy\nimport torchvision.transforms as transforms\nimport pandas as pd\nimport random\n\ntry:\n    from torchvision.transforms import InterpolationMode\n    BICUBIC = InterpolationMode.BICUBIC\nexcept ImportError:\n    BICUBIC = Image.BICUBIC\n\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n# create data\ntransform_train = transforms.Compose([\n    transforms.RandomResizedCrop(size=(224, 224), scale=(0.9, 1.0), ratio=(0.75, 1.3333), interpolation=BICUBIC,\n                                 antialias=None),  # 3 is bicubic\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\nclass FinetuneDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        ann = []\n        for meta_path in self.config['META']:\n            meta_l = json.load(open(meta_path))\n            print(f\"{meta_path}: len {len(meta_l)}\")\n            ann += meta_l\n        self.ann = ann\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n        data_item = self.ann[index]\n        if 'image' in data_item.keys():\n            filename = data_item['image']\n            question = data_item['conversations'][0]['value']\n            answer = data_item['conversations'][1]['value']\n            # < fill path substitution logics here>\n            # filename = url.replace(\"/data0/data/coco/\", \"/mnt/petrelfs/leimeng/datasets/coco/\")\n\n            image = Image.open(filename).convert('RGB')\n            image = self.transform(image)\n            format_instruction = question\n            format_input = None\n        else:\n            image = torch.zeros(3, 224, 224)\n            format_instruction = data_item['instruction'],\n            format_input = data_item['input']\n            answer = data_item['output']\n        input1 = llama.utils.format_prompt(format_instruction, format_input)\n        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image\n\n\nclass PretrainDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        images, captions = [], []\n        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []\n        for x, y in zip(images, captions):\n            self.data_list.append({'url': x, 'caption': y})\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.data_list)\n\n    def __getitem__(self, index):\n        sample = self.data_list[index]\n        image_path, caption = sample['url'], sample['caption']\n        if isinstance(caption, list):\n            caption = random.choice(caption)\n        caption = str(caption)\n\n        image = Image.open(image_path).convert('RGB')\n        image = self.transform(image)\n\n        format_instruction = \"Generate caption of this image\"\n        input1 = llama.utils.format_prompt(format_instruction, None)\n        input2 = input1 + caption\n\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image","source_hash":"21dd21acbfa1466236388e074f1d48472af0435a1e92452970f870bcaa21eae6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.data.dataset.FinetuneDataset","uri":"program://LLaMA-Adapter/class/imagebind_LLM.data.dataset.FinetuneDataset#L40-L96","kind":"class","name":"FinetuneDataset","path":"imagebind_LLM/data/dataset.py","language":"python","start_line":40,"end_line":96,"context_start_line":20,"context_end_line":116,"code":"PROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\n\n# create data\ntransform_train = transforms.Compose([\n    transforms.RandomResizedCrop(size=(224, 224), scale=(0.9, 1.0), ratio=(0.75, 1.3333), interpolation=BICUBIC,\n                                 antialias=None),  # 3 is bicubic\n    transforms.ToTensor(),\n    transforms.Normalize(mean=[0.48145466, 0.4578275, 0.40821073], std=[0.26862954, 0.26130258, 0.27577711])])\n\nclass FinetuneDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        ann = []\n        for meta_path in self.config['META']:\n            meta_l = json.load(open(meta_path))\n            print(f\"{meta_path}: len {len(meta_l)}\")\n            ann += meta_l\n        self.ann = ann\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.ann)\n\n    def __getitem__(self, index):\n        data_item = self.ann[index]\n        if 'image' in data_item.keys():\n            filename = data_item['image']\n            question = data_item['conversations'][0]['value']\n            answer = data_item['conversations'][1]['value']\n            # < fill path substitution logics here>\n            # filename = url.replace(\"/data0/data/coco/\", \"/mnt/petrelfs/leimeng/datasets/coco/\")\n\n            image = Image.open(filename).convert('RGB')\n            image = self.transform(image)\n            format_instruction = question\n            format_input = None\n        else:\n            image = torch.zeros(3, 224, 224)\n            format_instruction = data_item['instruction'],\n            format_input = data_item['input']\n            answer = data_item['output']\n        input1 = llama.utils.format_prompt(format_instruction, format_input)\n        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image\n\n\nclass PretrainDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        images, captions = [], []\n        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []","source_hash":"21dd21acbfa1466236388e074f1d48472af0435a1e92452970f870bcaa21eae6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.data.dataset.PretrainDataset","uri":"program://LLaMA-Adapter/class/imagebind_LLM.data.dataset.PretrainDataset#L99-L156","kind":"class","name":"PretrainDataset","path":"imagebind_LLM/data/dataset.py","language":"python","start_line":99,"end_line":156,"context_start_line":79,"context_end_line":156,"code":"        input1 = llama.utils.format_prompt(format_instruction, format_input)\n        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image\n\n\nclass PretrainDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        images, captions = [], []\n        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []\n        for x, y in zip(images, captions):\n            self.data_list.append({'url': x, 'caption': y})\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.data_list)\n\n    def __getitem__(self, index):\n        sample = self.data_list[index]\n        image_path, caption = sample['url'], sample['caption']\n        if isinstance(caption, list):\n            caption = random.choice(caption)\n        caption = str(caption)\n\n        image = Image.open(image_path).convert('RGB')\n        image = self.transform(image)\n\n        format_instruction = \"Generate caption of this image\"\n        input1 = llama.utils.format_prompt(format_instruction, None)\n        input2 = input1 + caption\n\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image","source_hash":"21dd21acbfa1466236388e074f1d48472af0435a1e92452970f870bcaa21eae6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.data.dataset.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.data.dataset.__init__#L100-L122","kind":"function","name":"__init__","path":"imagebind_LLM/data/dataset.py","language":"python","start_line":100,"end_line":122,"context_start_line":80,"context_end_line":142,"code":"        input2 = input1 + answer\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image\n\n\nclass PretrainDataset(Dataset):\n    def __init__(self, config_path, transform, max_words=30, tokenizer_path=None):\n        print(f\"read dataset config from {config_path}\")\n        with open(config_path, 'r') as f:\n            self.config = yaml.load(f, Loader=yaml.FullLoader)\n        print(\"DATASET CONFIG:\")\n        print(self.config)\n        images, captions = [], []\n        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []\n        for x, y in zip(images, captions):\n            self.data_list.append({'url': x, 'caption': y})\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.data_list)\n\n    def __getitem__(self, index):\n        sample = self.data_list[index]\n        image_path, caption = sample['url'], sample['caption']\n        if isinstance(caption, list):\n            caption = random.choice(caption)\n        caption = str(caption)\n\n        image = Image.open(image_path).convert('RGB')\n        image = self.transform(image)\n\n        format_instruction = \"Generate caption of this image\"\n        input1 = llama.utils.format_prompt(format_instruction, None)\n        input2 = input1 + caption\n\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)","source_hash":"21dd21acbfa1466236388e074f1d48472af0435a1e92452970f870bcaa21eae6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.data.dataset.__len__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.data.dataset.__len__#L124-L125","kind":"function","name":"__len__","path":"imagebind_LLM/data/dataset.py","language":"python","start_line":124,"end_line":125,"context_start_line":104,"context_end_line":145,"code":"        print(\"DATASET CONFIG:\")\n        print(self.config)\n        images, captions = [], []\n        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []\n        for x, y in zip(images, captions):\n            self.data_list.append({'url': x, 'caption': y})\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.data_list)\n\n    def __getitem__(self, index):\n        sample = self.data_list[index]\n        image_path, caption = sample['url'], sample['caption']\n        if isinstance(caption, list):\n            caption = random.choice(caption)\n        caption = str(caption)\n\n        image = Image.open(image_path).convert('RGB')\n        image = self.transform(image)\n\n        format_instruction = \"Generate caption of this image\"\n        input1 = llama.utils.format_prompt(format_instruction, None)\n        input2 = input1 + caption\n\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))","source_hash":"21dd21acbfa1466236388e074f1d48472af0435a1e92452970f870bcaa21eae6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.data.dataset.__getitem__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.data.dataset.__getitem__#L127-L156","kind":"function","name":"__getitem__","path":"imagebind_LLM/data/dataset.py","language":"python","start_line":127,"end_line":156,"context_start_line":107,"context_end_line":156,"code":"        for meta_path in self.config['META']:\n            images_this_meta, captions_this_meta = [], []\n            for chunk in pd.read_csv(meta_path, sep='\\t', lineterminator='\\n', chunksize=10 ** 6):\n                images_this_meta.extend(chunk['url'].tolist())\n                captions_this_meta.extend(chunk['caption'].tolist())\n            print(f\"{meta_path}: len {len(images_this_meta)}\")\n            images.extend(images_this_meta)\n            captions.extend(captions_this_meta)\n\n        self.data_list = []\n        for x, y in zip(images, captions):\n            self.data_list.append({'url': x, 'caption': y})\n        print(f\"total length: {len(self)}\")\n        self.transform = transform\n        self.max_words = max_words\n        self.tokenizer = Tokenizer(model_path=tokenizer_path)\n\n    def __len__(self):\n        return len(self.data_list)\n\n    def __getitem__(self, index):\n        sample = self.data_list[index]\n        image_path, caption = sample['url'], sample['caption']\n        if isinstance(caption, list):\n            caption = random.choice(caption)\n        caption = str(caption)\n\n        image = Image.open(image_path).convert('RGB')\n        image = self.transform(image)\n\n        format_instruction = \"Generate caption of this image\"\n        input1 = llama.utils.format_prompt(format_instruction, None)\n        input2 = input1 + caption\n\n        input1 = torch.tensor(self.tokenizer.encode(input1, bos=True, eos=False), dtype=torch.int64)\n        input2 = torch.tensor(self.tokenizer.encode(input2, bos=True, eos=True), dtype=torch.int64)\n        padding = self.max_words - input2.shape[0]\n        if padding > 0:\n            input2 = torch.cat((input2, torch.zeros(padding, dtype=torch.int64) - 1))\n        elif padding < 0:\n            input2 = input2[:self.max_words]\n        labels = copy.deepcopy(input2)\n        labels[:len(input1)] = -1\n        input2_mask = input2.ge(0)\n        label_mask = labels.ge(0)\n        input2[~input2_mask] = 0\n        labels[~label_mask] = 0\n        input2_mask = input2_mask.float()\n        label_mask = label_mask.float()\n        return input2, labels, input2_mask, image","source_hash":"21dd21acbfa1466236388e074f1d48472af0435a1e92452970f870bcaa21eae6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.demo","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.demo#L1-L55","kind":"module","name":"imagebind_LLM.ImageBind.demo","path":"imagebind_LLM/ImageBind/demo.py","language":"python","start_line":1,"end_line":55,"context_start_line":1,"context_end_line":55,"code":"import data\nimport torch\nfrom models import imagebind_model\nfrom models.imagebind_model import ModalityType\n\ntext_list=[\"A dog.\", \"A car\", \"A bird\"]\nimage_paths=[\".assets/dog_image.jpg\", \".assets/car_image.jpg\", \".assets/bird_image.jpg\"]\naudio_paths=[\".assets/dog_audio.wav\", \".assets/car_audio.wav\", \".assets/bird_audio.wav\"]\n\ndevice = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n\n# Instantiate model\nmodel = imagebind_model.imagebind_huge(pretrained=True)\nmodel.eval()\nmodel.to(device)\n\n# Load data\ninputs = {\n    ModalityType.TEXT: data.load_and_transform_text(text_list, device),\n    ModalityType.VISION: data.load_and_transform_vision_data(image_paths, device),\n    ModalityType.AUDIO: data.load_and_transform_audio_data(audio_paths, device),\n}\n\nwith torch.no_grad():\n    embeddings = model(inputs)\n\nprint(\n    \"Vision x Text: \",\n    torch.softmax(embeddings[ModalityType.VISION] @ embeddings[ModalityType.TEXT].T, dim=-1),\n)\nprint(\n    \"Audio x Text: \",\n    torch.softmax(embeddings[ModalityType.AUDIO] @ embeddings[ModalityType.TEXT].T, dim=-1),\n)\nprint(\n    \"Vision x Audio: \",\n    torch.softmax(embeddings[ModalityType.VISION] @ embeddings[ModalityType.AUDIO].T, dim=-1),\n)\n\n# Expected output:\n#\n# Vision x Text:\n# tensor([[9.9761e-01, 2.3694e-03, 1.8612e-05],\n#         [3.3836e-05, 9.9994e-01, 2.4118e-05],\n#         [4.7997e-05, 1.3496e-02, 9.8646e-01]])\n#\n# Audio x Text:\n# tensor([[1., 0., 0.],\n#         [0., 1., 0.],\n#         [0., 0., 1.]])\n#\n# Vision x Audio:\n# tensor([[0.8070, 0.1088, 0.0842],\n#         [0.1036, 0.7884, 0.1079],\n#         [0.0018, 0.0022, 0.9960]])","source_hash":"6164353bb6650d6e39c854828cd571d1ec6f55a9eaf83fd62c423394ef2f1375","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.data#L1-L371","kind":"module","name":"imagebind_LLM.ImageBind.data","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":1,"end_line":371,"context_start_line":1,"context_end_line":371,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport logging\nimport math\n\nimport torch\nimport torch.nn as nn\nimport torchaudio\nfrom PIL import Image\nimport open3d as o3d\nimport numpy\nfrom pytorchvideo import transforms as pv_transforms\nfrom pytorchvideo.data.clip_sampling import ConstantClipsPerVideoSampler\nfrom pytorchvideo.data.encoded_video import EncodedVideo\nfrom torchvision import transforms\nfrom torchvision.transforms._transforms_video import NormalizeVideo\n\nfrom .models.multimodal_preprocessors import SimpleTokenizer\n\nDEFAULT_AUDIO_FRAME_SHIFT_MS = 10  # in milliseconds\n\nBPE_PATH = \"ImageBind/bpe/bpe_simple_vocab_16e6.txt.gz\"\n\n\ndef waveform2melspec(waveform, sample_rate, num_mel_bins, target_length):\n    # Based on https://github.com/YuanGongND/ast/blob/d7d8b4b8e06cdaeb6c843cdb38794c1c7692234c/src/dataloader.py#L102\n    waveform -= waveform.mean()\n    fbank = torchaudio.compliance.kaldi.fbank(\n        waveform,\n        htk_compat=True,\n        sample_frequency=sample_rate,\n        use_energy=False,\n        window_type=\"hanning\",\n        num_mel_bins=num_mel_bins,\n        dither=0.0,\n        frame_length=25,\n        frame_shift=DEFAULT_AUDIO_FRAME_SHIFT_MS,\n    )\n    # Convert to [mel_bins, num_frames] shape\n    fbank = fbank.transpose(0, 1)\n    # Pad to target_length\n    n_frames = fbank.size(1)\n    p = target_length - n_frames\n    # if p is too large (say >20%), flash a warning\n    if abs(p) / n_frames > 0.2:\n        logging.warning(\n            \"Large gap between audio n_frames(%d) and \"\n            \"target_length (%d). Is the audio_target_length \"\n            \"setting correct?\",\n            n_frames,\n            target_length,\n        )\n    # cut and pad\n    if p > 0:\n        fbank = torch.nn.functional.pad(fbank, (0, p), mode=\"constant\", value=0)\n    elif p < 0:\n        fbank = fbank[:, 0:target_length]\n    # Convert to [1, mel_bins, num_frames] shape, essentially like a 1\n    # channel image\n    fbank = fbank.unsqueeze(0)\n    return fbank\n\n\ndef get_clip_timepoints(clip_sampler, duration):\n    # Read out all clips in this video\n    all_clips_timepoints = []\n    is_last_clip = False\n    end = 0.0\n    while not is_last_clip:\n        start, end, _, _, is_last_clip = clip_sampler(end, duration, annotation=None)\n        all_clips_timepoints.append((start, end))\n    return all_clips_timepoints\n\n\ndef load_and_transform_vision_data(image_paths, device):\n    if image_paths is None:\n        return None\n\n    image_ouputs = []\n    for image_path in image_paths:\n        data_transform = transforms.Compose(\n            [\n                transforms.Resize(\n                    224, interpolation=transforms.InterpolationMode.BICUBIC\n                ),\n                transforms.CenterCrop(224),\n                transforms.ToTensor(),\n                transforms.Normalize(\n                    mean=(0.48145466, 0.4578275, 0.40821073),\n                    std=(0.26862954, 0.26130258, 0.27577711),\n                ),\n            ]\n        )\n        with open(image_path, \"rb\") as fopen:\n            image = Image.open(fopen).convert(\"RGB\")\n\n        image = data_transform(image).to(device)\n        image_ouputs.append(image)\n    return torch.stack(image_ouputs, dim=0)\n\n\ndef load_and_transform_text(text, device):\n    if text is None:\n        return None\n    tokenizer = SimpleTokenizer(bpe_path=BPE_PATH)\n    tokens = [tokenizer(t).unsqueeze(0).to(device) for t in text]\n    tokens = torch.cat(tokens, dim=0)\n    return tokens\n\n\ndef load_and_transform_audio_data(\n    audio_paths,\n    device,\n    num_mel_bins=128,\n    target_length=204,\n    sample_rate=16000,\n    clip_duration=2,\n    clips_per_video=3,\n    mean=-4.268,\n    std=9.138,\n):\n    if audio_paths is None:\n        return None\n\n    audio_outputs = []\n    clip_sampler = ConstantClipsPerVideoSampler(\n        clip_duration=clip_duration, clips_per_video=clips_per_video\n    )\n\n    for audio_path in audio_paths:\n        waveform, sr = torchaudio.load(audio_path)\n        if sample_rate != sr:\n            waveform = torchaudio.functional.resample(\n                waveform, orig_freq=sr, new_freq=sample_rate\n            )\n        all_clips_timepoints = get_clip_timepoints(\n            clip_sampler, waveform.size(1) / sample_rate\n        )\n        all_clips = []\n        for clip_timepoints in all_clips_timepoints:\n            waveform_clip = waveform[\n                :,\n                int(clip_timepoints[0] * sample_rate) : int(\n                    clip_timepoints[1] * sample_rate\n                ),\n            ]\n            waveform_melspec = waveform2melspec(\n                waveform_clip, sample_rate, num_mel_bins, target_length\n            )\n            all_clips.append(waveform_melspec)\n\n        normalize = transforms.Normalize(mean=mean, std=std)\n        all_clips = [normalize(ac).to(device) for ac in all_clips]\n\n        all_clips = torch.stack(all_clips, dim=0)\n        audio_outputs.append(all_clips)\n\n    return torch.stack(audio_outputs, dim=0)\n\ndef load_and_transform_point_cloud_data(point_paths, device):\n    point_outputs = []\n\n    for point_path in point_paths:\n        if type(point_path) != str:\n            point_path = point_path.name\n        file_name = point_path\n        if '.ply' in file_name or '.pts' in file_name or '.pcd' in file_name or '.xyz' in file_name:\n            pcd = o3d.io.read_point_cloud(file_name)\n            point = numpy.asarray(pcd.points)\n        elif '.pt' in file_name:\n            point = torch.load(file_name).numpy()\n        elif '.npy' in file_name:\n            point = numpy.load(file_name)\n        elif '.obj' in file_name:\n            points = []\n            with open(file_name, 'r') as file:\n                for line in file:\n                    line = line.strip()\n                    if line.startswith('v '):\n                        data = line.split(' ')\n                        x = float(data[1])\n                        y = float(data[2])\n                        z = float(data[3])\n                        points.append((x, y, z))\n            point = numpy.asarray(points)\n        point = torch.tensor(point).to(device)\n        # point = point.type(torch.cuda.HalfTensor)\n        point_outputs.append(point)\n        \n    return torch.stack(point_outputs, dim=0)\n\ndef crop_boxes(boxes, x_offset, y_offset):\n    \"\"\"\n    Perform crop on the bounding boxes given the offsets.\n    Args:\n        boxes (ndarray or None): bounding boxes to perform crop. The dimension\n            is `num boxes` x 4.\n        x_offset (int): cropping offset in the x axis.\n        y_offset (int): cropping offset in the y axis.\n    Returns:\n        cropped_boxes (ndarray or None): the cropped boxes with dimension of\n            `num boxes` x 4.\n    \"\"\"\n    cropped_boxes = boxes.copy()\n    cropped_boxes[:, [0, 2]] = boxes[:, [0, 2]] - x_offset\n    cropped_boxes[:, [1, 3]] = boxes[:, [1, 3]] - y_offset\n\n    return cropped_boxes\n\n\ndef uniform_crop(images, size, spatial_idx, boxes=None, scale_size=None):\n    \"\"\"\n    Perform uniform spatial sampling on the images and corresponding boxes.\n    Args:\n        images (tensor): images to perform uniform crop. The dimension is\n            `num frames` x `channel` x `height` x `width`.\n        size (int): size of height and weight to crop the images.\n        spatial_idx (int): 0, 1, or 2 for left, center, and right crop if width\n            is larger than height. Or 0, 1, or 2 for top, center, and bottom\n            crop if height is larger than width.\n        boxes (ndarray or None): optional. Corresponding boxes to images.\n            Dimension is `num boxes` x 4.\n        scale_size (int): optinal. If not None, resize the images to scale_size before\n            performing any crop.\n    Returns:\n        cropped (tensor): images with dimension of\n            `num frames` x `channel` x `size` x `size`.\n        cropped_boxes (ndarray or None): the cropped boxes with dimension of\n            `num boxes` x 4.\n    \"\"\"\n    assert spatial_idx in [0, 1, 2]\n    ndim = len(images.shape)\n    if ndim == 3:\n        images = images.unsqueeze(0)\n    height = images.shape[2]\n    width = images.shape[3]\n\n    if scale_size is not None:\n        if width <= height:\n            width, height = scale_size, int(height / width * scale_size)\n        else:\n            width, height = int(width / height * scale_size), scale_size\n        images = torch.nn.functional.interpolate(\n            images,\n            size=(height, width),\n            mode=\"bilinear\",\n            align_corners=False,\n        )\n\n    y_offset = int(math.ceil((height - size) / 2))\n    x_offset = int(math.ceil((width - size) / 2))\n\n    if height > width:\n        if spatial_idx == 0:\n            y_offset = 0\n        elif spatial_idx == 2:\n            y_offset = height - size\n    else:\n        if spatial_idx == 0:\n            x_offset = 0\n        elif spatial_idx == 2:\n            x_offset = width - size\n    cropped = images[:, :, y_offset : y_offset + size, x_offset : x_offset + size]\n    cropped_boxes = crop_boxes(boxes, x_offset, y_offset) if boxes is not None else None\n    if ndim == 3:\n        cropped = cropped.squeeze(0)\n    return cropped, cropped_boxes\n\n\nclass SpatialCrop(nn.Module):\n    \"\"\"\n    Convert the video into 3 smaller clips spatially. Must be used after the\n        temporal crops to get spatial crops, and should be used with\n        -2 in the spatial crop at the slowfast augmentation stage (so full\n        frames are passed in here). Will return a larger list with the\n        3x spatial crops as well.\n    \"\"\"\n\n    def __init__(self, crop_size: int = 224, num_crops: int = 3):\n        super().__init__()\n        self.crop_size = crop_size\n        if num_crops == 3:\n            self.crops_to_ext = [0, 1, 2]\n            self.flipped_crops_to_ext = []\n        elif num_crops == 1:\n            self.crops_to_ext = [1]\n            self.flipped_crops_to_ext = []\n        else:\n            raise NotImplementedError(\"Nothing else supported yet\")\n\n    def forward(self, videos):\n        \"\"\"\n        Args:\n            videos: A list of C, T, H, W videos.\n        Returns:\n            videos: A list with 3x the number of elements. Each video converted\n                to C, T, H', W' by spatial cropping.\n        \"\"\"\n        assert isinstance(videos, list), \"Must be a list of videos after temporal crops\"\n        assert all([video.ndim == 4 for video in videos]), \"Must be (C,T,H,W)\"\n        res = []\n        for video in videos:\n            for spatial_idx in self.crops_to_ext:\n                res.append(uniform_crop(video, self.crop_size, spatial_idx)[0])\n            if not self.flipped_crops_to_ext:\n                continue\n            flipped_video = transforms.functional.hflip(video)\n            for spatial_idx in self.flipped_crops_to_ext:\n                res.append(uniform_crop(flipped_video, self.crop_size, spatial_idx)[0])\n        return res\n\n\ndef load_and_transform_video_data(\n    video_paths,\n    device,\n    clip_duration=2,\n    clips_per_video=5,\n    sample_rate=16000,\n):\n    if video_paths is None:\n        return None\n\n    video_outputs = []\n    video_transform = transforms.Compose(\n        [\n            pv_transforms.ShortSideScale(224),\n            NormalizeVideo(\n                mean=(0.48145466, 0.4578275, 0.40821073),\n                std=(0.26862954, 0.26130258, 0.27577711),\n            ),\n        ]\n    )\n\n    clip_sampler = ConstantClipsPerVideoSampler(\n        clip_duration=clip_duration, clips_per_video=clips_per_video\n    )\n    frame_sampler = pv_transforms.UniformTemporalSubsample(num_samples=clip_duration)\n\n    for video_path in video_paths:\n        video = EncodedVideo.from_path(\n            video_path,\n            decoder=\"pyav\",\n            decode_audio=False,\n            # **{\"sample_rate\": sample_rate},\n        )\n\n        all_clips_timepoints = get_clip_timepoints(clip_sampler, video.duration)\n\n        all_video = []\n        for clip_timepoints in all_clips_timepoints:\n            # Read the clip, get frames\n            clip = video.get_clip(clip_timepoints[0], clip_timepoints[1])\n            if clip is None:\n                raise ValueError(\"No clip found\")\n            video_clip = frame_sampler(clip[\"video\"])\n            video_clip = video_clip / 255.0  # since this is float, need 0-1\n\n            all_video.append(video_clip)\n\n        all_video = [video_transform(clip) for clip in all_video]\n        all_video = SpatialCrop(224, num_crops=3)(all_video)\n\n        all_video = torch.stack(all_video, dim=0)\n        video_outputs.append(all_video)\n\n    return torch.stack(video_outputs, dim=0).to(device)","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.waveform2melspec","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.waveform2melspec#L30-L66","kind":"function","name":"waveform2melspec","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":30,"end_line":66,"context_start_line":10,"context_end_line":86,"code":"\nimport torch\nimport torch.nn as nn\nimport torchaudio\nfrom PIL import Image\nimport open3d as o3d\nimport numpy\nfrom pytorchvideo import transforms as pv_transforms\nfrom pytorchvideo.data.clip_sampling import ConstantClipsPerVideoSampler\nfrom pytorchvideo.data.encoded_video import EncodedVideo\nfrom torchvision import transforms\nfrom torchvision.transforms._transforms_video import NormalizeVideo\n\nfrom .models.multimodal_preprocessors import SimpleTokenizer\n\nDEFAULT_AUDIO_FRAME_SHIFT_MS = 10  # in milliseconds\n\nBPE_PATH = \"ImageBind/bpe/bpe_simple_vocab_16e6.txt.gz\"\n\n\ndef waveform2melspec(waveform, sample_rate, num_mel_bins, target_length):\n    # Based on https://github.com/YuanGongND/ast/blob/d7d8b4b8e06cdaeb6c843cdb38794c1c7692234c/src/dataloader.py#L102\n    waveform -= waveform.mean()\n    fbank = torchaudio.compliance.kaldi.fbank(\n        waveform,\n        htk_compat=True,\n        sample_frequency=sample_rate,\n        use_energy=False,\n        window_type=\"hanning\",\n        num_mel_bins=num_mel_bins,\n        dither=0.0,\n        frame_length=25,\n        frame_shift=DEFAULT_AUDIO_FRAME_SHIFT_MS,\n    )\n    # Convert to [mel_bins, num_frames] shape\n    fbank = fbank.transpose(0, 1)\n    # Pad to target_length\n    n_frames = fbank.size(1)\n    p = target_length - n_frames\n    # if p is too large (say >20%), flash a warning\n    if abs(p) / n_frames > 0.2:\n        logging.warning(\n            \"Large gap between audio n_frames(%d) and \"\n            \"target_length (%d). Is the audio_target_length \"\n            \"setting correct?\",\n            n_frames,\n            target_length,\n        )\n    # cut and pad\n    if p > 0:\n        fbank = torch.nn.functional.pad(fbank, (0, p), mode=\"constant\", value=0)\n    elif p < 0:\n        fbank = fbank[:, 0:target_length]\n    # Convert to [1, mel_bins, num_frames] shape, essentially like a 1\n    # channel image\n    fbank = fbank.unsqueeze(0)\n    return fbank\n\n\ndef get_clip_timepoints(clip_sampler, duration):\n    # Read out all clips in this video\n    all_clips_timepoints = []\n    is_last_clip = False\n    end = 0.0\n    while not is_last_clip:\n        start, end, _, _, is_last_clip = clip_sampler(end, duration, annotation=None)\n        all_clips_timepoints.append((start, end))\n    return all_clips_timepoints\n\n\ndef load_and_transform_vision_data(image_paths, device):\n    if image_paths is None:\n        return None\n\n    image_ouputs = []\n    for image_path in image_paths:\n        data_transform = transforms.Compose(","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.get_clip_timepoints","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.get_clip_timepoints#L69-L77","kind":"function","name":"get_clip_timepoints","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":69,"end_line":77,"context_start_line":49,"context_end_line":97,"code":"    # if p is too large (say >20%), flash a warning\n    if abs(p) / n_frames > 0.2:\n        logging.warning(\n            \"Large gap between audio n_frames(%d) and \"\n            \"target_length (%d). Is the audio_target_length \"\n            \"setting correct?\",\n            n_frames,\n            target_length,\n        )\n    # cut and pad\n    if p > 0:\n        fbank = torch.nn.functional.pad(fbank, (0, p), mode=\"constant\", value=0)\n    elif p < 0:\n        fbank = fbank[:, 0:target_length]\n    # Convert to [1, mel_bins, num_frames] shape, essentially like a 1\n    # channel image\n    fbank = fbank.unsqueeze(0)\n    return fbank\n\n\ndef get_clip_timepoints(clip_sampler, duration):\n    # Read out all clips in this video\n    all_clips_timepoints = []\n    is_last_clip = False\n    end = 0.0\n    while not is_last_clip:\n        start, end, _, _, is_last_clip = clip_sampler(end, duration, annotation=None)\n        all_clips_timepoints.append((start, end))\n    return all_clips_timepoints\n\n\ndef load_and_transform_vision_data(image_paths, device):\n    if image_paths is None:\n        return None\n\n    image_ouputs = []\n    for image_path in image_paths:\n        data_transform = transforms.Compose(\n            [\n                transforms.Resize(\n                    224, interpolation=transforms.InterpolationMode.BICUBIC\n                ),\n                transforms.CenterCrop(224),\n                transforms.ToTensor(),\n                transforms.Normalize(\n                    mean=(0.48145466, 0.4578275, 0.40821073),\n                    std=(0.26862954, 0.26130258, 0.27577711),\n                ),\n            ]","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.load_and_transform_vision_data","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.load_and_transform_vision_data#L80-L104","kind":"function","name":"load_and_transform_vision_data","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":80,"end_line":104,"context_start_line":60,"context_end_line":124,"code":"        fbank = torch.nn.functional.pad(fbank, (0, p), mode=\"constant\", value=0)\n    elif p < 0:\n        fbank = fbank[:, 0:target_length]\n    # Convert to [1, mel_bins, num_frames] shape, essentially like a 1\n    # channel image\n    fbank = fbank.unsqueeze(0)\n    return fbank\n\n\ndef get_clip_timepoints(clip_sampler, duration):\n    # Read out all clips in this video\n    all_clips_timepoints = []\n    is_last_clip = False\n    end = 0.0\n    while not is_last_clip:\n        start, end, _, _, is_last_clip = clip_sampler(end, duration, annotation=None)\n        all_clips_timepoints.append((start, end))\n    return all_clips_timepoints\n\n\ndef load_and_transform_vision_data(image_paths, device):\n    if image_paths is None:\n        return None\n\n    image_ouputs = []\n    for image_path in image_paths:\n        data_transform = transforms.Compose(\n            [\n                transforms.Resize(\n                    224, interpolation=transforms.InterpolationMode.BICUBIC\n                ),\n                transforms.CenterCrop(224),\n                transforms.ToTensor(),\n                transforms.Normalize(\n                    mean=(0.48145466, 0.4578275, 0.40821073),\n                    std=(0.26862954, 0.26130258, 0.27577711),\n                ),\n            ]\n        )\n        with open(image_path, \"rb\") as fopen:\n            image = Image.open(fopen).convert(\"RGB\")\n\n        image = data_transform(image).to(device)\n        image_ouputs.append(image)\n    return torch.stack(image_ouputs, dim=0)\n\n\ndef load_and_transform_text(text, device):\n    if text is None:\n        return None\n    tokenizer = SimpleTokenizer(bpe_path=BPE_PATH)\n    tokens = [tokenizer(t).unsqueeze(0).to(device) for t in text]\n    tokens = torch.cat(tokens, dim=0)\n    return tokens\n\n\ndef load_and_transform_audio_data(\n    audio_paths,\n    device,\n    num_mel_bins=128,\n    target_length=204,\n    sample_rate=16000,\n    clip_duration=2,\n    clips_per_video=3,\n    mean=-4.268,","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.load_and_transform_text","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.load_and_transform_text#L107-L113","kind":"function","name":"load_and_transform_text","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":107,"end_line":113,"context_start_line":87,"context_end_line":133,"code":"            [\n                transforms.Resize(\n                    224, interpolation=transforms.InterpolationMode.BICUBIC\n                ),\n                transforms.CenterCrop(224),\n                transforms.ToTensor(),\n                transforms.Normalize(\n                    mean=(0.48145466, 0.4578275, 0.40821073),\n                    std=(0.26862954, 0.26130258, 0.27577711),\n                ),\n            ]\n        )\n        with open(image_path, \"rb\") as fopen:\n            image = Image.open(fopen).convert(\"RGB\")\n\n        image = data_transform(image).to(device)\n        image_ouputs.append(image)\n    return torch.stack(image_ouputs, dim=0)\n\n\ndef load_and_transform_text(text, device):\n    if text is None:\n        return None\n    tokenizer = SimpleTokenizer(bpe_path=BPE_PATH)\n    tokens = [tokenizer(t).unsqueeze(0).to(device) for t in text]\n    tokens = torch.cat(tokens, dim=0)\n    return tokens\n\n\ndef load_and_transform_audio_data(\n    audio_paths,\n    device,\n    num_mel_bins=128,\n    target_length=204,\n    sample_rate=16000,\n    clip_duration=2,\n    clips_per_video=3,\n    mean=-4.268,\n    std=9.138,\n):\n    if audio_paths is None:\n        return None\n\n    audio_outputs = []\n    clip_sampler = ConstantClipsPerVideoSampler(\n        clip_duration=clip_duration, clips_per_video=clips_per_video\n    )","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.load_and_transform_audio_data","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.load_and_transform_audio_data#L116-L163","kind":"function","name":"load_and_transform_audio_data","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":116,"end_line":163,"context_start_line":96,"context_end_line":183,"code":"                ),\n            ]\n        )\n        with open(image_path, \"rb\") as fopen:\n            image = Image.open(fopen).convert(\"RGB\")\n\n        image = data_transform(image).to(device)\n        image_ouputs.append(image)\n    return torch.stack(image_ouputs, dim=0)\n\n\ndef load_and_transform_text(text, device):\n    if text is None:\n        return None\n    tokenizer = SimpleTokenizer(bpe_path=BPE_PATH)\n    tokens = [tokenizer(t).unsqueeze(0).to(device) for t in text]\n    tokens = torch.cat(tokens, dim=0)\n    return tokens\n\n\ndef load_and_transform_audio_data(\n    audio_paths,\n    device,\n    num_mel_bins=128,\n    target_length=204,\n    sample_rate=16000,\n    clip_duration=2,\n    clips_per_video=3,\n    mean=-4.268,\n    std=9.138,\n):\n    if audio_paths is None:\n        return None\n\n    audio_outputs = []\n    clip_sampler = ConstantClipsPerVideoSampler(\n        clip_duration=clip_duration, clips_per_video=clips_per_video\n    )\n\n    for audio_path in audio_paths:\n        waveform, sr = torchaudio.load(audio_path)\n        if sample_rate != sr:\n            waveform = torchaudio.functional.resample(\n                waveform, orig_freq=sr, new_freq=sample_rate\n            )\n        all_clips_timepoints = get_clip_timepoints(\n            clip_sampler, waveform.size(1) / sample_rate\n        )\n        all_clips = []\n        for clip_timepoints in all_clips_timepoints:\n            waveform_clip = waveform[\n                :,\n                int(clip_timepoints[0] * sample_rate) : int(\n                    clip_timepoints[1] * sample_rate\n                ),\n            ]\n            waveform_melspec = waveform2melspec(\n                waveform_clip, sample_rate, num_mel_bins, target_length\n            )\n            all_clips.append(waveform_melspec)\n\n        normalize = transforms.Normalize(mean=mean, std=std)\n        all_clips = [normalize(ac).to(device) for ac in all_clips]\n\n        all_clips = torch.stack(all_clips, dim=0)\n        audio_outputs.append(all_clips)\n\n    return torch.stack(audio_outputs, dim=0)\n\ndef load_and_transform_point_cloud_data(point_paths, device):\n    point_outputs = []\n\n    for point_path in point_paths:\n        if type(point_path) != str:\n            point_path = point_path.name\n        file_name = point_path\n        if '.ply' in file_name or '.pts' in file_name or '.pcd' in file_name or '.xyz' in file_name:\n            pcd = o3d.io.read_point_cloud(file_name)\n            point = numpy.asarray(pcd.points)\n        elif '.pt' in file_name:\n            point = torch.load(file_name).numpy()\n        elif '.npy' in file_name:\n            point = numpy.load(file_name)\n        elif '.obj' in file_name:\n            points = []\n            with open(file_name, 'r') as file:\n                for line in file:\n                    line = line.strip()","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.load_and_transform_point_cloud_data","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.load_and_transform_point_cloud_data#L165-L195","kind":"function","name":"load_and_transform_point_cloud_data","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":165,"end_line":195,"context_start_line":145,"context_end_line":215,"code":"        for clip_timepoints in all_clips_timepoints:\n            waveform_clip = waveform[\n                :,\n                int(clip_timepoints[0] * sample_rate) : int(\n                    clip_timepoints[1] * sample_rate\n                ),\n            ]\n            waveform_melspec = waveform2melspec(\n                waveform_clip, sample_rate, num_mel_bins, target_length\n            )\n            all_clips.append(waveform_melspec)\n\n        normalize = transforms.Normalize(mean=mean, std=std)\n        all_clips = [normalize(ac).to(device) for ac in all_clips]\n\n        all_clips = torch.stack(all_clips, dim=0)\n        audio_outputs.append(all_clips)\n\n    return torch.stack(audio_outputs, dim=0)\n\ndef load_and_transform_point_cloud_data(point_paths, device):\n    point_outputs = []\n\n    for point_path in point_paths:\n        if type(point_path) != str:\n            point_path = point_path.name\n        file_name = point_path\n        if '.ply' in file_name or '.pts' in file_name or '.pcd' in file_name or '.xyz' in file_name:\n            pcd = o3d.io.read_point_cloud(file_name)\n            point = numpy.asarray(pcd.points)\n        elif '.pt' in file_name:\n            point = torch.load(file_name).numpy()\n        elif '.npy' in file_name:\n            point = numpy.load(file_name)\n        elif '.obj' in file_name:\n            points = []\n            with open(file_name, 'r') as file:\n                for line in file:\n                    line = line.strip()\n                    if line.startswith('v '):\n                        data = line.split(' ')\n                        x = float(data[1])\n                        y = float(data[2])\n                        z = float(data[3])\n                        points.append((x, y, z))\n            point = numpy.asarray(points)\n        point = torch.tensor(point).to(device)\n        # point = point.type(torch.cuda.HalfTensor)\n        point_outputs.append(point)\n        \n    return torch.stack(point_outputs, dim=0)\n\ndef crop_boxes(boxes, x_offset, y_offset):\n    \"\"\"\n    Perform crop on the bounding boxes given the offsets.\n    Args:\n        boxes (ndarray or None): bounding boxes to perform crop. The dimension\n            is `num boxes` x 4.\n        x_offset (int): cropping offset in the x axis.\n        y_offset (int): cropping offset in the y axis.\n    Returns:\n        cropped_boxes (ndarray or None): the cropped boxes with dimension of\n            `num boxes` x 4.\n    \"\"\"\n    cropped_boxes = boxes.copy()\n    cropped_boxes[:, [0, 2]] = boxes[:, [0, 2]] - x_offset\n    cropped_boxes[:, [1, 3]] = boxes[:, [1, 3]] - y_offset\n\n    return cropped_boxes\n\n","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.crop_boxes","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.crop_boxes#L197-L213","kind":"function","name":"crop_boxes","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":197,"end_line":213,"context_start_line":177,"context_end_line":233,"code":"        elif '.npy' in file_name:\n            point = numpy.load(file_name)\n        elif '.obj' in file_name:\n            points = []\n            with open(file_name, 'r') as file:\n                for line in file:\n                    line = line.strip()\n                    if line.startswith('v '):\n                        data = line.split(' ')\n                        x = float(data[1])\n                        y = float(data[2])\n                        z = float(data[3])\n                        points.append((x, y, z))\n            point = numpy.asarray(points)\n        point = torch.tensor(point).to(device)\n        # point = point.type(torch.cuda.HalfTensor)\n        point_outputs.append(point)\n        \n    return torch.stack(point_outputs, dim=0)\n\ndef crop_boxes(boxes, x_offset, y_offset):\n    \"\"\"\n    Perform crop on the bounding boxes given the offsets.\n    Args:\n        boxes (ndarray or None): bounding boxes to perform crop. The dimension\n            is `num boxes` x 4.\n        x_offset (int): cropping offset in the x axis.\n        y_offset (int): cropping offset in the y axis.\n    Returns:\n        cropped_boxes (ndarray or None): the cropped boxes with dimension of\n            `num boxes` x 4.\n    \"\"\"\n    cropped_boxes = boxes.copy()\n    cropped_boxes[:, [0, 2]] = boxes[:, [0, 2]] - x_offset\n    cropped_boxes[:, [1, 3]] = boxes[:, [1, 3]] - y_offset\n\n    return cropped_boxes\n\n\ndef uniform_crop(images, size, spatial_idx, boxes=None, scale_size=None):\n    \"\"\"\n    Perform uniform spatial sampling on the images and corresponding boxes.\n    Args:\n        images (tensor): images to perform uniform crop. The dimension is\n            `num frames` x `channel` x `height` x `width`.\n        size (int): size of height and weight to crop the images.\n        spatial_idx (int): 0, 1, or 2 for left, center, and right crop if width\n            is larger than height. Or 0, 1, or 2 for top, center, and bottom\n            crop if height is larger than width.\n        boxes (ndarray or None): optional. Corresponding boxes to images.\n            Dimension is `num boxes` x 4.\n        scale_size (int): optinal. If not None, resize the images to scale_size before\n            performing any crop.\n    Returns:\n        cropped (tensor): images with dimension of\n            `num frames` x `channel` x `size` x `size`.\n        cropped_boxes (ndarray or None): the cropped boxes with dimension of","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.uniform_crop","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.uniform_crop#L216-L272","kind":"function","name":"uniform_crop","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":216,"end_line":272,"context_start_line":196,"context_end_line":292,"code":"\ndef crop_boxes(boxes, x_offset, y_offset):\n    \"\"\"\n    Perform crop on the bounding boxes given the offsets.\n    Args:\n        boxes (ndarray or None): bounding boxes to perform crop. The dimension\n            is `num boxes` x 4.\n        x_offset (int): cropping offset in the x axis.\n        y_offset (int): cropping offset in the y axis.\n    Returns:\n        cropped_boxes (ndarray or None): the cropped boxes with dimension of\n            `num boxes` x 4.\n    \"\"\"\n    cropped_boxes = boxes.copy()\n    cropped_boxes[:, [0, 2]] = boxes[:, [0, 2]] - x_offset\n    cropped_boxes[:, [1, 3]] = boxes[:, [1, 3]] - y_offset\n\n    return cropped_boxes\n\n\ndef uniform_crop(images, size, spatial_idx, boxes=None, scale_size=None):\n    \"\"\"\n    Perform uniform spatial sampling on the images and corresponding boxes.\n    Args:\n        images (tensor): images to perform uniform crop. The dimension is\n            `num frames` x `channel` x `height` x `width`.\n        size (int): size of height and weight to crop the images.\n        spatial_idx (int): 0, 1, or 2 for left, center, and right crop if width\n            is larger than height. Or 0, 1, or 2 for top, center, and bottom\n            crop if height is larger than width.\n        boxes (ndarray or None): optional. Corresponding boxes to images.\n            Dimension is `num boxes` x 4.\n        scale_size (int): optinal. If not None, resize the images to scale_size before\n            performing any crop.\n    Returns:\n        cropped (tensor): images with dimension of\n            `num frames` x `channel` x `size` x `size`.\n        cropped_boxes (ndarray or None): the cropped boxes with dimension of\n            `num boxes` x 4.\n    \"\"\"\n    assert spatial_idx in [0, 1, 2]\n    ndim = len(images.shape)\n    if ndim == 3:\n        images = images.unsqueeze(0)\n    height = images.shape[2]\n    width = images.shape[3]\n\n    if scale_size is not None:\n        if width <= height:\n            width, height = scale_size, int(height / width * scale_size)\n        else:\n            width, height = int(width / height * scale_size), scale_size\n        images = torch.nn.functional.interpolate(\n            images,\n            size=(height, width),\n            mode=\"bilinear\",\n            align_corners=False,\n        )\n\n    y_offset = int(math.ceil((height - size) / 2))\n    x_offset = int(math.ceil((width - size) / 2))\n\n    if height > width:\n        if spatial_idx == 0:\n            y_offset = 0\n        elif spatial_idx == 2:\n            y_offset = height - size\n    else:\n        if spatial_idx == 0:\n            x_offset = 0\n        elif spatial_idx == 2:\n            x_offset = width - size\n    cropped = images[:, :, y_offset : y_offset + size, x_offset : x_offset + size]\n    cropped_boxes = crop_boxes(boxes, x_offset, y_offset) if boxes is not None else None\n    if ndim == 3:\n        cropped = cropped.squeeze(0)\n    return cropped, cropped_boxes\n\n\nclass SpatialCrop(nn.Module):\n    \"\"\"\n    Convert the video into 3 smaller clips spatially. Must be used after the\n        temporal crops to get spatial crops, and should be used with\n        -2 in the spatial crop at the slowfast augmentation stage (so full\n        frames are passed in here). Will return a larger list with the\n        3x spatial crops as well.\n    \"\"\"\n\n    def __init__(self, crop_size: int = 224, num_crops: int = 3):\n        super().__init__()\n        self.crop_size = crop_size\n        if num_crops == 3:\n            self.crops_to_ext = [0, 1, 2]\n            self.flipped_crops_to_ext = []\n        elif num_crops == 1:\n            self.crops_to_ext = [1]\n            self.flipped_crops_to_ext = []","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.SpatialCrop","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.data.SpatialCrop#L275-L315","kind":"class","name":"SpatialCrop","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":275,"end_line":315,"context_start_line":255,"context_end_line":335,"code":"    y_offset = int(math.ceil((height - size) / 2))\n    x_offset = int(math.ceil((width - size) / 2))\n\n    if height > width:\n        if spatial_idx == 0:\n            y_offset = 0\n        elif spatial_idx == 2:\n            y_offset = height - size\n    else:\n        if spatial_idx == 0:\n            x_offset = 0\n        elif spatial_idx == 2:\n            x_offset = width - size\n    cropped = images[:, :, y_offset : y_offset + size, x_offset : x_offset + size]\n    cropped_boxes = crop_boxes(boxes, x_offset, y_offset) if boxes is not None else None\n    if ndim == 3:\n        cropped = cropped.squeeze(0)\n    return cropped, cropped_boxes\n\n\nclass SpatialCrop(nn.Module):\n    \"\"\"\n    Convert the video into 3 smaller clips spatially. Must be used after the\n        temporal crops to get spatial crops, and should be used with\n        -2 in the spatial crop at the slowfast augmentation stage (so full\n        frames are passed in here). Will return a larger list with the\n        3x spatial crops as well.\n    \"\"\"\n\n    def __init__(self, crop_size: int = 224, num_crops: int = 3):\n        super().__init__()\n        self.crop_size = crop_size\n        if num_crops == 3:\n            self.crops_to_ext = [0, 1, 2]\n            self.flipped_crops_to_ext = []\n        elif num_crops == 1:\n            self.crops_to_ext = [1]\n            self.flipped_crops_to_ext = []\n        else:\n            raise NotImplementedError(\"Nothing else supported yet\")\n\n    def forward(self, videos):\n        \"\"\"\n        Args:\n            videos: A list of C, T, H, W videos.\n        Returns:\n            videos: A list with 3x the number of elements. Each video converted\n                to C, T, H', W' by spatial cropping.\n        \"\"\"\n        assert isinstance(videos, list), \"Must be a list of videos after temporal crops\"\n        assert all([video.ndim == 4 for video in videos]), \"Must be (C,T,H,W)\"\n        res = []\n        for video in videos:\n            for spatial_idx in self.crops_to_ext:\n                res.append(uniform_crop(video, self.crop_size, spatial_idx)[0])\n            if not self.flipped_crops_to_ext:\n                continue\n            flipped_video = transforms.functional.hflip(video)\n            for spatial_idx in self.flipped_crops_to_ext:\n                res.append(uniform_crop(flipped_video, self.crop_size, spatial_idx)[0])\n        return res\n\n\ndef load_and_transform_video_data(\n    video_paths,\n    device,\n    clip_duration=2,\n    clips_per_video=5,\n    sample_rate=16000,\n):\n    if video_paths is None:\n        return None\n\n    video_outputs = []\n    video_transform = transforms.Compose(\n        [\n            pv_transforms.ShortSideScale(224),\n            NormalizeVideo(\n                mean=(0.48145466, 0.4578275, 0.40821073),\n                std=(0.26862954, 0.26130258, 0.27577711),\n            ),","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.load_and_transform_video_data","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.load_and_transform_video_data#L318-L371","kind":"function","name":"load_and_transform_video_data","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":318,"end_line":371,"context_start_line":298,"context_end_line":371,"code":"        Args:\n            videos: A list of C, T, H, W videos.\n        Returns:\n            videos: A list with 3x the number of elements. Each video converted\n                to C, T, H', W' by spatial cropping.\n        \"\"\"\n        assert isinstance(videos, list), \"Must be a list of videos after temporal crops\"\n        assert all([video.ndim == 4 for video in videos]), \"Must be (C,T,H,W)\"\n        res = []\n        for video in videos:\n            for spatial_idx in self.crops_to_ext:\n                res.append(uniform_crop(video, self.crop_size, spatial_idx)[0])\n            if not self.flipped_crops_to_ext:\n                continue\n            flipped_video = transforms.functional.hflip(video)\n            for spatial_idx in self.flipped_crops_to_ext:\n                res.append(uniform_crop(flipped_video, self.crop_size, spatial_idx)[0])\n        return res\n\n\ndef load_and_transform_video_data(\n    video_paths,\n    device,\n    clip_duration=2,\n    clips_per_video=5,\n    sample_rate=16000,\n):\n    if video_paths is None:\n        return None\n\n    video_outputs = []\n    video_transform = transforms.Compose(\n        [\n            pv_transforms.ShortSideScale(224),\n            NormalizeVideo(\n                mean=(0.48145466, 0.4578275, 0.40821073),\n                std=(0.26862954, 0.26130258, 0.27577711),\n            ),\n        ]\n    )\n\n    clip_sampler = ConstantClipsPerVideoSampler(\n        clip_duration=clip_duration, clips_per_video=clips_per_video\n    )\n    frame_sampler = pv_transforms.UniformTemporalSubsample(num_samples=clip_duration)\n\n    for video_path in video_paths:\n        video = EncodedVideo.from_path(\n            video_path,\n            decoder=\"pyav\",\n            decode_audio=False,\n            # **{\"sample_rate\": sample_rate},\n        )\n\n        all_clips_timepoints = get_clip_timepoints(clip_sampler, video.duration)\n\n        all_video = []\n        for clip_timepoints in all_clips_timepoints:\n            # Read the clip, get frames\n            clip = video.get_clip(clip_timepoints[0], clip_timepoints[1])\n            if clip is None:\n                raise ValueError(\"No clip found\")\n            video_clip = frame_sampler(clip[\"video\"])\n            video_clip = video_clip / 255.0  # since this is float, need 0-1\n\n            all_video.append(video_clip)\n\n        all_video = [video_transform(clip) for clip in all_video]\n        all_video = SpatialCrop(224, num_crops=3)(all_video)\n\n        all_video = torch.stack(all_video, dim=0)\n        video_outputs.append(all_video)\n\n    return torch.stack(video_outputs, dim=0).to(device)","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.__init__#L284-L294","kind":"function","name":"__init__","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":284,"end_line":294,"context_start_line":264,"context_end_line":314,"code":"        if spatial_idx == 0:\n            x_offset = 0\n        elif spatial_idx == 2:\n            x_offset = width - size\n    cropped = images[:, :, y_offset : y_offset + size, x_offset : x_offset + size]\n    cropped_boxes = crop_boxes(boxes, x_offset, y_offset) if boxes is not None else None\n    if ndim == 3:\n        cropped = cropped.squeeze(0)\n    return cropped, cropped_boxes\n\n\nclass SpatialCrop(nn.Module):\n    \"\"\"\n    Convert the video into 3 smaller clips spatially. Must be used after the\n        temporal crops to get spatial crops, and should be used with\n        -2 in the spatial crop at the slowfast augmentation stage (so full\n        frames are passed in here). Will return a larger list with the\n        3x spatial crops as well.\n    \"\"\"\n\n    def __init__(self, crop_size: int = 224, num_crops: int = 3):\n        super().__init__()\n        self.crop_size = crop_size\n        if num_crops == 3:\n            self.crops_to_ext = [0, 1, 2]\n            self.flipped_crops_to_ext = []\n        elif num_crops == 1:\n            self.crops_to_ext = [1]\n            self.flipped_crops_to_ext = []\n        else:\n            raise NotImplementedError(\"Nothing else supported yet\")\n\n    def forward(self, videos):\n        \"\"\"\n        Args:\n            videos: A list of C, T, H, W videos.\n        Returns:\n            videos: A list with 3x the number of elements. Each video converted\n                to C, T, H', W' by spatial cropping.\n        \"\"\"\n        assert isinstance(videos, list), \"Must be a list of videos after temporal crops\"\n        assert all([video.ndim == 4 for video in videos]), \"Must be (C,T,H,W)\"\n        res = []\n        for video in videos:\n            for spatial_idx in self.crops_to_ext:\n                res.append(uniform_crop(video, self.crop_size, spatial_idx)[0])\n            if not self.flipped_crops_to_ext:\n                continue\n            flipped_video = transforms.functional.hflip(video)\n            for spatial_idx in self.flipped_crops_to_ext:\n                res.append(uniform_crop(flipped_video, self.crop_size, spatial_idx)[0])","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.data.forward","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.data.forward#L296-L315","kind":"function","name":"forward","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":296,"end_line":315,"context_start_line":276,"context_end_line":335,"code":"    \"\"\"\n    Convert the video into 3 smaller clips spatially. Must be used after the\n        temporal crops to get spatial crops, and should be used with\n        -2 in the spatial crop at the slowfast augmentation stage (so full\n        frames are passed in here). Will return a larger list with the\n        3x spatial crops as well.\n    \"\"\"\n\n    def __init__(self, crop_size: int = 224, num_crops: int = 3):\n        super().__init__()\n        self.crop_size = crop_size\n        if num_crops == 3:\n            self.crops_to_ext = [0, 1, 2]\n            self.flipped_crops_to_ext = []\n        elif num_crops == 1:\n            self.crops_to_ext = [1]\n            self.flipped_crops_to_ext = []\n        else:\n            raise NotImplementedError(\"Nothing else supported yet\")\n\n    def forward(self, videos):\n        \"\"\"\n        Args:\n            videos: A list of C, T, H, W videos.\n        Returns:\n            videos: A list with 3x the number of elements. Each video converted\n                to C, T, H', W' by spatial cropping.\n        \"\"\"\n        assert isinstance(videos, list), \"Must be a list of videos after temporal crops\"\n        assert all([video.ndim == 4 for video in videos]), \"Must be (C,T,H,W)\"\n        res = []\n        for video in videos:\n            for spatial_idx in self.crops_to_ext:\n                res.append(uniform_crop(video, self.crop_size, spatial_idx)[0])\n            if not self.flipped_crops_to_ext:\n                continue\n            flipped_video = transforms.functional.hflip(video)\n            for spatial_idx in self.flipped_crops_to_ext:\n                res.append(uniform_crop(flipped_video, self.crop_size, spatial_idx)[0])\n        return res\n\n\ndef load_and_transform_video_data(\n    video_paths,\n    device,\n    clip_duration=2,\n    clips_per_video=5,\n    sample_rate=16000,\n):\n    if video_paths is None:\n        return None\n\n    video_outputs = []\n    video_transform = transforms.Compose(\n        [\n            pv_transforms.ShortSideScale(224),\n            NormalizeVideo(\n                mean=(0.48145466, 0.4578275, 0.40821073),\n                std=(0.26862954, 0.26130258, 0.27577711),\n            ),","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.models.helpers#L1-L140","kind":"module","name":"imagebind_LLM.ImageBind.models.helpers","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":1,"end_line":140,"context_start_line":1,"context_end_line":140,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\n\nimport einops\nimport numpy as np\nimport torch\nimport torch.nn as nn\n\n\nclass Normalize(nn.Module):\n    def __init__(self, dim: int) -> None:\n        super().__init__()\n        self.dim = dim\n\n    def forward(self, x):\n        return torch.nn.functional.normalize(x, dim=self.dim, p=2)\n\n\nclass LearnableLogitScaling(nn.Module):\n    def __init__(\n        self,\n        logit_scale_init: float = 1 / 0.07,\n        learnable: bool = True,\n        max_logit_scale: float = 100,\n    ) -> None:\n        super().__init__()\n        self.max_logit_scale = max_logit_scale\n        self.logit_scale_init = logit_scale_init\n        self.learnable = learnable\n        log_logit_scale = torch.ones([]) * np.log(self.logit_scale_init)\n        if learnable:\n            self.log_logit_scale = nn.Parameter(log_logit_scale)\n        else:\n            self.register_buffer(\"log_logit_scale\", log_logit_scale)\n\n    def forward(self, x):\n        return torch.clip(self.log_logit_scale.exp(), max=self.max_logit_scale) * x\n\n    def extra_repr(self):\n        st = f\"logit_scale_init={self.logit_scale_init},learnable={self.learnable},\" \\\n             f\" max_logit_scale={self.max_logit_scale}\"\n        return st\n\n\nclass EinOpsRearrange(nn.Module):\n    def __init__(self, rearrange_expr: str, **kwargs) -> None:\n        super().__init__()\n        self.rearrange_expr = rearrange_expr\n        self.kwargs = kwargs\n\n    def forward(self, x):\n        assert isinstance(x, torch.Tensor)\n        return einops.rearrange(x, self.rearrange_expr, **self.kwargs)\n\n\nclass VerboseNNModule(nn.Module):\n    \"\"\"\n    Wrapper around nn.Module that prints registered buffers and parameter names.\n    \"\"\"\n\n    @staticmethod\n    def get_readable_tensor_repr(name: str, tensor: torch.Tensor) -> str:\n        st = (\n            \"(\"\n            + name\n            + \"): \"\n            + \"tensor(\"\n            + str(tuple(tensor[1].shape))\n            + \", requires_grad=\"\n            + str(tensor[1].requires_grad)\n            + \")\\n\"\n        )\n        return st\n\n    def extra_repr(self) -> str:\n        named_modules = set()\n        for p in self.named_modules():\n            named_modules.update([p[0]])\n        named_modules = list(named_modules)\n\n        string_repr = \"\"\n        for p in self.named_parameters():\n            name = p[0].split(\".\")[0]\n            if name not in named_modules:\n                string_repr += self.get_readable_tensor_repr(name, p)\n\n        for p in self.named_buffers():\n            name = p[0].split(\".\")[0]\n            string_repr += self.get_readable_tensor_repr(name, p)\n\n        return string_repr\n\n\ndef cast_if_src_dtype(\n    tensor: torch.Tensor, src_dtype: torch.dtype, tgt_dtype: torch.dtype\n):\n    updated = False\n    if tensor.dtype == src_dtype:\n        tensor = tensor.to(dtype=tgt_dtype)\n        updated = True\n    return tensor, updated\n\n\nclass QuickGELU(nn.Module):\n    # From https://github.com/openai/CLIP/blob/d50d76daa670286dd6cacf3bcd80b5e4823fc8e1/clip/model.py#L166\n    def forward(self, x: torch.Tensor):\n        return x * torch.sigmoid(1.702 * x)\n\n\nclass SelectElement(nn.Module):\n    def __init__(self, index) -> None:\n        super().__init__()\n        self.index = index\n\n    def forward(self, x):\n        assert x.ndim >= 3\n        return x[:, self.index, ...]\n\n\nclass SelectEOSAndProject(nn.Module):\n    \"\"\"\n    Text Pooling used in OpenCLIP\n    \"\"\"\n\n    def __init__(self, proj: nn.Module) -> None:\n        super().__init__()\n        self.proj = proj\n\n    def forward(self, x, seq_len):\n        assert x.ndim == 3\n        # x is of shape B x L x D\n        # take features from the eot embedding (eot_token is the highest number in each sequence)\n        x = x[torch.arange(x.shape[0]), seq_len]\n        x = self.proj(x)\n        return x","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.Normalize","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.helpers.Normalize#L15-L21","kind":"class","name":"Normalize","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":15,"end_line":21,"context_start_line":1,"context_end_line":41,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\n\nimport einops\nimport numpy as np\nimport torch\nimport torch.nn as nn\n\n\nclass Normalize(nn.Module):\n    def __init__(self, dim: int) -> None:\n        super().__init__()\n        self.dim = dim\n\n    def forward(self, x):\n        return torch.nn.functional.normalize(x, dim=self.dim, p=2)\n\n\nclass LearnableLogitScaling(nn.Module):\n    def __init__(\n        self,\n        logit_scale_init: float = 1 / 0.07,\n        learnable: bool = True,\n        max_logit_scale: float = 100,\n    ) -> None:\n        super().__init__()\n        self.max_logit_scale = max_logit_scale\n        self.logit_scale_init = logit_scale_init\n        self.learnable = learnable\n        log_logit_scale = torch.ones([]) * np.log(self.logit_scale_init)\n        if learnable:\n            self.log_logit_scale = nn.Parameter(log_logit_scale)\n        else:\n            self.register_buffer(\"log_logit_scale\", log_logit_scale)\n\n    def forward(self, x):","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.LearnableLogitScaling","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.helpers.LearnableLogitScaling#L24-L47","kind":"class","name":"LearnableLogitScaling","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":24,"end_line":47,"context_start_line":4,"context_end_line":67,"code":"\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\n\nimport einops\nimport numpy as np\nimport torch\nimport torch.nn as nn\n\n\nclass Normalize(nn.Module):\n    def __init__(self, dim: int) -> None:\n        super().__init__()\n        self.dim = dim\n\n    def forward(self, x):\n        return torch.nn.functional.normalize(x, dim=self.dim, p=2)\n\n\nclass LearnableLogitScaling(nn.Module):\n    def __init__(\n        self,\n        logit_scale_init: float = 1 / 0.07,\n        learnable: bool = True,\n        max_logit_scale: float = 100,\n    ) -> None:\n        super().__init__()\n        self.max_logit_scale = max_logit_scale\n        self.logit_scale_init = logit_scale_init\n        self.learnable = learnable\n        log_logit_scale = torch.ones([]) * np.log(self.logit_scale_init)\n        if learnable:\n            self.log_logit_scale = nn.Parameter(log_logit_scale)\n        else:\n            self.register_buffer(\"log_logit_scale\", log_logit_scale)\n\n    def forward(self, x):\n        return torch.clip(self.log_logit_scale.exp(), max=self.max_logit_scale) * x\n\n    def extra_repr(self):\n        st = f\"logit_scale_init={self.logit_scale_init},learnable={self.learnable},\" \\\n             f\" max_logit_scale={self.max_logit_scale}\"\n        return st\n\n\nclass EinOpsRearrange(nn.Module):\n    def __init__(self, rearrange_expr: str, **kwargs) -> None:\n        super().__init__()\n        self.rearrange_expr = rearrange_expr\n        self.kwargs = kwargs\n\n    def forward(self, x):\n        assert isinstance(x, torch.Tensor)\n        return einops.rearrange(x, self.rearrange_expr, **self.kwargs)\n\n\nclass VerboseNNModule(nn.Module):\n    \"\"\"\n    Wrapper around nn.Module that prints registered buffers and parameter names.\n    \"\"\"\n\n    @staticmethod\n    def get_readable_tensor_repr(name: str, tensor: torch.Tensor) -> str:","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.EinOpsRearrange","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.helpers.EinOpsRearrange#L50-L58","kind":"class","name":"EinOpsRearrange","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":50,"end_line":58,"context_start_line":30,"context_end_line":78,"code":"    ) -> None:\n        super().__init__()\n        self.max_logit_scale = max_logit_scale\n        self.logit_scale_init = logit_scale_init\n        self.learnable = learnable\n        log_logit_scale = torch.ones([]) * np.log(self.logit_scale_init)\n        if learnable:\n            self.log_logit_scale = nn.Parameter(log_logit_scale)\n        else:\n            self.register_buffer(\"log_logit_scale\", log_logit_scale)\n\n    def forward(self, x):\n        return torch.clip(self.log_logit_scale.exp(), max=self.max_logit_scale) * x\n\n    def extra_repr(self):\n        st = f\"logit_scale_init={self.logit_scale_init},learnable={self.learnable},\" \\\n             f\" max_logit_scale={self.max_logit_scale}\"\n        return st\n\n\nclass EinOpsRearrange(nn.Module):\n    def __init__(self, rearrange_expr: str, **kwargs) -> None:\n        super().__init__()\n        self.rearrange_expr = rearrange_expr\n        self.kwargs = kwargs\n\n    def forward(self, x):\n        assert isinstance(x, torch.Tensor)\n        return einops.rearrange(x, self.rearrange_expr, **self.kwargs)\n\n\nclass VerboseNNModule(nn.Module):\n    \"\"\"\n    Wrapper around nn.Module that prints registered buffers and parameter names.\n    \"\"\"\n\n    @staticmethod\n    def get_readable_tensor_repr(name: str, tensor: torch.Tensor) -> str:\n        st = (\n            \"(\"\n            + name\n            + \"): \"\n            + \"tensor(\"\n            + str(tuple(tensor[1].shape))\n            + \", requires_grad=\"\n            + str(tensor[1].requires_grad)\n            + \")\\n\"\n        )\n        return st","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.VerboseNNModule","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.helpers.VerboseNNModule#L61-L96","kind":"class","name":"VerboseNNModule","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":61,"end_line":96,"context_start_line":41,"context_end_line":116,"code":"    def forward(self, x):\n        return torch.clip(self.log_logit_scale.exp(), max=self.max_logit_scale) * x\n\n    def extra_repr(self):\n        st = f\"logit_scale_init={self.logit_scale_init},learnable={self.learnable},\" \\\n             f\" max_logit_scale={self.max_logit_scale}\"\n        return st\n\n\nclass EinOpsRearrange(nn.Module):\n    def __init__(self, rearrange_expr: str, **kwargs) -> None:\n        super().__init__()\n        self.rearrange_expr = rearrange_expr\n        self.kwargs = kwargs\n\n    def forward(self, x):\n        assert isinstance(x, torch.Tensor)\n        return einops.rearrange(x, self.rearrange_expr, **self.kwargs)\n\n\nclass VerboseNNModule(nn.Module):\n    \"\"\"\n    Wrapper around nn.Module that prints registered buffers and parameter names.\n    \"\"\"\n\n    @staticmethod\n    def get_readable_tensor_repr(name: str, tensor: torch.Tensor) -> str:\n        st = (\n            \"(\"\n            + name\n            + \"): \"\n            + \"tensor(\"\n            + str(tuple(tensor[1].shape))\n            + \", requires_grad=\"\n            + str(tensor[1].requires_grad)\n            + \")\\n\"\n        )\n        return st\n\n    def extra_repr(self) -> str:\n        named_modules = set()\n        for p in self.named_modules():\n            named_modules.update([p[0]])\n        named_modules = list(named_modules)\n\n        string_repr = \"\"\n        for p in self.named_parameters():\n            name = p[0].split(\".\")[0]\n            if name not in named_modules:\n                string_repr += self.get_readable_tensor_repr(name, p)\n\n        for p in self.named_buffers():\n            name = p[0].split(\".\")[0]\n            string_repr += self.get_readable_tensor_repr(name, p)\n\n        return string_repr\n\n\ndef cast_if_src_dtype(\n    tensor: torch.Tensor, src_dtype: torch.dtype, tgt_dtype: torch.dtype\n):\n    updated = False\n    if tensor.dtype == src_dtype:\n        tensor = tensor.to(dtype=tgt_dtype)\n        updated = True\n    return tensor, updated\n\n\nclass QuickGELU(nn.Module):\n    # From https://github.com/openai/CLIP/blob/d50d76daa670286dd6cacf3bcd80b5e4823fc8e1/clip/model.py#L166\n    def forward(self, x: torch.Tensor):\n        return x * torch.sigmoid(1.702 * x)\n\n\nclass SelectElement(nn.Module):\n    def __init__(self, index) -> None:","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.cast_if_src_dtype","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.helpers.cast_if_src_dtype#L99-L106","kind":"function","name":"cast_if_src_dtype","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":99,"end_line":106,"context_start_line":79,"context_end_line":126,"code":"\n    def extra_repr(self) -> str:\n        named_modules = set()\n        for p in self.named_modules():\n            named_modules.update([p[0]])\n        named_modules = list(named_modules)\n\n        string_repr = \"\"\n        for p in self.named_parameters():\n            name = p[0].split(\".\")[0]\n            if name not in named_modules:\n                string_repr += self.get_readable_tensor_repr(name, p)\n\n        for p in self.named_buffers():\n            name = p[0].split(\".\")[0]\n            string_repr += self.get_readable_tensor_repr(name, p)\n\n        return string_repr\n\n\ndef cast_if_src_dtype(\n    tensor: torch.Tensor, src_dtype: torch.dtype, tgt_dtype: torch.dtype\n):\n    updated = False\n    if tensor.dtype == src_dtype:\n        tensor = tensor.to(dtype=tgt_dtype)\n        updated = True\n    return tensor, updated\n\n\nclass QuickGELU(nn.Module):\n    # From https://github.com/openai/CLIP/blob/d50d76daa670286dd6cacf3bcd80b5e4823fc8e1/clip/model.py#L166\n    def forward(self, x: torch.Tensor):\n        return x * torch.sigmoid(1.702 * x)\n\n\nclass SelectElement(nn.Module):\n    def __init__(self, index) -> None:\n        super().__init__()\n        self.index = index\n\n    def forward(self, x):\n        assert x.ndim >= 3\n        return x[:, self.index, ...]\n\n\nclass SelectEOSAndProject(nn.Module):\n    \"\"\"","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.QuickGELU","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.helpers.QuickGELU#L109-L112","kind":"class","name":"QuickGELU","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":109,"end_line":112,"context_start_line":89,"context_end_line":132,"code":"            if name not in named_modules:\n                string_repr += self.get_readable_tensor_repr(name, p)\n\n        for p in self.named_buffers():\n            name = p[0].split(\".\")[0]\n            string_repr += self.get_readable_tensor_repr(name, p)\n\n        return string_repr\n\n\ndef cast_if_src_dtype(\n    tensor: torch.Tensor, src_dtype: torch.dtype, tgt_dtype: torch.dtype\n):\n    updated = False\n    if tensor.dtype == src_dtype:\n        tensor = tensor.to(dtype=tgt_dtype)\n        updated = True\n    return tensor, updated\n\n\nclass QuickGELU(nn.Module):\n    # From https://github.com/openai/CLIP/blob/d50d76daa670286dd6cacf3bcd80b5e4823fc8e1/clip/model.py#L166\n    def forward(self, x: torch.Tensor):\n        return x * torch.sigmoid(1.702 * x)\n\n\nclass SelectElement(nn.Module):\n    def __init__(self, index) -> None:\n        super().__init__()\n        self.index = index\n\n    def forward(self, x):\n        assert x.ndim >= 3\n        return x[:, self.index, ...]\n\n\nclass SelectEOSAndProject(nn.Module):\n    \"\"\"\n    Text Pooling used in OpenCLIP\n    \"\"\"\n\n    def __init__(self, proj: nn.Module) -> None:\n        super().__init__()\n        self.proj = proj","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.SelectElement","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.helpers.SelectElement#L115-L122","kind":"class","name":"SelectElement","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":115,"end_line":122,"context_start_line":95,"context_end_line":140,"code":"\n        return string_repr\n\n\ndef cast_if_src_dtype(\n    tensor: torch.Tensor, src_dtype: torch.dtype, tgt_dtype: torch.dtype\n):\n    updated = False\n    if tensor.dtype == src_dtype:\n        tensor = tensor.to(dtype=tgt_dtype)\n        updated = True\n    return tensor, updated\n\n\nclass QuickGELU(nn.Module):\n    # From https://github.com/openai/CLIP/blob/d50d76daa670286dd6cacf3bcd80b5e4823fc8e1/clip/model.py#L166\n    def forward(self, x: torch.Tensor):\n        return x * torch.sigmoid(1.702 * x)\n\n\nclass SelectElement(nn.Module):\n    def __init__(self, index) -> None:\n        super().__init__()\n        self.index = index\n\n    def forward(self, x):\n        assert x.ndim >= 3\n        return x[:, self.index, ...]\n\n\nclass SelectEOSAndProject(nn.Module):\n    \"\"\"\n    Text Pooling used in OpenCLIP\n    \"\"\"\n\n    def __init__(self, proj: nn.Module) -> None:\n        super().__init__()\n        self.proj = proj\n\n    def forward(self, x, seq_len):\n        assert x.ndim == 3\n        # x is of shape B x L x D\n        # take features from the eot embedding (eot_token is the highest number in each sequence)\n        x = x[torch.arange(x.shape[0]), seq_len]\n        x = self.proj(x)\n        return x","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.SelectEOSAndProject","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.helpers.SelectEOSAndProject#L125-L140","kind":"class","name":"SelectEOSAndProject","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":125,"end_line":140,"context_start_line":105,"context_end_line":140,"code":"        updated = True\n    return tensor, updated\n\n\nclass QuickGELU(nn.Module):\n    # From https://github.com/openai/CLIP/blob/d50d76daa670286dd6cacf3bcd80b5e4823fc8e1/clip/model.py#L166\n    def forward(self, x: torch.Tensor):\n        return x * torch.sigmoid(1.702 * x)\n\n\nclass SelectElement(nn.Module):\n    def __init__(self, index) -> None:\n        super().__init__()\n        self.index = index\n\n    def forward(self, x):\n        assert x.ndim >= 3\n        return x[:, self.index, ...]\n\n\nclass SelectEOSAndProject(nn.Module):\n    \"\"\"\n    Text Pooling used in OpenCLIP\n    \"\"\"\n\n    def __init__(self, proj: nn.Module) -> None:\n        super().__init__()\n        self.proj = proj\n\n    def forward(self, x, seq_len):\n        assert x.ndim == 3\n        # x is of shape B x L x D\n        # take features from the eot embedding (eot_token is the highest number in each sequence)\n        x = x[torch.arange(x.shape[0]), seq_len]\n        x = self.proj(x)\n        return x","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.helpers.__init__#L130-L132","kind":"function","name":"__init__","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":130,"end_line":132,"context_start_line":110,"context_end_line":140,"code":"    # From https://github.com/openai/CLIP/blob/d50d76daa670286dd6cacf3bcd80b5e4823fc8e1/clip/model.py#L166\n    def forward(self, x: torch.Tensor):\n        return x * torch.sigmoid(1.702 * x)\n\n\nclass SelectElement(nn.Module):\n    def __init__(self, index) -> None:\n        super().__init__()\n        self.index = index\n\n    def forward(self, x):\n        assert x.ndim >= 3\n        return x[:, self.index, ...]\n\n\nclass SelectEOSAndProject(nn.Module):\n    \"\"\"\n    Text Pooling used in OpenCLIP\n    \"\"\"\n\n    def __init__(self, proj: nn.Module) -> None:\n        super().__init__()\n        self.proj = proj\n\n    def forward(self, x, seq_len):\n        assert x.ndim == 3\n        # x is of shape B x L x D\n        # take features from the eot embedding (eot_token is the highest number in each sequence)\n        x = x[torch.arange(x.shape[0]), seq_len]\n        x = self.proj(x)\n        return x","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.forward","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.helpers.forward#L134-L140","kind":"function","name":"forward","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":134,"end_line":140,"context_start_line":114,"context_end_line":140,"code":"\nclass SelectElement(nn.Module):\n    def __init__(self, index) -> None:\n        super().__init__()\n        self.index = index\n\n    def forward(self, x):\n        assert x.ndim >= 3\n        return x[:, self.index, ...]\n\n\nclass SelectEOSAndProject(nn.Module):\n    \"\"\"\n    Text Pooling used in OpenCLIP\n    \"\"\"\n\n    def __init__(self, proj: nn.Module) -> None:\n        super().__init__()\n        self.proj = proj\n\n    def forward(self, x, seq_len):\n        assert x.ndim == 3\n        # x is of shape B x L x D\n        # take features from the eot embedding (eot_token is the highest number in each sequence)\n        x = x[torch.arange(x.shape[0]), seq_len]\n        x = self.proj(x)\n        return x","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.extra_repr","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.helpers.extra_repr#L80-L96","kind":"function","name":"extra_repr","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":80,"end_line":96,"context_start_line":60,"context_end_line":116,"code":"\nclass VerboseNNModule(nn.Module):\n    \"\"\"\n    Wrapper around nn.Module that prints registered buffers and parameter names.\n    \"\"\"\n\n    @staticmethod\n    def get_readable_tensor_repr(name: str, tensor: torch.Tensor) -> str:\n        st = (\n            \"(\"\n            + name\n            + \"): \"\n            + \"tensor(\"\n            + str(tuple(tensor[1].shape))\n            + \", requires_grad=\"\n            + str(tensor[1].requires_grad)\n            + \")\\n\"\n        )\n        return st\n\n    def extra_repr(self) -> str:\n        named_modules = set()\n        for p in self.named_modules():\n            named_modules.update([p[0]])\n        named_modules = list(named_modules)\n\n        string_repr = \"\"\n        for p in self.named_parameters():\n            name = p[0].split(\".\")[0]\n            if name not in named_modules:\n                string_repr += self.get_readable_tensor_repr(name, p)\n\n        for p in self.named_buffers():\n            name = p[0].split(\".\")[0]\n            string_repr += self.get_readable_tensor_repr(name, p)\n\n        return string_repr\n\n\ndef cast_if_src_dtype(\n    tensor: torch.Tensor, src_dtype: torch.dtype, tgt_dtype: torch.dtype\n):\n    updated = False\n    if tensor.dtype == src_dtype:\n        tensor = tensor.to(dtype=tgt_dtype)\n        updated = True\n    return tensor, updated\n\n\nclass QuickGELU(nn.Module):\n    # From https://github.com/openai/CLIP/blob/d50d76daa670286dd6cacf3bcd80b5e4823fc8e1/clip/model.py#L166\n    def forward(self, x: torch.Tensor):\n        return x * torch.sigmoid(1.702 * x)\n\n\nclass SelectElement(nn.Module):\n    def __init__(self, index) -> None:","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.helpers.get_readable_tensor_repr","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.helpers.get_readable_tensor_repr#L67-L78","kind":"function","name":"get_readable_tensor_repr","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":67,"end_line":78,"context_start_line":47,"context_end_line":98,"code":"        return st\n\n\nclass EinOpsRearrange(nn.Module):\n    def __init__(self, rearrange_expr: str, **kwargs) -> None:\n        super().__init__()\n        self.rearrange_expr = rearrange_expr\n        self.kwargs = kwargs\n\n    def forward(self, x):\n        assert isinstance(x, torch.Tensor)\n        return einops.rearrange(x, self.rearrange_expr, **self.kwargs)\n\n\nclass VerboseNNModule(nn.Module):\n    \"\"\"\n    Wrapper around nn.Module that prints registered buffers and parameter names.\n    \"\"\"\n\n    @staticmethod\n    def get_readable_tensor_repr(name: str, tensor: torch.Tensor) -> str:\n        st = (\n            \"(\"\n            + name\n            + \"): \"\n            + \"tensor(\"\n            + str(tuple(tensor[1].shape))\n            + \", requires_grad=\"\n            + str(tensor[1].requires_grad)\n            + \")\\n\"\n        )\n        return st\n\n    def extra_repr(self) -> str:\n        named_modules = set()\n        for p in self.named_modules():\n            named_modules.update([p[0]])\n        named_modules = list(named_modules)\n\n        string_repr = \"\"\n        for p in self.named_parameters():\n            name = p[0].split(\".\")[0]\n            if name not in named_modules:\n                string_repr += self.get_readable_tensor_repr(name, p)\n\n        for p in self.named_buffers():\n            name = p[0].split(\".\")[0]\n            string_repr += self.get_readable_tensor_repr(name, p)\n\n        return string_repr\n\n","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.transformer","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.models.transformer#L1-L280","kind":"module","name":"imagebind_LLM.ImageBind.models.transformer","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":1,"end_line":280,"context_start_line":1,"context_end_line":280,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\n# Code modified from\n# https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py ;\n# https://github.com/facebookresearch/deit/blob/main/models.py\n# and https://github.com/facebookresearch/vissl/blob/main/vissl/models/trunks/vision_transformer.py\n\n\nfrom functools import partial\nfrom typing import Callable, List, Optional\n\nimport torch\nimport torch.nn as nn\nimport torch.utils.checkpoint as checkpoint\nfrom timm.models.layers import DropPath, trunc_normal_\n\n\nclass Attention(nn.Module):\n    def __init__(\n        self,\n        dim,\n        num_heads=8,\n        qkv_bias=False,\n        qk_scale=None,\n        attn_drop=0.0,\n        proj_drop=0.0,\n    ):\n        super().__init__()\n        self.num_heads = num_heads\n        head_dim = dim // num_heads\n        # NOTE scale factor was wrong in my original version,\n        # can set manually to be compat with prev weights\n        self.scale = qk_scale or head_dim**-0.5\n\n        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)\n        self.attn_drop = nn.Dropout(attn_drop)\n        self.proj = nn.Linear(dim, dim)\n        self.proj_drop = nn.Dropout(proj_drop)\n\n    def forward(self, x):\n        B, N, C = x.shape\n        qkv = (\n            self.qkv(x)\n            .reshape(B, N, 3, self.num_heads, C // self.num_heads)\n            .permute(2, 0, 3, 1, 4)\n        )\n        q, k, v = (\n            qkv[0],\n            qkv[1],\n            qkv[2],\n        )  # make torchscript happy (cannot use tensor as tuple)\n\n        attn = (q @ k.transpose(-2, -1)) * self.scale\n        attn = attn.softmax(dim=-1)\n        attn = self.attn_drop(attn)\n\n        x = (attn @ v).transpose(1, 2).reshape(B, N, C)\n        x = self.proj(x)\n        x = self.proj_drop(x)\n        return x\n\n\nclass Mlp(nn.Module):\n    def __init__(\n        self,\n        in_features,\n        hidden_features=None,\n        out_features=None,\n        act_layer=nn.GELU,\n        drop=0.0,\n    ):\n        super().__init__()\n        out_features = out_features or in_features\n        hidden_features = hidden_features or in_features\n        self.fc1 = nn.Linear(in_features, hidden_features)\n        self.act = act_layer()\n        self.fc2 = nn.Linear(hidden_features, out_features)\n        self.drop = nn.Dropout(drop)\n\n    def forward(self, x):\n        x = self.fc1(x)\n        x = self.act(x)\n        x = self.drop(x)\n        x = self.fc2(x)\n        x = self.drop(x)\n        return x\n\n\nclass MultiheadAttention(nn.MultiheadAttention):\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        return super().forward(x, x, x, need_weights=False, attn_mask=attn_mask)[0]\n\n\nclass ViTAttention(Attention):\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        assert attn_mask is None\n        return super().forward(x)\n\n\nclass BlockWithMasking(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        attn_target: Callable,\n        mlp_ratio: int = 4,\n        act_layer: Callable = nn.GELU,\n        norm_layer: Callable = nn.LayerNorm,\n        ffn_dropout_rate: float = 0.0,\n        drop_path: float = 0.0,\n        layer_scale_type: Optional[str] = None,\n        layer_scale_init_value: float = 1e-4,\n    ):\n        super().__init__()\n\n        assert not isinstance(\n            attn_target, nn.Module\n        ), \"attn_target should be a Callable. Otherwise attn_target is shared across blocks!\"\n        self.attn = attn_target()\n        if drop_path > 0.0:\n            self.drop_path = DropPath(drop_path)\n        else:\n            self.drop_path = nn.Identity()\n        self.norm_1 = norm_layer(dim)\n        mlp_hidden_dim = int(mlp_ratio * dim)\n        self.mlp = Mlp(\n            in_features=dim,\n            hidden_features=mlp_hidden_dim,\n            act_layer=act_layer,\n            drop=ffn_dropout_rate,\n        )\n        self.norm_2 = norm_layer(dim)\n        self.layer_scale_type = layer_scale_type\n        if self.layer_scale_type is not None:\n            assert self.layer_scale_type in [\n                \"per_channel\",\n                \"scalar\",\n            ], f\"Found Layer scale type {self.layer_scale_type}\"\n            if self.layer_scale_type == \"per_channel\":\n                # one gamma value per channel\n                gamma_shape = [1, 1, dim]\n            elif self.layer_scale_type == \"scalar\":\n                # single gamma value for all channels\n                gamma_shape = [1, 1, 1]\n            # two gammas: for each part of the fwd in the encoder\n            self.layer_scale_gamma1 = nn.Parameter(\n                torch.ones(size=gamma_shape) * layer_scale_init_value,\n                requires_grad=True,\n            )\n            self.layer_scale_gamma2 = nn.Parameter(\n                torch.ones(size=gamma_shape) * layer_scale_init_value,\n                requires_grad=True,\n            )\n\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        if self.layer_scale_type is None:\n            x = x + self.drop_path(self.attn(self.norm_1(x), attn_mask))\n            x = x + self.drop_path(self.mlp(self.norm_2(x)))\n        else:\n            x = (\n                x\n                + self.drop_path(self.attn(self.norm_1(x), attn_mask))\n                * self.layer_scale_gamma1\n            )\n            x = x + self.drop_path(self.mlp(self.norm_2(x))) * self.layer_scale_gamma2\n        return x\n\n\n_LAYER_NORM = partial(nn.LayerNorm, eps=1e-6)\n\n\nclass SimpleTransformer(nn.Module):\n    def __init__(\n        self,\n        attn_target: Callable,\n        embed_dim: int,\n        num_blocks: int,\n        block: Callable = BlockWithMasking,\n        pre_transformer_layer: Optional[Callable] = None,\n        post_transformer_layer: Optional[Callable] = None,\n        drop_path_rate: float = 0.0,\n        drop_path_type: str = \"progressive\",\n        norm_layer: Callable = _LAYER_NORM,\n        mlp_ratio: int = 4,\n        ffn_dropout_rate: float = 0.0,\n        layer_scale_type: Optional[str] = None,  # from cait; possible values are None, \"per_channel\", \"scalar\"\n        layer_scale_init_value: float = 1e-4,  # from cait; float\n        weight_init_style: str = \"jax\",  # possible values jax or pytorch\n    ):\n        \"\"\"\n        Simple Transformer with the following features\n        1. Supports masked attention\n        2. Supports DropPath\n        3. Supports LayerScale\n        4. Supports Dropout in Attention and FFN\n        5. Makes few assumptions about the input except that it is a Tensor\n        \"\"\"\n        super().__init__()\n        self.pre_transformer_layer = pre_transformer_layer\n        if drop_path_type == \"progressive\":\n            dpr = [x.item() for x in torch.linspace(0, drop_path_rate, num_blocks)]\n        elif drop_path_type == \"uniform\":\n            dpr = [drop_path_rate for i in range(num_blocks)]\n        else:\n            raise ValueError(f\"Unknown drop_path_type: {drop_path_type}\")\n\n        self.blocks = nn.Sequential(\n            *[\n                block(\n                    dim=embed_dim,\n                    attn_target=attn_target,\n                    mlp_ratio=mlp_ratio,\n                    ffn_dropout_rate=ffn_dropout_rate,\n                    drop_path=dpr[i],\n                    norm_layer=norm_layer,\n                    layer_scale_type=layer_scale_type,\n                    layer_scale_init_value=layer_scale_init_value,\n                )\n                for i in range(num_blocks)\n            ]\n        )\n        self.post_transformer_layer = post_transformer_layer\n        self.weight_init_style = weight_init_style\n        self.apply(self._init_weights)\n\n    def _init_weights(self, m):\n        if isinstance(m, nn.Linear):\n            if self.weight_init_style == \"jax\":\n                # Based on MAE and official Jax ViT implementation\n                torch.nn.init.xavier_uniform_(m.weight)\n            elif self.weight_init_style == \"pytorch\":\n                # PyTorch ViT uses trunc_normal_\n                trunc_normal_(m.weight, std=0.02)\n\n            if m.bias is not None:\n                nn.init.constant_(m.bias, 0)\n        elif isinstance(m, (nn.LayerNorm)):\n            nn.init.constant_(m.bias, 0)\n            nn.init.constant_(m.weight, 1.0)\n\n    def forward(\n        self,\n        tokens: torch.Tensor,\n        attn_mask: torch.Tensor = None,\n        use_checkpoint: bool = False,\n        checkpoint_every_n: int = 1,\n        checkpoint_blk_ids: Optional[List[int]] = None,\n    ):\n        \"\"\"\n        Inputs\n        - tokens: data of shape N x L x D (or L x N x D depending on the attention implementation)\n        - attn: mask of shape L x L\n\n        Output\n        - x: data of shape N x L x D (or L x N x D depending on the attention implementation)\n        \"\"\"\n        if self.pre_transformer_layer:\n            tokens = self.pre_transformer_layer(tokens)\n        if use_checkpoint and checkpoint_blk_ids is None:\n            checkpoint_blk_ids = [\n                blk_id\n                for blk_id in range(len(self.blocks))\n                if blk_id % checkpoint_every_n == 0\n            ]\n        if checkpoint_blk_ids:\n            checkpoint_blk_ids = set(checkpoint_blk_ids)\n        for blk_id, blk in enumerate(self.blocks):\n            if use_checkpoint and blk_id in checkpoint_blk_ids:\n                tokens = checkpoint.checkpoint(\n                    blk, tokens, attn_mask, use_reentrant=False\n                )\n            else:\n                tokens = blk(tokens, attn_mask=attn_mask)\n        if self.post_transformer_layer:\n            tokens = self.post_transformer_layer(tokens)\n        return tokens","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.transformer.Attention","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.transformer.Attention#L23-L65","kind":"class","name":"Attention","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":23,"end_line":65,"context_start_line":3,"context_end_line":85,"code":"# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\n# Code modified from\n# https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py ;\n# https://github.com/facebookresearch/deit/blob/main/models.py\n# and https://github.com/facebookresearch/vissl/blob/main/vissl/models/trunks/vision_transformer.py\n\n\nfrom functools import partial\nfrom typing import Callable, List, Optional\n\nimport torch\nimport torch.nn as nn\nimport torch.utils.checkpoint as checkpoint\nfrom timm.models.layers import DropPath, trunc_normal_\n\n\nclass Attention(nn.Module):\n    def __init__(\n        self,\n        dim,\n        num_heads=8,\n        qkv_bias=False,\n        qk_scale=None,\n        attn_drop=0.0,\n        proj_drop=0.0,\n    ):\n        super().__init__()\n        self.num_heads = num_heads\n        head_dim = dim // num_heads\n        # NOTE scale factor was wrong in my original version,\n        # can set manually to be compat with prev weights\n        self.scale = qk_scale or head_dim**-0.5\n\n        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)\n        self.attn_drop = nn.Dropout(attn_drop)\n        self.proj = nn.Linear(dim, dim)\n        self.proj_drop = nn.Dropout(proj_drop)\n\n    def forward(self, x):\n        B, N, C = x.shape\n        qkv = (\n            self.qkv(x)\n            .reshape(B, N, 3, self.num_heads, C // self.num_heads)\n            .permute(2, 0, 3, 1, 4)\n        )\n        q, k, v = (\n            qkv[0],\n            qkv[1],\n            qkv[2],\n        )  # make torchscript happy (cannot use tensor as tuple)\n\n        attn = (q @ k.transpose(-2, -1)) * self.scale\n        attn = attn.softmax(dim=-1)\n        attn = self.attn_drop(attn)\n\n        x = (attn @ v).transpose(1, 2).reshape(B, N, C)\n        x = self.proj(x)\n        x = self.proj_drop(x)\n        return x\n\n\nclass Mlp(nn.Module):\n    def __init__(\n        self,\n        in_features,\n        hidden_features=None,\n        out_features=None,\n        act_layer=nn.GELU,\n        drop=0.0,\n    ):\n        super().__init__()\n        out_features = out_features or in_features\n        hidden_features = hidden_features or in_features\n        self.fc1 = nn.Linear(in_features, hidden_features)\n        self.act = act_layer()\n        self.fc2 = nn.Linear(hidden_features, out_features)\n        self.drop = nn.Dropout(drop)\n\n    def forward(self, x):","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.transformer.Mlp","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.transformer.Mlp#L68-L91","kind":"class","name":"Mlp","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":68,"end_line":91,"context_start_line":48,"context_end_line":111,"code":"            self.qkv(x)\n            .reshape(B, N, 3, self.num_heads, C // self.num_heads)\n            .permute(2, 0, 3, 1, 4)\n        )\n        q, k, v = (\n            qkv[0],\n            qkv[1],\n            qkv[2],\n        )  # make torchscript happy (cannot use tensor as tuple)\n\n        attn = (q @ k.transpose(-2, -1)) * self.scale\n        attn = attn.softmax(dim=-1)\n        attn = self.attn_drop(attn)\n\n        x = (attn @ v).transpose(1, 2).reshape(B, N, C)\n        x = self.proj(x)\n        x = self.proj_drop(x)\n        return x\n\n\nclass Mlp(nn.Module):\n    def __init__(\n        self,\n        in_features,\n        hidden_features=None,\n        out_features=None,\n        act_layer=nn.GELU,\n        drop=0.0,\n    ):\n        super().__init__()\n        out_features = out_features or in_features\n        hidden_features = hidden_features or in_features\n        self.fc1 = nn.Linear(in_features, hidden_features)\n        self.act = act_layer()\n        self.fc2 = nn.Linear(hidden_features, out_features)\n        self.drop = nn.Dropout(drop)\n\n    def forward(self, x):\n        x = self.fc1(x)\n        x = self.act(x)\n        x = self.drop(x)\n        x = self.fc2(x)\n        x = self.drop(x)\n        return x\n\n\nclass MultiheadAttention(nn.MultiheadAttention):\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        return super().forward(x, x, x, need_weights=False, attn_mask=attn_mask)[0]\n\n\nclass ViTAttention(Attention):\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        assert attn_mask is None\n        return super().forward(x)\n\n\nclass BlockWithMasking(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        attn_target: Callable,\n        mlp_ratio: int = 4,\n        act_layer: Callable = nn.GELU,","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.transformer.MultiheadAttention","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.transformer.MultiheadAttention#L94-L96","kind":"class","name":"MultiheadAttention","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":94,"end_line":96,"context_start_line":74,"context_end_line":116,"code":"        act_layer=nn.GELU,\n        drop=0.0,\n    ):\n        super().__init__()\n        out_features = out_features or in_features\n        hidden_features = hidden_features or in_features\n        self.fc1 = nn.Linear(in_features, hidden_features)\n        self.act = act_layer()\n        self.fc2 = nn.Linear(hidden_features, out_features)\n        self.drop = nn.Dropout(drop)\n\n    def forward(self, x):\n        x = self.fc1(x)\n        x = self.act(x)\n        x = self.drop(x)\n        x = self.fc2(x)\n        x = self.drop(x)\n        return x\n\n\nclass MultiheadAttention(nn.MultiheadAttention):\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        return super().forward(x, x, x, need_weights=False, attn_mask=attn_mask)[0]\n\n\nclass ViTAttention(Attention):\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        assert attn_mask is None\n        return super().forward(x)\n\n\nclass BlockWithMasking(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        attn_target: Callable,\n        mlp_ratio: int = 4,\n        act_layer: Callable = nn.GELU,\n        norm_layer: Callable = nn.LayerNorm,\n        ffn_dropout_rate: float = 0.0,\n        drop_path: float = 0.0,\n        layer_scale_type: Optional[str] = None,\n        layer_scale_init_value: float = 1e-4,","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.transformer.ViTAttention","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.transformer.ViTAttention#L99-L102","kind":"class","name":"ViTAttention","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":99,"end_line":102,"context_start_line":79,"context_end_line":122,"code":"        hidden_features = hidden_features or in_features\n        self.fc1 = nn.Linear(in_features, hidden_features)\n        self.act = act_layer()\n        self.fc2 = nn.Linear(hidden_features, out_features)\n        self.drop = nn.Dropout(drop)\n\n    def forward(self, x):\n        x = self.fc1(x)\n        x = self.act(x)\n        x = self.drop(x)\n        x = self.fc2(x)\n        x = self.drop(x)\n        return x\n\n\nclass MultiheadAttention(nn.MultiheadAttention):\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        return super().forward(x, x, x, need_weights=False, attn_mask=attn_mask)[0]\n\n\nclass ViTAttention(Attention):\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        assert attn_mask is None\n        return super().forward(x)\n\n\nclass BlockWithMasking(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        attn_target: Callable,\n        mlp_ratio: int = 4,\n        act_layer: Callable = nn.GELU,\n        norm_layer: Callable = nn.LayerNorm,\n        ffn_dropout_rate: float = 0.0,\n        drop_path: float = 0.0,\n        layer_scale_type: Optional[str] = None,\n        layer_scale_init_value: float = 1e-4,\n    ):\n        super().__init__()\n\n        assert not isinstance(\n            attn_target, nn.Module\n        ), \"attn_target should be a Callable. Otherwise attn_target is shared across blocks!\"","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.transformer.BlockWithMasking","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.transformer.BlockWithMasking#L105-L170","kind":"class","name":"BlockWithMasking","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":105,"end_line":170,"context_start_line":85,"context_end_line":190,"code":"    def forward(self, x):\n        x = self.fc1(x)\n        x = self.act(x)\n        x = self.drop(x)\n        x = self.fc2(x)\n        x = self.drop(x)\n        return x\n\n\nclass MultiheadAttention(nn.MultiheadAttention):\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        return super().forward(x, x, x, need_weights=False, attn_mask=attn_mask)[0]\n\n\nclass ViTAttention(Attention):\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        assert attn_mask is None\n        return super().forward(x)\n\n\nclass BlockWithMasking(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        attn_target: Callable,\n        mlp_ratio: int = 4,\n        act_layer: Callable = nn.GELU,\n        norm_layer: Callable = nn.LayerNorm,\n        ffn_dropout_rate: float = 0.0,\n        drop_path: float = 0.0,\n        layer_scale_type: Optional[str] = None,\n        layer_scale_init_value: float = 1e-4,\n    ):\n        super().__init__()\n\n        assert not isinstance(\n            attn_target, nn.Module\n        ), \"attn_target should be a Callable. Otherwise attn_target is shared across blocks!\"\n        self.attn = attn_target()\n        if drop_path > 0.0:\n            self.drop_path = DropPath(drop_path)\n        else:\n            self.drop_path = nn.Identity()\n        self.norm_1 = norm_layer(dim)\n        mlp_hidden_dim = int(mlp_ratio * dim)\n        self.mlp = Mlp(\n            in_features=dim,\n            hidden_features=mlp_hidden_dim,\n            act_layer=act_layer,\n            drop=ffn_dropout_rate,\n        )\n        self.norm_2 = norm_layer(dim)\n        self.layer_scale_type = layer_scale_type\n        if self.layer_scale_type is not None:\n            assert self.layer_scale_type in [\n                \"per_channel\",\n                \"scalar\",\n            ], f\"Found Layer scale type {self.layer_scale_type}\"\n            if self.layer_scale_type == \"per_channel\":\n                # one gamma value per channel\n                gamma_shape = [1, 1, dim]\n            elif self.layer_scale_type == \"scalar\":\n                # single gamma value for all channels\n                gamma_shape = [1, 1, 1]\n            # two gammas: for each part of the fwd in the encoder\n            self.layer_scale_gamma1 = nn.Parameter(\n                torch.ones(size=gamma_shape) * layer_scale_init_value,\n                requires_grad=True,\n            )\n            self.layer_scale_gamma2 = nn.Parameter(\n                torch.ones(size=gamma_shape) * layer_scale_init_value,\n                requires_grad=True,\n            )\n\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        if self.layer_scale_type is None:\n            x = x + self.drop_path(self.attn(self.norm_1(x), attn_mask))\n            x = x + self.drop_path(self.mlp(self.norm_2(x)))\n        else:\n            x = (\n                x\n                + self.drop_path(self.attn(self.norm_1(x), attn_mask))\n                * self.layer_scale_gamma1\n            )\n            x = x + self.drop_path(self.mlp(self.norm_2(x))) * self.layer_scale_gamma2\n        return x\n\n\n_LAYER_NORM = partial(nn.LayerNorm, eps=1e-6)\n\n\nclass SimpleTransformer(nn.Module):\n    def __init__(\n        self,\n        attn_target: Callable,\n        embed_dim: int,\n        num_blocks: int,\n        block: Callable = BlockWithMasking,\n        pre_transformer_layer: Optional[Callable] = None,\n        post_transformer_layer: Optional[Callable] = None,\n        drop_path_rate: float = 0.0,\n        drop_path_type: str = \"progressive\",\n        norm_layer: Callable = _LAYER_NORM,\n        mlp_ratio: int = 4,\n        ffn_dropout_rate: float = 0.0,\n        layer_scale_type: Optional[str] = None,  # from cait; possible values are None, \"per_channel\", \"scalar\"","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.transformer.SimpleTransformer","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.transformer.SimpleTransformer#L176-L280","kind":"class","name":"SimpleTransformer","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":176,"end_line":280,"context_start_line":156,"context_end_line":280,"code":"                requires_grad=True,\n            )\n\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        if self.layer_scale_type is None:\n            x = x + self.drop_path(self.attn(self.norm_1(x), attn_mask))\n            x = x + self.drop_path(self.mlp(self.norm_2(x)))\n        else:\n            x = (\n                x\n                + self.drop_path(self.attn(self.norm_1(x), attn_mask))\n                * self.layer_scale_gamma1\n            )\n            x = x + self.drop_path(self.mlp(self.norm_2(x))) * self.layer_scale_gamma2\n        return x\n\n\n_LAYER_NORM = partial(nn.LayerNorm, eps=1e-6)\n\n\nclass SimpleTransformer(nn.Module):\n    def __init__(\n        self,\n        attn_target: Callable,\n        embed_dim: int,\n        num_blocks: int,\n        block: Callable = BlockWithMasking,\n        pre_transformer_layer: Optional[Callable] = None,\n        post_transformer_layer: Optional[Callable] = None,\n        drop_path_rate: float = 0.0,\n        drop_path_type: str = \"progressive\",\n        norm_layer: Callable = _LAYER_NORM,\n        mlp_ratio: int = 4,\n        ffn_dropout_rate: float = 0.0,\n        layer_scale_type: Optional[str] = None,  # from cait; possible values are None, \"per_channel\", \"scalar\"\n        layer_scale_init_value: float = 1e-4,  # from cait; float\n        weight_init_style: str = \"jax\",  # possible values jax or pytorch\n    ):\n        \"\"\"\n        Simple Transformer with the following features\n        1. Supports masked attention\n        2. Supports DropPath\n        3. Supports LayerScale\n        4. Supports Dropout in Attention and FFN\n        5. Makes few assumptions about the input except that it is a Tensor\n        \"\"\"\n        super().__init__()\n        self.pre_transformer_layer = pre_transformer_layer\n        if drop_path_type == \"progressive\":\n            dpr = [x.item() for x in torch.linspace(0, drop_path_rate, num_blocks)]\n        elif drop_path_type == \"uniform\":\n            dpr = [drop_path_rate for i in range(num_blocks)]\n        else:\n            raise ValueError(f\"Unknown drop_path_type: {drop_path_type}\")\n\n        self.blocks = nn.Sequential(\n            *[\n                block(\n                    dim=embed_dim,\n                    attn_target=attn_target,\n                    mlp_ratio=mlp_ratio,\n                    ffn_dropout_rate=ffn_dropout_rate,\n                    drop_path=dpr[i],\n                    norm_layer=norm_layer,\n                    layer_scale_type=layer_scale_type,\n                    layer_scale_init_value=layer_scale_init_value,\n                )\n                for i in range(num_blocks)\n            ]\n        )\n        self.post_transformer_layer = post_transformer_layer\n        self.weight_init_style = weight_init_style\n        self.apply(self._init_weights)\n\n    def _init_weights(self, m):\n        if isinstance(m, nn.Linear):\n            if self.weight_init_style == \"jax\":\n                # Based on MAE and official Jax ViT implementation\n                torch.nn.init.xavier_uniform_(m.weight)\n            elif self.weight_init_style == \"pytorch\":\n                # PyTorch ViT uses trunc_normal_\n                trunc_normal_(m.weight, std=0.02)\n\n            if m.bias is not None:\n                nn.init.constant_(m.bias, 0)\n        elif isinstance(m, (nn.LayerNorm)):\n            nn.init.constant_(m.bias, 0)\n            nn.init.constant_(m.weight, 1.0)\n\n    def forward(\n        self,\n        tokens: torch.Tensor,\n        attn_mask: torch.Tensor = None,\n        use_checkpoint: bool = False,\n        checkpoint_every_n: int = 1,\n        checkpoint_blk_ids: Optional[List[int]] = None,\n    ):\n        \"\"\"\n        Inputs\n        - tokens: data of shape N x L x D (or L x N x D depending on the attention implementation)\n        - attn: mask of shape L x L\n\n        Output\n        - x: data of shape N x L x D (or L x N x D depending on the attention implementation)\n        \"\"\"\n        if self.pre_transformer_layer:\n            tokens = self.pre_transformer_layer(tokens)\n        if use_checkpoint and checkpoint_blk_ids is None:\n            checkpoint_blk_ids = [\n                blk_id\n                for blk_id in range(len(self.blocks))\n                if blk_id % checkpoint_every_n == 0\n            ]\n        if checkpoint_blk_ids:\n            checkpoint_blk_ids = set(checkpoint_blk_ids)\n        for blk_id, blk in enumerate(self.blocks):\n            if use_checkpoint and blk_id in checkpoint_blk_ids:\n                tokens = checkpoint.checkpoint(\n                    blk, tokens, attn_mask, use_reentrant=False\n                )\n            else:\n                tokens = blk(tokens, attn_mask=attn_mask)\n        if self.post_transformer_layer:\n            tokens = self.post_transformer_layer(tokens)\n        return tokens","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.transformer.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.transformer.__init__#L177-L228","kind":"function","name":"__init__","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":177,"end_line":228,"context_start_line":157,"context_end_line":248,"code":"            )\n\n    def forward(self, x: torch.Tensor, attn_mask: torch.Tensor):\n        if self.layer_scale_type is None:\n            x = x + self.drop_path(self.attn(self.norm_1(x), attn_mask))\n            x = x + self.drop_path(self.mlp(self.norm_2(x)))\n        else:\n            x = (\n                x\n                + self.drop_path(self.attn(self.norm_1(x), attn_mask))\n                * self.layer_scale_gamma1\n            )\n            x = x + self.drop_path(self.mlp(self.norm_2(x))) * self.layer_scale_gamma2\n        return x\n\n\n_LAYER_NORM = partial(nn.LayerNorm, eps=1e-6)\n\n\nclass SimpleTransformer(nn.Module):\n    def __init__(\n        self,\n        attn_target: Callable,\n        embed_dim: int,\n        num_blocks: int,\n        block: Callable = BlockWithMasking,\n        pre_transformer_layer: Optional[Callable] = None,\n        post_transformer_layer: Optional[Callable] = None,\n        drop_path_rate: float = 0.0,\n        drop_path_type: str = \"progressive\",\n        norm_layer: Callable = _LAYER_NORM,\n        mlp_ratio: int = 4,\n        ffn_dropout_rate: float = 0.0,\n        layer_scale_type: Optional[str] = None,  # from cait; possible values are None, \"per_channel\", \"scalar\"\n        layer_scale_init_value: float = 1e-4,  # from cait; float\n        weight_init_style: str = \"jax\",  # possible values jax or pytorch\n    ):\n        \"\"\"\n        Simple Transformer with the following features\n        1. Supports masked attention\n        2. Supports DropPath\n        3. Supports LayerScale\n        4. Supports Dropout in Attention and FFN\n        5. Makes few assumptions about the input except that it is a Tensor\n        \"\"\"\n        super().__init__()\n        self.pre_transformer_layer = pre_transformer_layer\n        if drop_path_type == \"progressive\":\n            dpr = [x.item() for x in torch.linspace(0, drop_path_rate, num_blocks)]\n        elif drop_path_type == \"uniform\":\n            dpr = [drop_path_rate for i in range(num_blocks)]\n        else:\n            raise ValueError(f\"Unknown drop_path_type: {drop_path_type}\")\n\n        self.blocks = nn.Sequential(\n            *[\n                block(\n                    dim=embed_dim,\n                    attn_target=attn_target,\n                    mlp_ratio=mlp_ratio,\n                    ffn_dropout_rate=ffn_dropout_rate,\n                    drop_path=dpr[i],\n                    norm_layer=norm_layer,\n                    layer_scale_type=layer_scale_type,\n                    layer_scale_init_value=layer_scale_init_value,\n                )\n                for i in range(num_blocks)\n            ]\n        )\n        self.post_transformer_layer = post_transformer_layer\n        self.weight_init_style = weight_init_style\n        self.apply(self._init_weights)\n\n    def _init_weights(self, m):\n        if isinstance(m, nn.Linear):\n            if self.weight_init_style == \"jax\":\n                # Based on MAE and official Jax ViT implementation\n                torch.nn.init.xavier_uniform_(m.weight)\n            elif self.weight_init_style == \"pytorch\":\n                # PyTorch ViT uses trunc_normal_\n                trunc_normal_(m.weight, std=0.02)\n\n            if m.bias is not None:\n                nn.init.constant_(m.bias, 0)\n        elif isinstance(m, (nn.LayerNorm)):\n            nn.init.constant_(m.bias, 0)\n            nn.init.constant_(m.weight, 1.0)\n\n    def forward(\n        self,\n        tokens: torch.Tensor,\n        attn_mask: torch.Tensor = None,","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.transformer.forward","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.transformer.forward#L245-L280","kind":"function","name":"forward","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":245,"end_line":280,"context_start_line":225,"context_end_line":280,"code":"        )\n        self.post_transformer_layer = post_transformer_layer\n        self.weight_init_style = weight_init_style\n        self.apply(self._init_weights)\n\n    def _init_weights(self, m):\n        if isinstance(m, nn.Linear):\n            if self.weight_init_style == \"jax\":\n                # Based on MAE and official Jax ViT implementation\n                torch.nn.init.xavier_uniform_(m.weight)\n            elif self.weight_init_style == \"pytorch\":\n                # PyTorch ViT uses trunc_normal_\n                trunc_normal_(m.weight, std=0.02)\n\n            if m.bias is not None:\n                nn.init.constant_(m.bias, 0)\n        elif isinstance(m, (nn.LayerNorm)):\n            nn.init.constant_(m.bias, 0)\n            nn.init.constant_(m.weight, 1.0)\n\n    def forward(\n        self,\n        tokens: torch.Tensor,\n        attn_mask: torch.Tensor = None,\n        use_checkpoint: bool = False,\n        checkpoint_every_n: int = 1,\n        checkpoint_blk_ids: Optional[List[int]] = None,\n    ):\n        \"\"\"\n        Inputs\n        - tokens: data of shape N x L x D (or L x N x D depending on the attention implementation)\n        - attn: mask of shape L x L\n\n        Output\n        - x: data of shape N x L x D (or L x N x D depending on the attention implementation)\n        \"\"\"\n        if self.pre_transformer_layer:\n            tokens = self.pre_transformer_layer(tokens)\n        if use_checkpoint and checkpoint_blk_ids is None:\n            checkpoint_blk_ids = [\n                blk_id\n                for blk_id in range(len(self.blocks))\n                if blk_id % checkpoint_every_n == 0\n            ]\n        if checkpoint_blk_ids:\n            checkpoint_blk_ids = set(checkpoint_blk_ids)\n        for blk_id, blk in enumerate(self.blocks):\n            if use_checkpoint and blk_id in checkpoint_blk_ids:\n                tokens = checkpoint.checkpoint(\n                    blk, tokens, attn_mask, use_reentrant=False\n                )\n            else:\n                tokens = blk(tokens, attn_mask=attn_mask)\n        if self.post_transformer_layer:\n            tokens = self.post_transformer_layer(tokens)\n        return tokens","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.transformer._init_weights","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.transformer._init_weights#L230-L243","kind":"function","name":"_init_weights","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":230,"end_line":243,"context_start_line":210,"context_end_line":263,"code":"\n        self.blocks = nn.Sequential(\n            *[\n                block(\n                    dim=embed_dim,\n                    attn_target=attn_target,\n                    mlp_ratio=mlp_ratio,\n                    ffn_dropout_rate=ffn_dropout_rate,\n                    drop_path=dpr[i],\n                    norm_layer=norm_layer,\n                    layer_scale_type=layer_scale_type,\n                    layer_scale_init_value=layer_scale_init_value,\n                )\n                for i in range(num_blocks)\n            ]\n        )\n        self.post_transformer_layer = post_transformer_layer\n        self.weight_init_style = weight_init_style\n        self.apply(self._init_weights)\n\n    def _init_weights(self, m):\n        if isinstance(m, nn.Linear):\n            if self.weight_init_style == \"jax\":\n                # Based on MAE and official Jax ViT implementation\n                torch.nn.init.xavier_uniform_(m.weight)\n            elif self.weight_init_style == \"pytorch\":\n                # PyTorch ViT uses trunc_normal_\n                trunc_normal_(m.weight, std=0.02)\n\n            if m.bias is not None:\n                nn.init.constant_(m.bias, 0)\n        elif isinstance(m, (nn.LayerNorm)):\n            nn.init.constant_(m.bias, 0)\n            nn.init.constant_(m.weight, 1.0)\n\n    def forward(\n        self,\n        tokens: torch.Tensor,\n        attn_mask: torch.Tensor = None,\n        use_checkpoint: bool = False,\n        checkpoint_every_n: int = 1,\n        checkpoint_blk_ids: Optional[List[int]] = None,\n    ):\n        \"\"\"\n        Inputs\n        - tokens: data of shape N x L x D (or L x N x D depending on the attention implementation)\n        - attn: mask of shape L x L\n\n        Output\n        - x: data of shape N x L x D (or L x N x D depending on the attention implementation)\n        \"\"\"\n        if self.pre_transformer_layer:\n            tokens = self.pre_transformer_layer(tokens)\n        if use_checkpoint and checkpoint_blk_ids is None:","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.models.multimodal_preprocessors#L1-L685","kind":"module","name":"imagebind_LLM.ImageBind.models.multimodal_preprocessors","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":1,"end_line":685,"context_start_line":1,"context_end_line":685,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport gzip\nimport html\nimport io\nimport math\nfrom functools import lru_cache\nfrom typing import Callable, List, Optional, Tuple\n\nimport ftfy\nimport numpy as np\nimport regex as re\nimport torch\nimport torch.nn as nn\nfrom iopath.common.file_io import g_pathmgr\nfrom timm.models.layers import trunc_normal_\n\nfrom .helpers import VerboseNNModule, cast_if_src_dtype\n\n\ndef get_sinusoid_encoding_table(n_position, d_hid):\n    \"\"\"Sinusoid position encoding table\"\"\"\n\n    # TODO: make it with torch instead of numpy\n    def get_position_angle_vec(position):\n        return [\n            position / np.power(10000, 2 * (hid_j // 2) / d_hid)\n            for hid_j in range(d_hid)\n        ]\n\n    sinusoid_table = np.array(\n        [get_position_angle_vec(pos_i) for pos_i in range(n_position)]\n    )\n    sinusoid_table[:, 0::2] = np.sin(sinusoid_table[:, 0::2])  # dim 2i\n    sinusoid_table[:, 1::2] = np.cos(sinusoid_table[:, 1::2])  # dim 2i+1\n\n    return torch.FloatTensor(sinusoid_table).unsqueeze(0)\n\n\ndef interpolate_pos_encoding_2d(target_spatial_size, pos_embed):\n    N = pos_embed.shape[1]\n    if N == target_spatial_size:\n        return pos_embed\n    dim = pos_embed.shape[-1]\n    # nn.functional.interpolate doesn't work with bfloat16 so we cast to float32\n    pos_embed, updated = cast_if_src_dtype(pos_embed, torch.bfloat16, torch.float32)\n    pos_embed = nn.functional.interpolate(\n        pos_embed.reshape(1, int(math.sqrt(N)), int(math.sqrt(N)), dim).permute(\n            0, 3, 1, 2\n        ),\n        scale_factor=math.sqrt(target_spatial_size / N),\n        mode=\"bicubic\",\n    )\n    if updated:\n        pos_embed, _ = cast_if_src_dtype(pos_embed, torch.float32, torch.bfloat16)\n    pos_embed = pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)\n    return pos_embed\n\n\ndef interpolate_pos_encoding(\n    npatch_per_img,\n    pos_embed,\n    patches_layout,\n    input_shape=None,\n    first_patch_idx=1,\n):\n    assert first_patch_idx == 0 or first_patch_idx == 1, \"there is 1 CLS token or none\"\n    N = pos_embed.shape[1] - first_patch_idx  # since it's 1 if cls_token exists\n    if npatch_per_img == N:\n        return pos_embed\n\n    assert (\n        patches_layout[-1] == patches_layout[-2]\n    ), \"Interpolation of pos embed not supported for non-square layouts\"\n\n    class_emb = pos_embed[:, :first_patch_idx]\n    pos_embed = pos_embed[:, first_patch_idx:]\n\n    if input_shape is None or patches_layout[0] == 1:\n        # simple 2D pos embedding, no temporal component\n        pos_embed = interpolate_pos_encoding_2d(npatch_per_img, pos_embed)\n    elif patches_layout[0] > 1:\n        # pos embed has a temporal component\n        assert len(input_shape) == 4, \"temporal interpolation not supported\"\n        # we only support 2D interpolation in this case\n        num_frames = patches_layout[0]\n        num_spatial_tokens = patches_layout[1] * patches_layout[2]\n        pos_embed = pos_embed.view(1, num_frames, num_spatial_tokens, -1)\n        # interpolate embedding for zeroth frame\n        pos_embed = interpolate_pos_encoding_2d(\n            npatch_per_img, pos_embed[0, 0, ...].unsqueeze(0)\n        )\n    else:\n        raise ValueError(\"This type of interpolation isn't implemented\")\n\n    return torch.cat((class_emb, pos_embed), dim=1)\n\n\ndef _get_pos_embedding(\n    npatch_per_img,\n    pos_embed,\n    patches_layout,\n    input_shape,\n    first_patch_idx=1,\n):\n    pos_embed = interpolate_pos_encoding(\n        npatch_per_img,\n        pos_embed,\n        patches_layout,\n        input_shape=input_shape,\n        first_patch_idx=first_patch_idx,\n    )\n    return pos_embed\n\n\nclass PatchEmbedGeneric(nn.Module):\n    \"\"\"\n    PatchEmbed from Hydra\n    \"\"\"\n\n    def __init__(self, proj_stem, norm_layer: Optional[nn.Module] = None):\n        super().__init__()\n\n        if len(proj_stem) > 1:\n            self.proj = nn.Sequential(*proj_stem)\n        else:\n            # Special case to be able to load pre-trained models that were\n            # trained with a standard stem\n            self.proj = proj_stem[0]\n        self.norm_layer = norm_layer\n\n    def get_patch_layout(self, img_size):\n        with torch.no_grad():\n            dummy_img = torch.zeros(\n                [\n                    1,\n                ]\n                + img_size\n            )\n            dummy_out = self.proj(dummy_img)\n        embed_dim = dummy_out.shape[1]\n        patches_layout = tuple(dummy_out.shape[2:])\n        num_patches = np.prod(patches_layout)\n        return patches_layout, num_patches, embed_dim\n\n    def forward(self, x):\n        x = self.proj(x)\n        # B C (T) H W -> B (T)HW C\n        x = x.flatten(2).transpose(1, 2)\n        if self.norm_layer is not None:\n            x = self.norm_layer(x)\n        return x\n\n\nclass SpatioTemporalPosEmbeddingHelper(VerboseNNModule):\n    def __init__(\n        self,\n        patches_layout: List,\n        num_patches: int,\n        num_cls_tokens: int,\n        embed_dim: int,\n        learnable: bool,\n    ) -> None:\n        super().__init__()\n        self.num_cls_tokens = num_cls_tokens\n        self.patches_layout = patches_layout\n        self.num_patches = num_patches\n        self.num_tokens = num_cls_tokens + num_patches\n        self.learnable = learnable\n        if self.learnable:\n            self.pos_embed = nn.Parameter(torch.zeros(1, self.num_tokens, embed_dim))\n            trunc_normal_(self.pos_embed, std=0.02)\n        else:\n            self.register_buffer(\n                \"pos_embed\", get_sinusoid_encoding_table(self.num_tokens, embed_dim)\n            )\n\n    def get_pos_embedding(self, vision_input, all_vision_tokens):\n        input_shape = vision_input.shape\n        pos_embed = _get_pos_embedding(\n            all_vision_tokens.size(1) - self.num_cls_tokens,\n            pos_embed=self.pos_embed,\n            patches_layout=self.patches_layout,\n            input_shape=input_shape,\n            first_patch_idx=self.num_cls_tokens,\n        )\n        return pos_embed\n\n\nclass RGBDTPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        rgbt_stem: PatchEmbedGeneric,\n        depth_stem: Optional[PatchEmbedGeneric],\n        img_size: Tuple = (3, 224, 224),\n        num_cls_tokens: int = 1,\n        pos_embed_fn: Optional[Callable] = None,\n        use_type_embed: bool = False,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        stem = rgbt_stem if rgbt_stem is not None else depth_stem\n        (\n            self.patches_layout,\n            self.num_patches,\n            self.embed_dim,\n        ) = stem.get_patch_layout(img_size)\n        self.rgbt_stem = rgbt_stem\n        self.depth_stem = depth_stem\n        self.use_pos_embed = pos_embed_fn is not None\n        self.use_type_embed = use_type_embed\n        self.num_cls_tokens = num_cls_tokens\n\n        if self.use_pos_embed:\n            self.pos_embedding_helper = pos_embed_fn(\n                patches_layout=self.patches_layout,\n                num_cls_tokens=num_cls_tokens,\n                num_patches=self.num_patches,\n                embed_dim=self.embed_dim,\n            )\n        if self.num_cls_tokens > 0:\n            self.cls_token = nn.Parameter(\n                torch.zeros(1, self.num_cls_tokens, self.embed_dim)\n            )\n        if self.use_type_embed:\n            self.type_embed = nn.Parameter(torch.zeros(1, 1, self.embed_dim))\n\n        self.init_parameters(init_param_style)\n\n    @torch.no_grad()\n    def init_parameters(self, init_param_style):\n        if init_param_style == \"openclip\":\n            # OpenCLIP style initialization\n            scale = self.embed_dim**-0.5\n            if self.use_pos_embed:\n                nn.init.normal_(self.pos_embedding_helper.pos_embed)\n                self.pos_embedding_helper.pos_embed *= scale\n\n            if self.num_cls_tokens > 0:\n                nn.init.normal_(self.cls_token)\n                self.cls_token *= scale\n        elif init_param_style == \"vit\":\n            self.cls_token.data.fill_(0)\n        else:\n            raise ValueError(f\"Unknown init {init_param_style}\")\n\n        if self.use_type_embed:\n            nn.init.normal_(self.type_embed)\n\n    def tokenize_input_and_cls_pos(self, input, stem, mask):\n        # tokens is of shape B x L x D\n        tokens = stem(input)\n        assert tokens.ndim == 3\n        assert tokens.shape[2] == self.embed_dim\n        B = tokens.shape[0]\n        if self.num_cls_tokens > 0:\n            class_tokens = self.cls_token.expand(\n                B, -1, -1\n            )  # stole class_tokens impl from Phil Wang, thanks\n            tokens = torch.cat((class_tokens, tokens), dim=1)\n        if self.use_pos_embed:\n            pos_embed = self.pos_embedding_helper.get_pos_embedding(input, tokens)\n            tokens = tokens + pos_embed\n        if self.use_type_embed:\n            tokens = tokens + self.type_embed.expand(B, -1, -1)\n        return tokens\n\n    def forward(self, vision=None, depth=None, patch_mask=None):\n        if patch_mask is not None:\n            raise NotImplementedError()\n\n        if vision is not None:\n            vision_tokens = self.tokenize_input_and_cls_pos(\n                vision, self.rgbt_stem, patch_mask\n            )\n\n        if depth is not None:\n            depth_tokens = self.tokenize_input_and_cls_pos(\n                depth, self.depth_stem, patch_mask\n            )\n\n        # aggregate tokens\n        if vision is not None and depth is not None:\n            final_tokens = vision_tokens + depth_tokens\n        else:\n            final_tokens = vision_tokens if vision is not None else depth_tokens\n        return_dict = {\n            \"trunk\": {\n                \"tokens\": final_tokens,\n            },\n            \"head\": {},\n        }\n        return return_dict\n\n\nclass AudioPreprocessor(RGBDTPreprocessor):\n    def __init__(self, audio_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=audio_stem, depth_stem=None, **kwargs)\n\n    def forward(self, audio=None):\n        return super().forward(vision=audio)\n\n\nclass ThermalPreprocessor(RGBDTPreprocessor):\n    def __init__(self, thermal_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=thermal_stem, depth_stem=None, **kwargs)\n\n    def forward(self, thermal=None):\n        return super().forward(vision=thermal)\n\n\ndef build_causal_attention_mask(context_length):\n    # lazily create causal attention mask, with full attention between the vision tokens\n    # pytorch uses additive attention mask; fill with -inf\n    mask = torch.empty(context_length, context_length, requires_grad=False)\n    mask.fill_(float(\"-inf\"))\n    mask.triu_(1)  # zero out the lower diagonal\n    return mask\n\n\nclass TextPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        vocab_size: int,\n        context_length: int,\n        embed_dim: int,\n        causal_masking: bool,\n        supply_seq_len_to_head: bool = True,\n        num_cls_tokens: int = 0,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        self.vocab_size = vocab_size\n        self.context_length = context_length\n        self.token_embedding = nn.Embedding(vocab_size, embed_dim)\n        self.pos_embed = nn.Parameter(\n            torch.empty(1, self.context_length + num_cls_tokens, embed_dim)\n        )\n        self.causal_masking = causal_masking\n        if self.causal_masking:\n            mask = build_causal_attention_mask(self.context_length)\n            # register the mask as a buffer so it can be moved to the right device\n            self.register_buffer(\"mask\", mask)\n\n        self.supply_seq_len_to_head = supply_seq_len_to_head\n        self.num_cls_tokens = num_cls_tokens\n        self.embed_dim = embed_dim\n        if num_cls_tokens > 0:\n            assert self.causal_masking is False, \"Masking + CLS token isn't implemented\"\n            self.cls_token = nn.Parameter(\n                torch.zeros(1, self.num_cls_tokens, embed_dim)\n            )\n\n        self.init_parameters(init_param_style)\n\n    @torch.no_grad()\n    def init_parameters(self, init_param_style=\"openclip\"):\n        # OpenCLIP style initialization\n        nn.init.normal_(self.token_embedding.weight, std=0.02)\n        nn.init.normal_(self.pos_embed, std=0.01)\n\n        if init_param_style == \"openclip\":\n            # OpenCLIP style initialization\n            scale = self.embed_dim**-0.5\n            if self.num_cls_tokens > 0:\n                nn.init.normal_(self.cls_token)\n                self.cls_token *= scale\n        elif init_param_style == \"vit\":\n            self.cls_token.data.fill_(0)\n        else:\n            raise ValueError(f\"Unknown init {init_param_style}\")\n\n    def forward(self, text):\n        # text tokens are of shape B x L x D\n        text_tokens = self.token_embedding(text)\n        # concat CLS tokens if any\n        if self.num_cls_tokens > 0:\n            B = text_tokens.shape[0]\n            class_tokens = self.cls_token.expand(\n                B, -1, -1\n            )  # stole class_tokens impl from Phil Wang, thanks\n            text_tokens = torch.cat((class_tokens, text_tokens), dim=1)\n        text_tokens = text_tokens + self.pos_embed\n        return_dict = {\n            \"trunk\": {\n                \"tokens\": text_tokens,\n            },\n            \"head\": {},\n        }\n        # Compute sequence length after adding CLS tokens\n        if self.supply_seq_len_to_head:\n            text_lengths = text.argmax(dim=-1)\n            return_dict[\"head\"] = {\n                \"seq_len\": text_lengths,\n            }\n        if self.causal_masking:\n            return_dict[\"trunk\"].update({\"attn_mask\": self.mask})\n        return return_dict\n\n\nclass Im2Video(nn.Module):\n    \"\"\"Convert an image into a trivial video.\"\"\"\n\n    def __init__(self, time_dim=2):\n        super().__init__()\n        self.time_dim = time_dim\n\n    def forward(self, x):\n        if x.ndim == 4:\n            # B, C, H, W -> B, C, T, H, W\n            return x.unsqueeze(self.time_dim)\n        elif x.ndim == 5:\n            return x\n        else:\n            raise ValueError(f\"Dimension incorrect {x.shape}\")\n\n\nclass PadIm2Video(Im2Video):\n    def __init__(self, ntimes, pad_type, time_dim=2):\n        super().__init__(time_dim=time_dim)\n        assert ntimes > 0\n        assert pad_type in [\"zero\", \"repeat\"]\n        self.ntimes = ntimes\n        self.pad_type = pad_type\n\n    def forward(self, x):\n        x = super().forward(x)\n        if x.shape[self.time_dim] == 1:\n            if self.pad_type == \"repeat\":\n                new_shape = [1] * len(x.shape)\n                new_shape[self.time_dim] = self.ntimes\n                x = x.repeat(new_shape)\n            elif self.pad_type == \"zero\":\n                padarg = [0, 0] * len(x.shape)\n                padarg[2 * self.time_dim + 1] = self.ntimes - x.shape[self.time_dim]\n                x = nn.functional.pad(x, padarg)\n        return x\n\n\n# Modified from github.com/openai/CLIP\n@lru_cache()\ndef bytes_to_unicode():\n    \"\"\"\n    Returns list of utf-8 byte and a corresponding list of unicode strings.\n    The reversible bpe codes work on unicode strings.\n    This means you need a large # of unicode characters in your vocab if you want to avoid UNKs.\n    When you're at something like a 10B token dataset you end up needing around 5K for decent coverage.\n    This is a signficant percentage of your normal, say, 32K bpe vocab.\n    To avoid that, we want lookup tables between utf-8 bytes and unicode strings.\n    And avoids mapping to whitespace/control characters the bpe code barfs on.\n    \"\"\"\n    bs = (\n        list(range(ord(\"!\"), ord(\"~\") + 1))\n        + list(range(ord(\"¡\"), ord(\"¬\") + 1))\n        + list(range(ord(\"®\"), ord(\"ÿ\") + 1))\n    )\n    cs = bs[:]\n    n = 0\n    for b in range(2**8):\n        if b not in bs:\n            bs.append(b)\n            cs.append(2**8 + n)\n            n += 1\n    cs = [chr(n) for n in cs]\n    return dict(zip(bs, cs))\n\n\ndef get_pairs(word):\n    \"\"\"Return set of symbol pairs in a word.\n    Word is represented as tuple of symbols (symbols being variable-length strings).\n    \"\"\"\n    pairs = set()\n    prev_char = word[0]\n    for char in word[1:]:\n        pairs.add((prev_char, char))\n        prev_char = char\n    return pairs\n\n\ndef basic_clean(text):\n    text = ftfy.fix_text(text)\n    text = html.unescape(html.unescape(text))\n    return text.strip()\n\n\ndef whitespace_clean(text):\n    text = re.sub(r\"\\s+\", \" \", text)\n    text = text.strip()\n    return text\n\n\nclass SimpleTokenizer(object):\n    def __init__(self, bpe_path: str, context_length=77):\n        self.byte_encoder = bytes_to_unicode()\n        self.byte_decoder = {v: k for k, v in self.byte_encoder.items()}\n\n        with g_pathmgr.open(bpe_path, \"rb\") as fh:\n            bpe_bytes = io.BytesIO(fh.read())\n            merges: List[str] = gzip.open(bpe_bytes).read().decode(\"utf-8\").split(\"\\n\")\n        merges = merges[1 : 49152 - 256 - 2 + 1]\n        merges: List[Tuple[str, ...]] = [tuple(merge.split()) for merge in merges]\n        vocab = list(bytes_to_unicode().values())\n        vocab = vocab + [v + \"</w>\" for v in vocab]\n        for merge in merges:\n            vocab.append(\"\".join(merge))\n        vocab.extend([\"<|startoftext|>\", \"<|endoftext|>\"])\n        self.encoder = dict(zip(vocab, range(len(vocab))))\n        self.decoder = {v: k for k, v in self.encoder.items()}\n        self.bpe_ranks = dict(zip(merges, range(len(merges))))\n        self.cache = {\n            \"<|startoftext|>\": \"<|startoftext|>\",\n            \"<|endoftext|>\": \"<|endoftext|>\",\n        }\n        self.pat = re.compile(\n            r\"\"\"<\\|startoftext\\|>|<\\|endoftext\\|>|'s|'t|'re|'ve|'m|'ll|'d|[\\p{L}]+|[\\p{N}]|[^\\s\\p{L}\\p{N}]+\"\"\",\n            re.IGNORECASE,\n        )\n        self.context_length = context_length\n\n    def bpe(self, token):\n        if token in self.cache:\n            return self.cache[token]\n        word = tuple(token[:-1]) + (token[-1] + \"</w>\",)\n        pairs = get_pairs(word)\n\n        if not pairs:\n            return token + \"</w>\"\n\n        while True:\n            bigram = min(pairs, key=lambda pair: self.bpe_ranks.get(pair, float(\"inf\")))\n            if bigram not in self.bpe_ranks:\n                break\n            first, second = bigram\n            new_word = []\n            i = 0\n            while i < len(word):\n                try:\n                    j = word.index(first, i)\n                    new_word.extend(word[i:j])\n                    i = j\n                except:\n                    new_word.extend(word[i:])\n                    break\n\n                if word[i] == first and i < len(word) - 1 and word[i + 1] == second:\n                    new_word.append(first + second)\n                    i += 2\n                else:\n                    new_word.append(word[i])\n                    i += 1\n            new_word = tuple(new_word)\n            word = new_word\n            if len(word) == 1:\n                break\n            else:\n                pairs = get_pairs(word)\n        word = \" \".join(word)\n        self.cache[token] = word\n        return word\n\n    def encode(self, text):\n        bpe_tokens = []\n        text = whitespace_clean(basic_clean(text)).lower()\n        for token in re.findall(self.pat, text):\n            token = \"\".join(self.byte_encoder[b] for b in token.encode(\"utf-8\"))\n            bpe_tokens.extend(\n                self.encoder[bpe_token] for bpe_token in self.bpe(token).split(\" \")\n            )\n        return bpe_tokens\n\n    def decode(self, tokens):\n        text = \"\".join([self.decoder[token] for token in tokens])\n        text = (\n            bytearray([self.byte_decoder[c] for c in text])\n            .decode(\"utf-8\", errors=\"replace\")\n            .replace(\"</w>\", \" \")\n        )\n        return text\n\n    def __call__(self, texts, context_length=None):\n        if not context_length:\n            context_length = self.conte\n# ... truncated ...","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":true}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.get_sinusoid_encoding_table","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.get_sinusoid_encoding_table#L26-L42","kind":"function","name":"get_sinusoid_encoding_table","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":26,"end_line":42,"context_start_line":6,"context_end_line":62,"code":"# LICENSE file in the root directory of this source tree.\n\nimport gzip\nimport html\nimport io\nimport math\nfrom functools import lru_cache\nfrom typing import Callable, List, Optional, Tuple\n\nimport ftfy\nimport numpy as np\nimport regex as re\nimport torch\nimport torch.nn as nn\nfrom iopath.common.file_io import g_pathmgr\nfrom timm.models.layers import trunc_normal_\n\nfrom .helpers import VerboseNNModule, cast_if_src_dtype\n\n\ndef get_sinusoid_encoding_table(n_position, d_hid):\n    \"\"\"Sinusoid position encoding table\"\"\"\n\n    # TODO: make it with torch instead of numpy\n    def get_position_angle_vec(position):\n        return [\n            position / np.power(10000, 2 * (hid_j // 2) / d_hid)\n            for hid_j in range(d_hid)\n        ]\n\n    sinusoid_table = np.array(\n        [get_position_angle_vec(pos_i) for pos_i in range(n_position)]\n    )\n    sinusoid_table[:, 0::2] = np.sin(sinusoid_table[:, 0::2])  # dim 2i\n    sinusoid_table[:, 1::2] = np.cos(sinusoid_table[:, 1::2])  # dim 2i+1\n\n    return torch.FloatTensor(sinusoid_table).unsqueeze(0)\n\n\ndef interpolate_pos_encoding_2d(target_spatial_size, pos_embed):\n    N = pos_embed.shape[1]\n    if N == target_spatial_size:\n        return pos_embed\n    dim = pos_embed.shape[-1]\n    # nn.functional.interpolate doesn't work with bfloat16 so we cast to float32\n    pos_embed, updated = cast_if_src_dtype(pos_embed, torch.bfloat16, torch.float32)\n    pos_embed = nn.functional.interpolate(\n        pos_embed.reshape(1, int(math.sqrt(N)), int(math.sqrt(N)), dim).permute(\n            0, 3, 1, 2\n        ),\n        scale_factor=math.sqrt(target_spatial_size / N),\n        mode=\"bicubic\",\n    )\n    if updated:\n        pos_embed, _ = cast_if_src_dtype(pos_embed, torch.float32, torch.bfloat16)\n    pos_embed = pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)\n    return pos_embed","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.interpolate_pos_encoding_2d","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.interpolate_pos_encoding_2d#L45-L62","kind":"function","name":"interpolate_pos_encoding_2d","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":45,"end_line":62,"context_start_line":25,"context_end_line":82,"code":"\ndef get_sinusoid_encoding_table(n_position, d_hid):\n    \"\"\"Sinusoid position encoding table\"\"\"\n\n    # TODO: make it with torch instead of numpy\n    def get_position_angle_vec(position):\n        return [\n            position / np.power(10000, 2 * (hid_j // 2) / d_hid)\n            for hid_j in range(d_hid)\n        ]\n\n    sinusoid_table = np.array(\n        [get_position_angle_vec(pos_i) for pos_i in range(n_position)]\n    )\n    sinusoid_table[:, 0::2] = np.sin(sinusoid_table[:, 0::2])  # dim 2i\n    sinusoid_table[:, 1::2] = np.cos(sinusoid_table[:, 1::2])  # dim 2i+1\n\n    return torch.FloatTensor(sinusoid_table).unsqueeze(0)\n\n\ndef interpolate_pos_encoding_2d(target_spatial_size, pos_embed):\n    N = pos_embed.shape[1]\n    if N == target_spatial_size:\n        return pos_embed\n    dim = pos_embed.shape[-1]\n    # nn.functional.interpolate doesn't work with bfloat16 so we cast to float32\n    pos_embed, updated = cast_if_src_dtype(pos_embed, torch.bfloat16, torch.float32)\n    pos_embed = nn.functional.interpolate(\n        pos_embed.reshape(1, int(math.sqrt(N)), int(math.sqrt(N)), dim).permute(\n            0, 3, 1, 2\n        ),\n        scale_factor=math.sqrt(target_spatial_size / N),\n        mode=\"bicubic\",\n    )\n    if updated:\n        pos_embed, _ = cast_if_src_dtype(pos_embed, torch.float32, torch.bfloat16)\n    pos_embed = pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)\n    return pos_embed\n\n\ndef interpolate_pos_encoding(\n    npatch_per_img,\n    pos_embed,\n    patches_layout,\n    input_shape=None,\n    first_patch_idx=1,\n):\n    assert first_patch_idx == 0 or first_patch_idx == 1, \"there is 1 CLS token or none\"\n    N = pos_embed.shape[1] - first_patch_idx  # since it's 1 if cls_token exists\n    if npatch_per_img == N:\n        return pos_embed\n\n    assert (\n        patches_layout[-1] == patches_layout[-2]\n    ), \"Interpolation of pos embed not supported for non-square layouts\"\n\n    class_emb = pos_embed[:, :first_patch_idx]\n    pos_embed = pos_embed[:, first_patch_idx:]","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.interpolate_pos_encoding","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.interpolate_pos_encoding#L65-L101","kind":"function","name":"interpolate_pos_encoding","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":65,"end_line":101,"context_start_line":45,"context_end_line":121,"code":"def interpolate_pos_encoding_2d(target_spatial_size, pos_embed):\n    N = pos_embed.shape[1]\n    if N == target_spatial_size:\n        return pos_embed\n    dim = pos_embed.shape[-1]\n    # nn.functional.interpolate doesn't work with bfloat16 so we cast to float32\n    pos_embed, updated = cast_if_src_dtype(pos_embed, torch.bfloat16, torch.float32)\n    pos_embed = nn.functional.interpolate(\n        pos_embed.reshape(1, int(math.sqrt(N)), int(math.sqrt(N)), dim).permute(\n            0, 3, 1, 2\n        ),\n        scale_factor=math.sqrt(target_spatial_size / N),\n        mode=\"bicubic\",\n    )\n    if updated:\n        pos_embed, _ = cast_if_src_dtype(pos_embed, torch.float32, torch.bfloat16)\n    pos_embed = pos_embed.permute(0, 2, 3, 1).view(1, -1, dim)\n    return pos_embed\n\n\ndef interpolate_pos_encoding(\n    npatch_per_img,\n    pos_embed,\n    patches_layout,\n    input_shape=None,\n    first_patch_idx=1,\n):\n    assert first_patch_idx == 0 or first_patch_idx == 1, \"there is 1 CLS token or none\"\n    N = pos_embed.shape[1] - first_patch_idx  # since it's 1 if cls_token exists\n    if npatch_per_img == N:\n        return pos_embed\n\n    assert (\n        patches_layout[-1] == patches_layout[-2]\n    ), \"Interpolation of pos embed not supported for non-square layouts\"\n\n    class_emb = pos_embed[:, :first_patch_idx]\n    pos_embed = pos_embed[:, first_patch_idx:]\n\n    if input_shape is None or patches_layout[0] == 1:\n        # simple 2D pos embedding, no temporal component\n        pos_embed = interpolate_pos_encoding_2d(npatch_per_img, pos_embed)\n    elif patches_layout[0] > 1:\n        # pos embed has a temporal component\n        assert len(input_shape) == 4, \"temporal interpolation not supported\"\n        # we only support 2D interpolation in this case\n        num_frames = patches_layout[0]\n        num_spatial_tokens = patches_layout[1] * patches_layout[2]\n        pos_embed = pos_embed.view(1, num_frames, num_spatial_tokens, -1)\n        # interpolate embedding for zeroth frame\n        pos_embed = interpolate_pos_encoding_2d(\n            npatch_per_img, pos_embed[0, 0, ...].unsqueeze(0)\n        )\n    else:\n        raise ValueError(\"This type of interpolation isn't implemented\")\n\n    return torch.cat((class_emb, pos_embed), dim=1)\n\n\ndef _get_pos_embedding(\n    npatch_per_img,\n    pos_embed,\n    patches_layout,\n    input_shape,\n    first_patch_idx=1,\n):\n    pos_embed = interpolate_pos_encoding(\n        npatch_per_img,\n        pos_embed,\n        patches_layout,\n        input_shape=input_shape,\n        first_patch_idx=first_patch_idx,\n    )\n    return pos_embed\n\n\nclass PatchEmbedGeneric(nn.Module):","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors._get_pos_embedding","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors._get_pos_embedding#L104-L118","kind":"function","name":"_get_pos_embedding","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":104,"end_line":118,"context_start_line":84,"context_end_line":138,"code":"    if input_shape is None or patches_layout[0] == 1:\n        # simple 2D pos embedding, no temporal component\n        pos_embed = interpolate_pos_encoding_2d(npatch_per_img, pos_embed)\n    elif patches_layout[0] > 1:\n        # pos embed has a temporal component\n        assert len(input_shape) == 4, \"temporal interpolation not supported\"\n        # we only support 2D interpolation in this case\n        num_frames = patches_layout[0]\n        num_spatial_tokens = patches_layout[1] * patches_layout[2]\n        pos_embed = pos_embed.view(1, num_frames, num_spatial_tokens, -1)\n        # interpolate embedding for zeroth frame\n        pos_embed = interpolate_pos_encoding_2d(\n            npatch_per_img, pos_embed[0, 0, ...].unsqueeze(0)\n        )\n    else:\n        raise ValueError(\"This type of interpolation isn't implemented\")\n\n    return torch.cat((class_emb, pos_embed), dim=1)\n\n\ndef _get_pos_embedding(\n    npatch_per_img,\n    pos_embed,\n    patches_layout,\n    input_shape,\n    first_patch_idx=1,\n):\n    pos_embed = interpolate_pos_encoding(\n        npatch_per_img,\n        pos_embed,\n        patches_layout,\n        input_shape=input_shape,\n        first_patch_idx=first_patch_idx,\n    )\n    return pos_embed\n\n\nclass PatchEmbedGeneric(nn.Module):\n    \"\"\"\n    PatchEmbed from Hydra\n    \"\"\"\n\n    def __init__(self, proj_stem, norm_layer: Optional[nn.Module] = None):\n        super().__init__()\n\n        if len(proj_stem) > 1:\n            self.proj = nn.Sequential(*proj_stem)\n        else:\n            # Special case to be able to load pre-trained models that were\n            # trained with a standard stem\n            self.proj = proj_stem[0]\n        self.norm_layer = norm_layer\n\n    def get_patch_layout(self, img_size):\n        with torch.no_grad():","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.PatchEmbedGeneric","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.multimodal_preprocessors.PatchEmbedGeneric#L121-L157","kind":"class","name":"PatchEmbedGeneric","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":121,"end_line":157,"context_start_line":101,"context_end_line":177,"code":"    return torch.cat((class_emb, pos_embed), dim=1)\n\n\ndef _get_pos_embedding(\n    npatch_per_img,\n    pos_embed,\n    patches_layout,\n    input_shape,\n    first_patch_idx=1,\n):\n    pos_embed = interpolate_pos_encoding(\n        npatch_per_img,\n        pos_embed,\n        patches_layout,\n        input_shape=input_shape,\n        first_patch_idx=first_patch_idx,\n    )\n    return pos_embed\n\n\nclass PatchEmbedGeneric(nn.Module):\n    \"\"\"\n    PatchEmbed from Hydra\n    \"\"\"\n\n    def __init__(self, proj_stem, norm_layer: Optional[nn.Module] = None):\n        super().__init__()\n\n        if len(proj_stem) > 1:\n            self.proj = nn.Sequential(*proj_stem)\n        else:\n            # Special case to be able to load pre-trained models that were\n            # trained with a standard stem\n            self.proj = proj_stem[0]\n        self.norm_layer = norm_layer\n\n    def get_patch_layout(self, img_size):\n        with torch.no_grad():\n            dummy_img = torch.zeros(\n                [\n                    1,\n                ]\n                + img_size\n            )\n            dummy_out = self.proj(dummy_img)\n        embed_dim = dummy_out.shape[1]\n        patches_layout = tuple(dummy_out.shape[2:])\n        num_patches = np.prod(patches_layout)\n        return patches_layout, num_patches, embed_dim\n\n    def forward(self, x):\n        x = self.proj(x)\n        # B C (T) H W -> B (T)HW C\n        x = x.flatten(2).transpose(1, 2)\n        if self.norm_layer is not None:\n            x = self.norm_layer(x)\n        return x\n\n\nclass SpatioTemporalPosEmbeddingHelper(VerboseNNModule):\n    def __init__(\n        self,\n        patches_layout: List,\n        num_patches: int,\n        num_cls_tokens: int,\n        embed_dim: int,\n        learnable: bool,\n    ) -> None:\n        super().__init__()\n        self.num_cls_tokens = num_cls_tokens\n        self.patches_layout = patches_layout\n        self.num_patches = num_patches\n        self.num_tokens = num_cls_tokens + num_patches\n        self.learnable = learnable\n        if self.learnable:\n            self.pos_embed = nn.Parameter(torch.zeros(1, self.num_tokens, embed_dim))\n            trunc_normal_(self.pos_embed, std=0.02)","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.SpatioTemporalPosEmbeddingHelper","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.multimodal_preprocessors.SpatioTemporalPosEmbeddingHelper#L160-L192","kind":"class","name":"SpatioTemporalPosEmbeddingHelper","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":160,"end_line":192,"context_start_line":140,"context_end_line":212,"code":"                [\n                    1,\n                ]\n                + img_size\n            )\n            dummy_out = self.proj(dummy_img)\n        embed_dim = dummy_out.shape[1]\n        patches_layout = tuple(dummy_out.shape[2:])\n        num_patches = np.prod(patches_layout)\n        return patches_layout, num_patches, embed_dim\n\n    def forward(self, x):\n        x = self.proj(x)\n        # B C (T) H W -> B (T)HW C\n        x = x.flatten(2).transpose(1, 2)\n        if self.norm_layer is not None:\n            x = self.norm_layer(x)\n        return x\n\n\nclass SpatioTemporalPosEmbeddingHelper(VerboseNNModule):\n    def __init__(\n        self,\n        patches_layout: List,\n        num_patches: int,\n        num_cls_tokens: int,\n        embed_dim: int,\n        learnable: bool,\n    ) -> None:\n        super().__init__()\n        self.num_cls_tokens = num_cls_tokens\n        self.patches_layout = patches_layout\n        self.num_patches = num_patches\n        self.num_tokens = num_cls_tokens + num_patches\n        self.learnable = learnable\n        if self.learnable:\n            self.pos_embed = nn.Parameter(torch.zeros(1, self.num_tokens, embed_dim))\n            trunc_normal_(self.pos_embed, std=0.02)\n        else:\n            self.register_buffer(\n                \"pos_embed\", get_sinusoid_encoding_table(self.num_tokens, embed_dim)\n            )\n\n    def get_pos_embedding(self, vision_input, all_vision_tokens):\n        input_shape = vision_input.shape\n        pos_embed = _get_pos_embedding(\n            all_vision_tokens.size(1) - self.num_cls_tokens,\n            pos_embed=self.pos_embed,\n            patches_layout=self.patches_layout,\n            input_shape=input_shape,\n            first_patch_idx=self.num_cls_tokens,\n        )\n        return pos_embed\n\n\nclass RGBDTPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        rgbt_stem: PatchEmbedGeneric,\n        depth_stem: Optional[PatchEmbedGeneric],\n        img_size: Tuple = (3, 224, 224),\n        num_cls_tokens: int = 1,\n        pos_embed_fn: Optional[Callable] = None,\n        use_type_embed: bool = False,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        stem = rgbt_stem if rgbt_stem is not None else depth_stem\n        (\n            self.patches_layout,\n            self.num_patches,\n            self.embed_dim,\n        ) = stem.get_patch_layout(img_size)","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.RGBDTPreprocessor","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.multimodal_preprocessors.RGBDTPreprocessor#L195-L298","kind":"class","name":"RGBDTPreprocessor","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":195,"end_line":298,"context_start_line":175,"context_end_line":318,"code":"        if self.learnable:\n            self.pos_embed = nn.Parameter(torch.zeros(1, self.num_tokens, embed_dim))\n            trunc_normal_(self.pos_embed, std=0.02)\n        else:\n            self.register_buffer(\n                \"pos_embed\", get_sinusoid_encoding_table(self.num_tokens, embed_dim)\n            )\n\n    def get_pos_embedding(self, vision_input, all_vision_tokens):\n        input_shape = vision_input.shape\n        pos_embed = _get_pos_embedding(\n            all_vision_tokens.size(1) - self.num_cls_tokens,\n            pos_embed=self.pos_embed,\n            patches_layout=self.patches_layout,\n            input_shape=input_shape,\n            first_patch_idx=self.num_cls_tokens,\n        )\n        return pos_embed\n\n\nclass RGBDTPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        rgbt_stem: PatchEmbedGeneric,\n        depth_stem: Optional[PatchEmbedGeneric],\n        img_size: Tuple = (3, 224, 224),\n        num_cls_tokens: int = 1,\n        pos_embed_fn: Optional[Callable] = None,\n        use_type_embed: bool = False,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        stem = rgbt_stem if rgbt_stem is not None else depth_stem\n        (\n            self.patches_layout,\n            self.num_patches,\n            self.embed_dim,\n        ) = stem.get_patch_layout(img_size)\n        self.rgbt_stem = rgbt_stem\n        self.depth_stem = depth_stem\n        self.use_pos_embed = pos_embed_fn is not None\n        self.use_type_embed = use_type_embed\n        self.num_cls_tokens = num_cls_tokens\n\n        if self.use_pos_embed:\n            self.pos_embedding_helper = pos_embed_fn(\n                patches_layout=self.patches_layout,\n                num_cls_tokens=num_cls_tokens,\n                num_patches=self.num_patches,\n                embed_dim=self.embed_dim,\n            )\n        if self.num_cls_tokens > 0:\n            self.cls_token = nn.Parameter(\n                torch.zeros(1, self.num_cls_tokens, self.embed_dim)\n            )\n        if self.use_type_embed:\n            self.type_embed = nn.Parameter(torch.zeros(1, 1, self.embed_dim))\n\n        self.init_parameters(init_param_style)\n\n    @torch.no_grad()\n    def init_parameters(self, init_param_style):\n        if init_param_style == \"openclip\":\n            # OpenCLIP style initialization\n            scale = self.embed_dim**-0.5\n            if self.use_pos_embed:\n                nn.init.normal_(self.pos_embedding_helper.pos_embed)\n                self.pos_embedding_helper.pos_embed *= scale\n\n            if self.num_cls_tokens > 0:\n                nn.init.normal_(self.cls_token)\n                self.cls_token *= scale\n        elif init_param_style == \"vit\":\n            self.cls_token.data.fill_(0)\n        else:\n            raise ValueError(f\"Unknown init {init_param_style}\")\n\n        if self.use_type_embed:\n            nn.init.normal_(self.type_embed)\n\n    def tokenize_input_and_cls_pos(self, input, stem, mask):\n        # tokens is of shape B x L x D\n        tokens = stem(input)\n        assert tokens.ndim == 3\n        assert tokens.shape[2] == self.embed_dim\n        B = tokens.shape[0]\n        if self.num_cls_tokens > 0:\n            class_tokens = self.cls_token.expand(\n                B, -1, -1\n            )  # stole class_tokens impl from Phil Wang, thanks\n            tokens = torch.cat((class_tokens, tokens), dim=1)\n        if self.use_pos_embed:\n            pos_embed = self.pos_embedding_helper.get_pos_embedding(input, tokens)\n            tokens = tokens + pos_embed\n        if self.use_type_embed:\n            tokens = tokens + self.type_embed.expand(B, -1, -1)\n        return tokens\n\n    def forward(self, vision=None, depth=None, patch_mask=None):\n        if patch_mask is not None:\n            raise NotImplementedError()\n\n        if vision is not None:\n            vision_tokens = self.tokenize_input_and_cls_pos(\n                vision, self.rgbt_stem, patch_mask\n            )\n\n        if depth is not None:\n            depth_tokens = self.tokenize_input_and_cls_pos(\n                depth, self.depth_stem, patch_mask\n            )\n\n        # aggregate tokens\n        if vision is not None and depth is not None:\n            final_tokens = vision_tokens + depth_tokens\n        else:\n            final_tokens = vision_tokens if vision is not None else depth_tokens\n        return_dict = {\n            \"trunk\": {\n                \"tokens\": final_tokens,\n            },\n            \"head\": {},\n        }\n        return return_dict\n\n\nclass AudioPreprocessor(RGBDTPreprocessor):\n    def __init__(self, audio_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=audio_stem, depth_stem=None, **kwargs)\n\n    def forward(self, audio=None):\n        return super().forward(vision=audio)\n\n\nclass ThermalPreprocessor(RGBDTPreprocessor):\n    def __init__(self, thermal_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=thermal_stem, depth_stem=None, **kwargs)\n\n    def forward(self, thermal=None):\n        return super().forward(vision=thermal)\n\n\ndef build_causal_attention_mask(context_length):\n    # lazily create causal attention mask, with full attention between the vision tokens","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.AudioPreprocessor","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.multimodal_preprocessors.AudioPreprocessor#L301-L306","kind":"class","name":"AudioPreprocessor","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":301,"end_line":306,"context_start_line":281,"context_end_line":326,"code":"\n        if depth is not None:\n            depth_tokens = self.tokenize_input_and_cls_pos(\n                depth, self.depth_stem, patch_mask\n            )\n\n        # aggregate tokens\n        if vision is not None and depth is not None:\n            final_tokens = vision_tokens + depth_tokens\n        else:\n            final_tokens = vision_tokens if vision is not None else depth_tokens\n        return_dict = {\n            \"trunk\": {\n                \"tokens\": final_tokens,\n            },\n            \"head\": {},\n        }\n        return return_dict\n\n\nclass AudioPreprocessor(RGBDTPreprocessor):\n    def __init__(self, audio_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=audio_stem, depth_stem=None, **kwargs)\n\n    def forward(self, audio=None):\n        return super().forward(vision=audio)\n\n\nclass ThermalPreprocessor(RGBDTPreprocessor):\n    def __init__(self, thermal_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=thermal_stem, depth_stem=None, **kwargs)\n\n    def forward(self, thermal=None):\n        return super().forward(vision=thermal)\n\n\ndef build_causal_attention_mask(context_length):\n    # lazily create causal attention mask, with full attention between the vision tokens\n    # pytorch uses additive attention mask; fill with -inf\n    mask = torch.empty(context_length, context_length, requires_grad=False)\n    mask.fill_(float(\"-inf\"))\n    mask.triu_(1)  # zero out the lower diagonal\n    return mask\n\n\nclass TextPreprocessor(VerboseNNModule):","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.ThermalPreprocessor","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.multimodal_preprocessors.ThermalPreprocessor#L309-L314","kind":"class","name":"ThermalPreprocessor","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":309,"end_line":314,"context_start_line":289,"context_end_line":334,"code":"            final_tokens = vision_tokens + depth_tokens\n        else:\n            final_tokens = vision_tokens if vision is not None else depth_tokens\n        return_dict = {\n            \"trunk\": {\n                \"tokens\": final_tokens,\n            },\n            \"head\": {},\n        }\n        return return_dict\n\n\nclass AudioPreprocessor(RGBDTPreprocessor):\n    def __init__(self, audio_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=audio_stem, depth_stem=None, **kwargs)\n\n    def forward(self, audio=None):\n        return super().forward(vision=audio)\n\n\nclass ThermalPreprocessor(RGBDTPreprocessor):\n    def __init__(self, thermal_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=thermal_stem, depth_stem=None, **kwargs)\n\n    def forward(self, thermal=None):\n        return super().forward(vision=thermal)\n\n\ndef build_causal_attention_mask(context_length):\n    # lazily create causal attention mask, with full attention between the vision tokens\n    # pytorch uses additive attention mask; fill with -inf\n    mask = torch.empty(context_length, context_length, requires_grad=False)\n    mask.fill_(float(\"-inf\"))\n    mask.triu_(1)  # zero out the lower diagonal\n    return mask\n\n\nclass TextPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        vocab_size: int,\n        context_length: int,\n        embed_dim: int,\n        causal_masking: bool,\n        supply_seq_len_to_head: bool = True,\n        num_cls_tokens: int = 0,","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.build_causal_attention_mask","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.build_causal_attention_mask#L317-L323","kind":"function","name":"build_causal_attention_mask","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":317,"end_line":323,"context_start_line":297,"context_end_line":343,"code":"        }\n        return return_dict\n\n\nclass AudioPreprocessor(RGBDTPreprocessor):\n    def __init__(self, audio_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=audio_stem, depth_stem=None, **kwargs)\n\n    def forward(self, audio=None):\n        return super().forward(vision=audio)\n\n\nclass ThermalPreprocessor(RGBDTPreprocessor):\n    def __init__(self, thermal_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=thermal_stem, depth_stem=None, **kwargs)\n\n    def forward(self, thermal=None):\n        return super().forward(vision=thermal)\n\n\ndef build_causal_attention_mask(context_length):\n    # lazily create causal attention mask, with full attention between the vision tokens\n    # pytorch uses additive attention mask; fill with -inf\n    mask = torch.empty(context_length, context_length, requires_grad=False)\n    mask.fill_(float(\"-inf\"))\n    mask.triu_(1)  # zero out the lower diagonal\n    return mask\n\n\nclass TextPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        vocab_size: int,\n        context_length: int,\n        embed_dim: int,\n        causal_masking: bool,\n        supply_seq_len_to_head: bool = True,\n        num_cls_tokens: int = 0,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        self.vocab_size = vocab_size\n        self.context_length = context_length\n        self.token_embedding = nn.Embedding(vocab_size, embed_dim)\n        self.pos_embed = nn.Parameter(\n            torch.empty(1, self.context_length + num_cls_tokens, embed_dim)\n        )","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.TextPreprocessor","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.multimodal_preprocessors.TextPreprocessor#L326-L403","kind":"class","name":"TextPreprocessor","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":326,"end_line":403,"context_start_line":306,"context_end_line":423,"code":"        return super().forward(vision=audio)\n\n\nclass ThermalPreprocessor(RGBDTPreprocessor):\n    def __init__(self, thermal_stem: PatchEmbedGeneric, **kwargs) -> None:\n        super().__init__(rgbt_stem=thermal_stem, depth_stem=None, **kwargs)\n\n    def forward(self, thermal=None):\n        return super().forward(vision=thermal)\n\n\ndef build_causal_attention_mask(context_length):\n    # lazily create causal attention mask, with full attention between the vision tokens\n    # pytorch uses additive attention mask; fill with -inf\n    mask = torch.empty(context_length, context_length, requires_grad=False)\n    mask.fill_(float(\"-inf\"))\n    mask.triu_(1)  # zero out the lower diagonal\n    return mask\n\n\nclass TextPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        vocab_size: int,\n        context_length: int,\n        embed_dim: int,\n        causal_masking: bool,\n        supply_seq_len_to_head: bool = True,\n        num_cls_tokens: int = 0,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        self.vocab_size = vocab_size\n        self.context_length = context_length\n        self.token_embedding = nn.Embedding(vocab_size, embed_dim)\n        self.pos_embed = nn.Parameter(\n            torch.empty(1, self.context_length + num_cls_tokens, embed_dim)\n        )\n        self.causal_masking = causal_masking\n        if self.causal_masking:\n            mask = build_causal_attention_mask(self.context_length)\n            # register the mask as a buffer so it can be moved to the right device\n            self.register_buffer(\"mask\", mask)\n\n        self.supply_seq_len_to_head = supply_seq_len_to_head\n        self.num_cls_tokens = num_cls_tokens\n        self.embed_dim = embed_dim\n        if num_cls_tokens > 0:\n            assert self.causal_masking is False, \"Masking + CLS token isn't implemented\"\n            self.cls_token = nn.Parameter(\n                torch.zeros(1, self.num_cls_tokens, embed_dim)\n            )\n\n        self.init_parameters(init_param_style)\n\n    @torch.no_grad()\n    def init_parameters(self, init_param_style=\"openclip\"):\n        # OpenCLIP style initialization\n        nn.init.normal_(self.token_embedding.weight, std=0.02)\n        nn.init.normal_(self.pos_embed, std=0.01)\n\n        if init_param_style == \"openclip\":\n            # OpenCLIP style initialization\n            scale = self.embed_dim**-0.5\n            if self.num_cls_tokens > 0:\n                nn.init.normal_(self.cls_token)\n                self.cls_token *= scale\n        elif init_param_style == \"vit\":\n            self.cls_token.data.fill_(0)\n        else:\n            raise ValueError(f\"Unknown init {init_param_style}\")\n\n    def forward(self, text):\n        # text tokens are of shape B x L x D\n        text_tokens = self.token_embedding(text)\n        # concat CLS tokens if any\n        if self.num_cls_tokens > 0:\n            B = text_tokens.shape[0]\n            class_tokens = self.cls_token.expand(\n                B, -1, -1\n            )  # stole class_tokens impl from Phil Wang, thanks\n            text_tokens = torch.cat((class_tokens, text_tokens), dim=1)\n        text_tokens = text_tokens + self.pos_embed\n        return_dict = {\n            \"trunk\": {\n                \"tokens\": text_tokens,\n            },\n            \"head\": {},\n        }\n        # Compute sequence length after adding CLS tokens\n        if self.supply_seq_len_to_head:\n            text_lengths = text.argmax(dim=-1)\n            return_dict[\"head\"] = {\n                \"seq_len\": text_lengths,\n            }\n        if self.causal_masking:\n            return_dict[\"trunk\"].update({\"attn_mask\": self.mask})\n        return return_dict\n\n\nclass Im2Video(nn.Module):\n    \"\"\"Convert an image into a trivial video.\"\"\"\n\n    def __init__(self, time_dim=2):\n        super().__init__()\n        self.time_dim = time_dim\n\n    def forward(self, x):\n        if x.ndim == 4:\n            # B, C, H, W -> B, C, T, H, W\n            return x.unsqueeze(self.time_dim)\n        elif x.ndim == 5:\n            return x\n        else:\n            raise ValueError(f\"Dimension incorrect {x.shape}\")\n\n\nclass PadIm2Video(Im2Video):","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.Im2Video","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.multimodal_preprocessors.Im2Video#L406-L420","kind":"class","name":"Im2Video","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":406,"end_line":420,"context_start_line":386,"context_end_line":440,"code":"            )  # stole class_tokens impl from Phil Wang, thanks\n            text_tokens = torch.cat((class_tokens, text_tokens), dim=1)\n        text_tokens = text_tokens + self.pos_embed\n        return_dict = {\n            \"trunk\": {\n                \"tokens\": text_tokens,\n            },\n            \"head\": {},\n        }\n        # Compute sequence length after adding CLS tokens\n        if self.supply_seq_len_to_head:\n            text_lengths = text.argmax(dim=-1)\n            return_dict[\"head\"] = {\n                \"seq_len\": text_lengths,\n            }\n        if self.causal_masking:\n            return_dict[\"trunk\"].update({\"attn_mask\": self.mask})\n        return return_dict\n\n\nclass Im2Video(nn.Module):\n    \"\"\"Convert an image into a trivial video.\"\"\"\n\n    def __init__(self, time_dim=2):\n        super().__init__()\n        self.time_dim = time_dim\n\n    def forward(self, x):\n        if x.ndim == 4:\n            # B, C, H, W -> B, C, T, H, W\n            return x.unsqueeze(self.time_dim)\n        elif x.ndim == 5:\n            return x\n        else:\n            raise ValueError(f\"Dimension incorrect {x.shape}\")\n\n\nclass PadIm2Video(Im2Video):\n    def __init__(self, ntimes, pad_type, time_dim=2):\n        super().__init__(time_dim=time_dim)\n        assert ntimes > 0\n        assert pad_type in [\"zero\", \"repeat\"]\n        self.ntimes = ntimes\n        self.pad_type = pad_type\n\n    def forward(self, x):\n        x = super().forward(x)\n        if x.shape[self.time_dim] == 1:\n            if self.pad_type == \"repeat\":\n                new_shape = [1] * len(x.shape)\n                new_shape[self.time_dim] = self.ntimes\n                x = x.repeat(new_shape)\n            elif self.pad_type == \"zero\":\n                padarg = [0, 0] * len(x.shape)\n                padarg[2 * self.time_dim + 1] = self.ntimes - x.shape[self.time_dim]","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.PadIm2Video","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.multimodal_preprocessors.PadIm2Video#L423-L442","kind":"class","name":"PadIm2Video","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":423,"end_line":442,"context_start_line":403,"context_end_line":462,"code":"        return return_dict\n\n\nclass Im2Video(nn.Module):\n    \"\"\"Convert an image into a trivial video.\"\"\"\n\n    def __init__(self, time_dim=2):\n        super().__init__()\n        self.time_dim = time_dim\n\n    def forward(self, x):\n        if x.ndim == 4:\n            # B, C, H, W -> B, C, T, H, W\n            return x.unsqueeze(self.time_dim)\n        elif x.ndim == 5:\n            return x\n        else:\n            raise ValueError(f\"Dimension incorrect {x.shape}\")\n\n\nclass PadIm2Video(Im2Video):\n    def __init__(self, ntimes, pad_type, time_dim=2):\n        super().__init__(time_dim=time_dim)\n        assert ntimes > 0\n        assert pad_type in [\"zero\", \"repeat\"]\n        self.ntimes = ntimes\n        self.pad_type = pad_type\n\n    def forward(self, x):\n        x = super().forward(x)\n        if x.shape[self.time_dim] == 1:\n            if self.pad_type == \"repeat\":\n                new_shape = [1] * len(x.shape)\n                new_shape[self.time_dim] = self.ntimes\n                x = x.repeat(new_shape)\n            elif self.pad_type == \"zero\":\n                padarg = [0, 0] * len(x.shape)\n                padarg[2 * self.time_dim + 1] = self.ntimes - x.shape[self.time_dim]\n                x = nn.functional.pad(x, padarg)\n        return x\n\n\n# Modified from github.com/openai/CLIP\n@lru_cache()\ndef bytes_to_unicode():\n    \"\"\"\n    Returns list of utf-8 byte and a corresponding list of unicode strings.\n    The reversible bpe codes work on unicode strings.\n    This means you need a large # of unicode characters in your vocab if you want to avoid UNKs.\n    When you're at something like a 10B token dataset you end up needing around 5K for decent coverage.\n    This is a signficant percentage of your normal, say, 32K bpe vocab.\n    To avoid that, we want lookup tables between utf-8 bytes and unicode strings.\n    And avoids mapping to whitespace/control characters the bpe code barfs on.\n    \"\"\"\n    bs = (\n        list(range(ord(\"!\"), ord(\"~\") + 1))\n        + list(range(ord(\"¡\"), ord(\"¬\") + 1))\n        + list(range(ord(\"®\"), ord(\"ÿ\") + 1))\n    )\n    cs = bs[:]","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.bytes_to_unicode","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.bytes_to_unicode#L447-L470","kind":"function","name":"bytes_to_unicode","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":447,"end_line":470,"context_start_line":427,"context_end_line":490,"code":"        assert pad_type in [\"zero\", \"repeat\"]\n        self.ntimes = ntimes\n        self.pad_type = pad_type\n\n    def forward(self, x):\n        x = super().forward(x)\n        if x.shape[self.time_dim] == 1:\n            if self.pad_type == \"repeat\":\n                new_shape = [1] * len(x.shape)\n                new_shape[self.time_dim] = self.ntimes\n                x = x.repeat(new_shape)\n            elif self.pad_type == \"zero\":\n                padarg = [0, 0] * len(x.shape)\n                padarg[2 * self.time_dim + 1] = self.ntimes - x.shape[self.time_dim]\n                x = nn.functional.pad(x, padarg)\n        return x\n\n\n# Modified from github.com/openai/CLIP\n@lru_cache()\ndef bytes_to_unicode():\n    \"\"\"\n    Returns list of utf-8 byte and a corresponding list of unicode strings.\n    The reversible bpe codes work on unicode strings.\n    This means you need a large # of unicode characters in your vocab if you want to avoid UNKs.\n    When you're at something like a 10B token dataset you end up needing around 5K for decent coverage.\n    This is a signficant percentage of your normal, say, 32K bpe vocab.\n    To avoid that, we want lookup tables between utf-8 bytes and unicode strings.\n    And avoids mapping to whitespace/control characters the bpe code barfs on.\n    \"\"\"\n    bs = (\n        list(range(ord(\"!\"), ord(\"~\") + 1))\n        + list(range(ord(\"¡\"), ord(\"¬\") + 1))\n        + list(range(ord(\"®\"), ord(\"ÿ\") + 1))\n    )\n    cs = bs[:]\n    n = 0\n    for b in range(2**8):\n        if b not in bs:\n            bs.append(b)\n            cs.append(2**8 + n)\n            n += 1\n    cs = [chr(n) for n in cs]\n    return dict(zip(bs, cs))\n\n\ndef get_pairs(word):\n    \"\"\"Return set of symbol pairs in a word.\n    Word is represented as tuple of symbols (symbols being variable-length strings).\n    \"\"\"\n    pairs = set()\n    prev_char = word[0]\n    for char in word[1:]:\n        pairs.add((prev_char, char))\n        prev_char = char\n    return pairs\n\n\ndef basic_clean(text):\n    text = ftfy.fix_text(text)\n    text = html.unescape(html.unescape(text))\n    return text.strip()\n\n","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.get_pairs","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.get_pairs#L473-L482","kind":"function","name":"get_pairs","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":473,"end_line":482,"context_start_line":453,"context_end_line":502,"code":"    This is a signficant percentage of your normal, say, 32K bpe vocab.\n    To avoid that, we want lookup tables between utf-8 bytes and unicode strings.\n    And avoids mapping to whitespace/control characters the bpe code barfs on.\n    \"\"\"\n    bs = (\n        list(range(ord(\"!\"), ord(\"~\") + 1))\n        + list(range(ord(\"¡\"), ord(\"¬\") + 1))\n        + list(range(ord(\"®\"), ord(\"ÿ\") + 1))\n    )\n    cs = bs[:]\n    n = 0\n    for b in range(2**8):\n        if b not in bs:\n            bs.append(b)\n            cs.append(2**8 + n)\n            n += 1\n    cs = [chr(n) for n in cs]\n    return dict(zip(bs, cs))\n\n\ndef get_pairs(word):\n    \"\"\"Return set of symbol pairs in a word.\n    Word is represented as tuple of symbols (symbols being variable-length strings).\n    \"\"\"\n    pairs = set()\n    prev_char = word[0]\n    for char in word[1:]:\n        pairs.add((prev_char, char))\n        prev_char = char\n    return pairs\n\n\ndef basic_clean(text):\n    text = ftfy.fix_text(text)\n    text = html.unescape(html.unescape(text))\n    return text.strip()\n\n\ndef whitespace_clean(text):\n    text = re.sub(r\"\\s+\", \" \", text)\n    text = text.strip()\n    return text\n\n\nclass SimpleTokenizer(object):\n    def __init__(self, bpe_path: str, context_length=77):\n        self.byte_encoder = bytes_to_unicode()\n        self.byte_decoder = {v: k for k, v in self.byte_encoder.items()}\n\n        with g_pathmgr.open(bpe_path, \"rb\") as fh:","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.basic_clean","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.basic_clean#L485-L488","kind":"function","name":"basic_clean","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":485,"end_line":488,"context_start_line":465,"context_end_line":508,"code":"        if b not in bs:\n            bs.append(b)\n            cs.append(2**8 + n)\n            n += 1\n    cs = [chr(n) for n in cs]\n    return dict(zip(bs, cs))\n\n\ndef get_pairs(word):\n    \"\"\"Return set of symbol pairs in a word.\n    Word is represented as tuple of symbols (symbols being variable-length strings).\n    \"\"\"\n    pairs = set()\n    prev_char = word[0]\n    for char in word[1:]:\n        pairs.add((prev_char, char))\n        prev_char = char\n    return pairs\n\n\ndef basic_clean(text):\n    text = ftfy.fix_text(text)\n    text = html.unescape(html.unescape(text))\n    return text.strip()\n\n\ndef whitespace_clean(text):\n    text = re.sub(r\"\\s+\", \" \", text)\n    text = text.strip()\n    return text\n\n\nclass SimpleTokenizer(object):\n    def __init__(self, bpe_path: str, context_length=77):\n        self.byte_encoder = bytes_to_unicode()\n        self.byte_decoder = {v: k for k, v in self.byte_encoder.items()}\n\n        with g_pathmgr.open(bpe_path, \"rb\") as fh:\n            bpe_bytes = io.BytesIO(fh.read())\n            merges: List[str] = gzip.open(bpe_bytes).read().decode(\"utf-8\").split(\"\\n\")\n        merges = merges[1 : 49152 - 256 - 2 + 1]\n        merges: List[Tuple[str, ...]] = [tuple(merge.split()) for merge in merges]\n        vocab = list(bytes_to_unicode().values())\n        vocab = vocab + [v + \"</w>\" for v in vocab]","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.whitespace_clean","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.whitespace_clean#L491-L494","kind":"function","name":"whitespace_clean","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":491,"end_line":494,"context_start_line":471,"context_end_line":514,"code":"\n\ndef get_pairs(word):\n    \"\"\"Return set of symbol pairs in a word.\n    Word is represented as tuple of symbols (symbols being variable-length strings).\n    \"\"\"\n    pairs = set()\n    prev_char = word[0]\n    for char in word[1:]:\n        pairs.add((prev_char, char))\n        prev_char = char\n    return pairs\n\n\ndef basic_clean(text):\n    text = ftfy.fix_text(text)\n    text = html.unescape(html.unescape(text))\n    return text.strip()\n\n\ndef whitespace_clean(text):\n    text = re.sub(r\"\\s+\", \" \", text)\n    text = text.strip()\n    return text\n\n\nclass SimpleTokenizer(object):\n    def __init__(self, bpe_path: str, context_length=77):\n        self.byte_encoder = bytes_to_unicode()\n        self.byte_decoder = {v: k for k, v in self.byte_encoder.items()}\n\n        with g_pathmgr.open(bpe_path, \"rb\") as fh:\n            bpe_bytes = io.BytesIO(fh.read())\n            merges: List[str] = gzip.open(bpe_bytes).read().decode(\"utf-8\").split(\"\\n\")\n        merges = merges[1 : 49152 - 256 - 2 + 1]\n        merges: List[Tuple[str, ...]] = [tuple(merge.split()) for merge in merges]\n        vocab = list(bytes_to_unicode().values())\n        vocab = vocab + [v + \"</w>\" for v in vocab]\n        for merge in merges:\n            vocab.append(\"\".join(merge))\n        vocab.extend([\"<|startoftext|>\", \"<|endoftext|>\"])\n        self.encoder = dict(zip(vocab, range(len(vocab))))\n        self.decoder = {v: k for k, v in self.encoder.items()}\n        self.bpe_ranks = dict(zip(merges, range(len(merges))))","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.SimpleTokenizer","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.multimodal_preprocessors.SimpleTokenizer#L497-L603","kind":"class","name":"SimpleTokenizer","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":497,"end_line":603,"context_start_line":477,"context_end_line":623,"code":"    pairs = set()\n    prev_char = word[0]\n    for char in word[1:]:\n        pairs.add((prev_char, char))\n        prev_char = char\n    return pairs\n\n\ndef basic_clean(text):\n    text = ftfy.fix_text(text)\n    text = html.unescape(html.unescape(text))\n    return text.strip()\n\n\ndef whitespace_clean(text):\n    text = re.sub(r\"\\s+\", \" \", text)\n    text = text.strip()\n    return text\n\n\nclass SimpleTokenizer(object):\n    def __init__(self, bpe_path: str, context_length=77):\n        self.byte_encoder = bytes_to_unicode()\n        self.byte_decoder = {v: k for k, v in self.byte_encoder.items()}\n\n        with g_pathmgr.open(bpe_path, \"rb\") as fh:\n            bpe_bytes = io.BytesIO(fh.read())\n            merges: List[str] = gzip.open(bpe_bytes).read().decode(\"utf-8\").split(\"\\n\")\n        merges = merges[1 : 49152 - 256 - 2 + 1]\n        merges: List[Tuple[str, ...]] = [tuple(merge.split()) for merge in merges]\n        vocab = list(bytes_to_unicode().values())\n        vocab = vocab + [v + \"</w>\" for v in vocab]\n        for merge in merges:\n            vocab.append(\"\".join(merge))\n        vocab.extend([\"<|startoftext|>\", \"<|endoftext|>\"])\n        self.encoder = dict(zip(vocab, range(len(vocab))))\n        self.decoder = {v: k for k, v in self.encoder.items()}\n        self.bpe_ranks = dict(zip(merges, range(len(merges))))\n        self.cache = {\n            \"<|startoftext|>\": \"<|startoftext|>\",\n            \"<|endoftext|>\": \"<|endoftext|>\",\n        }\n        self.pat = re.compile(\n            r\"\"\"<\\|startoftext\\|>|<\\|endoftext\\|>|'s|'t|'re|'ve|'m|'ll|'d|[\\p{L}]+|[\\p{N}]|[^\\s\\p{L}\\p{N}]+\"\"\",\n            re.IGNORECASE,\n        )\n        self.context_length = context_length\n\n    def bpe(self, token):\n        if token in self.cache:\n            return self.cache[token]\n        word = tuple(token[:-1]) + (token[-1] + \"</w>\",)\n        pairs = get_pairs(word)\n\n        if not pairs:\n            return token + \"</w>\"\n\n        while True:\n            bigram = min(pairs, key=lambda pair: self.bpe_ranks.get(pair, float(\"inf\")))\n            if bigram not in self.bpe_ranks:\n                break\n            first, second = bigram\n            new_word = []\n            i = 0\n            while i < len(word):\n                try:\n                    j = word.index(first, i)\n                    new_word.extend(word[i:j])\n                    i = j\n                except:\n                    new_word.extend(word[i:])\n                    break\n\n                if word[i] == first and i < len(word) - 1 and word[i + 1] == second:\n                    new_word.append(first + second)\n                    i += 2\n                else:\n                    new_word.append(word[i])\n                    i += 1\n            new_word = tuple(new_word)\n            word = new_word\n            if len(word) == 1:\n                break\n            else:\n                pairs = get_pairs(word)\n        word = \" \".join(word)\n        self.cache[token] = word\n        return word\n\n    def encode(self, text):\n        bpe_tokens = []\n        text = whitespace_clean(basic_clean(text)).lower()\n        for token in re.findall(self.pat, text):\n            token = \"\".join(self.byte_encoder[b] for b in token.encode(\"utf-8\"))\n            bpe_tokens.extend(\n                self.encoder[bpe_token] for bpe_token in self.bpe(token).split(\" \")\n            )\n        return bpe_tokens\n\n    def decode(self, tokens):\n        text = \"\".join([self.decoder[token] for token in tokens])\n        text = (\n            bytearray([self.byte_decoder[c] for c in text])\n            .decode(\"utf-8\", errors=\"replace\")\n            .replace(\"</w>\", \" \")\n        )\n        return text\n\n    def __call__(self, texts, context_length=None):\n        if not context_length:\n            context_length = self.context_length\n\n        if isinstance(texts, str):\n            texts = [texts]\n\n        sot_token = self.encoder[\"<|startoftext|>\"]\n        eot_token = self.encoder[\"<|endoftext|>\"]\n        all_tokens = [[sot_token] + self.encode(text) + [eot_token] for text in texts]\n        result = torch.zeros(len(all_tokens), context_length, dtype=torch.long)\n\n        for i, tokens in enumerate(all_tokens):\n            tokens = tokens[:context_length]\n            result[i, : len(tokens)] = torch.tensor(tokens)\n\n        if len(result) == 1:\n            return result[0]\n        return result\n\n\nclass IMUPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        kernel_size: int,\n        imu_stem: PatchEmbedGeneric,\n        embed_dim: int,\n        img_size: Tuple = (6, 2000),\n        num_cls_tokens: int = 1,\n        pos_embed_fn: Optional[Callable] = None,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        self.imu_stem = imu_stem\n        self.embed_dim = embed_dim\n        self.use_pos_embed = pos_embed_fn is not None\n        self.num_cls_tokens = num_cls_tokens\n        self.kernel_size = kernel_size\n        self.pos_embed = nn.Parameter(","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.IMUPreprocessor","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.multimodal_preprocessors.IMUPreprocessor#L606-L685","kind":"class","name":"IMUPreprocessor","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":606,"end_line":685,"context_start_line":586,"context_end_line":685,"code":"        if not context_length:\n            context_length = self.context_length\n\n        if isinstance(texts, str):\n            texts = [texts]\n\n        sot_token = self.encoder[\"<|startoftext|>\"]\n        eot_token = self.encoder[\"<|endoftext|>\"]\n        all_tokens = [[sot_token] + self.encode(text) + [eot_token] for text in texts]\n        result = torch.zeros(len(all_tokens), context_length, dtype=torch.long)\n\n        for i, tokens in enumerate(all_tokens):\n            tokens = tokens[:context_length]\n            result[i, : len(tokens)] = torch.tensor(tokens)\n\n        if len(result) == 1:\n            return result[0]\n        return result\n\n\nclass IMUPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        kernel_size: int,\n        imu_stem: PatchEmbedGeneric,\n        embed_dim: int,\n        img_size: Tuple = (6, 2000),\n        num_cls_tokens: int = 1,\n        pos_embed_fn: Optional[Callable] = None,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        self.imu_stem = imu_stem\n        self.embed_dim = embed_dim\n        self.use_pos_embed = pos_embed_fn is not None\n        self.num_cls_tokens = num_cls_tokens\n        self.kernel_size = kernel_size\n        self.pos_embed = nn.Parameter(\n            torch.empty(1, (img_size[1] // kernel_size) + num_cls_tokens, embed_dim)\n        )\n\n        if self.num_cls_tokens > 0:\n            self.cls_token = nn.Parameter(\n                torch.zeros(1, self.num_cls_tokens, self.embed_dim)\n            )\n\n        self.init_parameters(init_param_style)\n\n    @torch.no_grad()\n    def init_parameters(self, init_param_style):\n        nn.init.normal_(self.pos_embed, std=0.01)\n\n        if init_param_style == \"openclip\":\n            # OpenCLIP style initialization\n            scale = self.embed_dim**-0.5\n\n            if self.num_cls_tokens > 0:\n                nn.init.normal_(self.cls_token)\n                self.cls_token *= scale\n        elif init_param_style == \"vit\":\n            self.cls_token.data.fill_(0)\n        else:\n            raise ValueError(f\"Unknown init {init_param_style}\")\n\n    def tokenize_input_and_cls_pos(self, input, stem):\n        # tokens is of shape B x L x D\n        tokens = stem.norm_layer(stem.proj(input))\n        assert tokens.ndim == 3\n        assert tokens.shape[2] == self.embed_dim\n        B = tokens.shape[0]\n        if self.num_cls_tokens > 0:\n            class_tokens = self.cls_token.expand(\n                B, -1, -1\n            )  # stole class_tokens impl from Phil Wang, thanks\n            tokens = torch.cat((class_tokens, tokens), dim=1)\n        if self.use_pos_embed:\n            tokens = tokens + self.pos_embed\n        return tokens\n\n    def forward(self, imu):\n        # Patchify\n        imu = imu.unfold(\n            -1,\n            self.kernel_size,\n            self.kernel_size,\n        ).permute(0, 2, 1, 3)\n        imu = imu.reshape(imu.size(0), imu.size(1), -1)\n\n        imu_tokens = self.tokenize_input_and_cls_pos(\n            imu,\n            self.imu_stem,\n        )\n\n        return_dict = {\n            \"trunk\": {\n                \"tokens\": imu_tokens,\n            },\n            \"head\": {},\n        }\n        return return_dict","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.get_position_angle_vec","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.get_position_angle_vec#L30-L34","kind":"function","name":"get_position_angle_vec","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":30,"end_line":34,"context_start_line":10,"context_end_line":54,"code":"import io\nimport math\nfrom functools import lru_cache\nfrom typing import Callable, List, Optional, Tuple\n\nimport ftfy\nimport numpy as np\nimport regex as re\nimport torch\nimport torch.nn as nn\nfrom iopath.common.file_io import g_pathmgr\nfrom timm.models.layers import trunc_normal_\n\nfrom .helpers import VerboseNNModule, cast_if_src_dtype\n\n\ndef get_sinusoid_encoding_table(n_position, d_hid):\n    \"\"\"Sinusoid position encoding table\"\"\"\n\n    # TODO: make it with torch instead of numpy\n    def get_position_angle_vec(position):\n        return [\n            position / np.power(10000, 2 * (hid_j // 2) / d_hid)\n            for hid_j in range(d_hid)\n        ]\n\n    sinusoid_table = np.array(\n        [get_position_angle_vec(pos_i) for pos_i in range(n_position)]\n    )\n    sinusoid_table[:, 0::2] = np.sin(sinusoid_table[:, 0::2])  # dim 2i\n    sinusoid_table[:, 1::2] = np.cos(sinusoid_table[:, 1::2])  # dim 2i+1\n\n    return torch.FloatTensor(sinusoid_table).unsqueeze(0)\n\n\ndef interpolate_pos_encoding_2d(target_spatial_size, pos_embed):\n    N = pos_embed.shape[1]\n    if N == target_spatial_size:\n        return pos_embed\n    dim = pos_embed.shape[-1]\n    # nn.functional.interpolate doesn't work with bfloat16 so we cast to float32\n    pos_embed, updated = cast_if_src_dtype(pos_embed, torch.bfloat16, torch.float32)\n    pos_embed = nn.functional.interpolate(\n        pos_embed.reshape(1, int(math.sqrt(N)), int(math.sqrt(N)), dim).permute(\n            0, 3, 1, 2","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.__init__#L607-L632","kind":"function","name":"__init__","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":607,"end_line":632,"context_start_line":587,"context_end_line":652,"code":"            context_length = self.context_length\n\n        if isinstance(texts, str):\n            texts = [texts]\n\n        sot_token = self.encoder[\"<|startoftext|>\"]\n        eot_token = self.encoder[\"<|endoftext|>\"]\n        all_tokens = [[sot_token] + self.encode(text) + [eot_token] for text in texts]\n        result = torch.zeros(len(all_tokens), context_length, dtype=torch.long)\n\n        for i, tokens in enumerate(all_tokens):\n            tokens = tokens[:context_length]\n            result[i, : len(tokens)] = torch.tensor(tokens)\n\n        if len(result) == 1:\n            return result[0]\n        return result\n\n\nclass IMUPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        kernel_size: int,\n        imu_stem: PatchEmbedGeneric,\n        embed_dim: int,\n        img_size: Tuple = (6, 2000),\n        num_cls_tokens: int = 1,\n        pos_embed_fn: Optional[Callable] = None,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        self.imu_stem = imu_stem\n        self.embed_dim = embed_dim\n        self.use_pos_embed = pos_embed_fn is not None\n        self.num_cls_tokens = num_cls_tokens\n        self.kernel_size = kernel_size\n        self.pos_embed = nn.Parameter(\n            torch.empty(1, (img_size[1] // kernel_size) + num_cls_tokens, embed_dim)\n        )\n\n        if self.num_cls_tokens > 0:\n            self.cls_token = nn.Parameter(\n                torch.zeros(1, self.num_cls_tokens, self.embed_dim)\n            )\n\n        self.init_parameters(init_param_style)\n\n    @torch.no_grad()\n    def init_parameters(self, init_param_style):\n        nn.init.normal_(self.pos_embed, std=0.01)\n\n        if init_param_style == \"openclip\":\n            # OpenCLIP style initialization\n            scale = self.embed_dim**-0.5\n\n            if self.num_cls_tokens > 0:\n                nn.init.normal_(self.cls_token)\n                self.cls_token *= scale\n        elif init_param_style == \"vit\":\n            self.cls_token.data.fill_(0)\n        else:\n            raise ValueError(f\"Unknown init {init_param_style}\")\n\n    def tokenize_input_and_cls_pos(self, input, stem):\n        # tokens is of shape B x L x D\n        tokens = stem.norm_layer(stem.proj(input))","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.get_patch_layout","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.get_patch_layout#L137-L149","kind":"function","name":"get_patch_layout","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":137,"end_line":149,"context_start_line":117,"context_end_line":169,"code":"    )\n    return pos_embed\n\n\nclass PatchEmbedGeneric(nn.Module):\n    \"\"\"\n    PatchEmbed from Hydra\n    \"\"\"\n\n    def __init__(self, proj_stem, norm_layer: Optional[nn.Module] = None):\n        super().__init__()\n\n        if len(proj_stem) > 1:\n            self.proj = nn.Sequential(*proj_stem)\n        else:\n            # Special case to be able to load pre-trained models that were\n            # trained with a standard stem\n            self.proj = proj_stem[0]\n        self.norm_layer = norm_layer\n\n    def get_patch_layout(self, img_size):\n        with torch.no_grad():\n            dummy_img = torch.zeros(\n                [\n                    1,\n                ]\n                + img_size\n            )\n            dummy_out = self.proj(dummy_img)\n        embed_dim = dummy_out.shape[1]\n        patches_layout = tuple(dummy_out.shape[2:])\n        num_patches = np.prod(patches_layout)\n        return patches_layout, num_patches, embed_dim\n\n    def forward(self, x):\n        x = self.proj(x)\n        # B C (T) H W -> B (T)HW C\n        x = x.flatten(2).transpose(1, 2)\n        if self.norm_layer is not None:\n            x = self.norm_layer(x)\n        return x\n\n\nclass SpatioTemporalPosEmbeddingHelper(VerboseNNModule):\n    def __init__(\n        self,\n        patches_layout: List,\n        num_patches: int,\n        num_cls_tokens: int,\n        embed_dim: int,\n        learnable: bool,\n    ) -> None:\n        super().__init__()","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.forward","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.forward#L665-L685","kind":"function","name":"forward","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":665,"end_line":685,"context_start_line":645,"context_end_line":685,"code":"        elif init_param_style == \"vit\":\n            self.cls_token.data.fill_(0)\n        else:\n            raise ValueError(f\"Unknown init {init_param_style}\")\n\n    def tokenize_input_and_cls_pos(self, input, stem):\n        # tokens is of shape B x L x D\n        tokens = stem.norm_layer(stem.proj(input))\n        assert tokens.ndim == 3\n        assert tokens.shape[2] == self.embed_dim\n        B = tokens.shape[0]\n        if self.num_cls_tokens > 0:\n            class_tokens = self.cls_token.expand(\n                B, -1, -1\n            )  # stole class_tokens impl from Phil Wang, thanks\n            tokens = torch.cat((class_tokens, tokens), dim=1)\n        if self.use_pos_embed:\n            tokens = tokens + self.pos_embed\n        return tokens\n\n    def forward(self, imu):\n        # Patchify\n        imu = imu.unfold(\n            -1,\n            self.kernel_size,\n            self.kernel_size,\n        ).permute(0, 2, 1, 3)\n        imu = imu.reshape(imu.size(0), imu.size(1), -1)\n\n        imu_tokens = self.tokenize_input_and_cls_pos(\n            imu,\n            self.imu_stem,\n        )\n\n        return_dict = {\n            \"trunk\": {\n                \"tokens\": imu_tokens,\n            },\n            \"head\": {},\n        }\n        return return_dict","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.get_pos_embedding","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.get_pos_embedding#L183-L192","kind":"function","name":"get_pos_embedding","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":183,"end_line":192,"context_start_line":163,"context_end_line":212,"code":"        patches_layout: List,\n        num_patches: int,\n        num_cls_tokens: int,\n        embed_dim: int,\n        learnable: bool,\n    ) -> None:\n        super().__init__()\n        self.num_cls_tokens = num_cls_tokens\n        self.patches_layout = patches_layout\n        self.num_patches = num_patches\n        self.num_tokens = num_cls_tokens + num_patches\n        self.learnable = learnable\n        if self.learnable:\n            self.pos_embed = nn.Parameter(torch.zeros(1, self.num_tokens, embed_dim))\n            trunc_normal_(self.pos_embed, std=0.02)\n        else:\n            self.register_buffer(\n                \"pos_embed\", get_sinusoid_encoding_table(self.num_tokens, embed_dim)\n            )\n\n    def get_pos_embedding(self, vision_input, all_vision_tokens):\n        input_shape = vision_input.shape\n        pos_embed = _get_pos_embedding(\n            all_vision_tokens.size(1) - self.num_cls_tokens,\n            pos_embed=self.pos_embed,\n            patches_layout=self.patches_layout,\n            input_shape=input_shape,\n            first_patch_idx=self.num_cls_tokens,\n        )\n        return pos_embed\n\n\nclass RGBDTPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        rgbt_stem: PatchEmbedGeneric,\n        depth_stem: Optional[PatchEmbedGeneric],\n        img_size: Tuple = (3, 224, 224),\n        num_cls_tokens: int = 1,\n        pos_embed_fn: Optional[Callable] = None,\n        use_type_embed: bool = False,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        stem = rgbt_stem if rgbt_stem is not None else depth_stem\n        (\n            self.patches_layout,\n            self.num_patches,\n            self.embed_dim,\n        ) = stem.get_patch_layout(img_size)","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.init_parameters","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.init_parameters#L635-L648","kind":"function","name":"init_parameters","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":635,"end_line":648,"context_start_line":615,"context_end_line":668,"code":"        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        self.imu_stem = imu_stem\n        self.embed_dim = embed_dim\n        self.use_pos_embed = pos_embed_fn is not None\n        self.num_cls_tokens = num_cls_tokens\n        self.kernel_size = kernel_size\n        self.pos_embed = nn.Parameter(\n            torch.empty(1, (img_size[1] // kernel_size) + num_cls_tokens, embed_dim)\n        )\n\n        if self.num_cls_tokens > 0:\n            self.cls_token = nn.Parameter(\n                torch.zeros(1, self.num_cls_tokens, self.embed_dim)\n            )\n\n        self.init_parameters(init_param_style)\n\n    @torch.no_grad()\n    def init_parameters(self, init_param_style):\n        nn.init.normal_(self.pos_embed, std=0.01)\n\n        if init_param_style == \"openclip\":\n            # OpenCLIP style initialization\n            scale = self.embed_dim**-0.5\n\n            if self.num_cls_tokens > 0:\n                nn.init.normal_(self.cls_token)\n                self.cls_token *= scale\n        elif init_param_style == \"vit\":\n            self.cls_token.data.fill_(0)\n        else:\n            raise ValueError(f\"Unknown init {init_param_style}\")\n\n    def tokenize_input_and_cls_pos(self, input, stem):\n        # tokens is of shape B x L x D\n        tokens = stem.norm_layer(stem.proj(input))\n        assert tokens.ndim == 3\n        assert tokens.shape[2] == self.embed_dim\n        B = tokens.shape[0]\n        if self.num_cls_tokens > 0:\n            class_tokens = self.cls_token.expand(\n                B, -1, -1\n            )  # stole class_tokens impl from Phil Wang, thanks\n            tokens = torch.cat((class_tokens, tokens), dim=1)\n        if self.use_pos_embed:\n            tokens = tokens + self.pos_embed\n        return tokens\n\n    def forward(self, imu):\n        # Patchify\n        imu = imu.unfold(\n            -1,","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.tokenize_input_and_cls_pos","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.tokenize_input_and_cls_pos#L650-L663","kind":"function","name":"tokenize_input_and_cls_pos","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":650,"end_line":663,"context_start_line":630,"context_end_line":683,"code":"            )\n\n        self.init_parameters(init_param_style)\n\n    @torch.no_grad()\n    def init_parameters(self, init_param_style):\n        nn.init.normal_(self.pos_embed, std=0.01)\n\n        if init_param_style == \"openclip\":\n            # OpenCLIP style initialization\n            scale = self.embed_dim**-0.5\n\n            if self.num_cls_tokens > 0:\n                nn.init.normal_(self.cls_token)\n                self.cls_token *= scale\n        elif init_param_style == \"vit\":\n            self.cls_token.data.fill_(0)\n        else:\n            raise ValueError(f\"Unknown init {init_param_style}\")\n\n    def tokenize_input_and_cls_pos(self, input, stem):\n        # tokens is of shape B x L x D\n        tokens = stem.norm_layer(stem.proj(input))\n        assert tokens.ndim == 3\n        assert tokens.shape[2] == self.embed_dim\n        B = tokens.shape[0]\n        if self.num_cls_tokens > 0:\n            class_tokens = self.cls_token.expand(\n                B, -1, -1\n            )  # stole class_tokens impl from Phil Wang, thanks\n            tokens = torch.cat((class_tokens, tokens), dim=1)\n        if self.use_pos_embed:\n            tokens = tokens + self.pos_embed\n        return tokens\n\n    def forward(self, imu):\n        # Patchify\n        imu = imu.unfold(\n            -1,\n            self.kernel_size,\n            self.kernel_size,\n        ).permute(0, 2, 1, 3)\n        imu = imu.reshape(imu.size(0), imu.size(1), -1)\n\n        imu_tokens = self.tokenize_input_and_cls_pos(\n            imu,\n            self.imu_stem,\n        )\n\n        return_dict = {\n            \"trunk\": {\n                \"tokens\": imu_tokens,\n            },\n            \"head\": {},","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.bpe","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.bpe#L525-L564","kind":"function","name":"bpe","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":525,"end_line":564,"context_start_line":505,"context_end_line":584,"code":"        merges = merges[1 : 49152 - 256 - 2 + 1]\n        merges: List[Tuple[str, ...]] = [tuple(merge.split()) for merge in merges]\n        vocab = list(bytes_to_unicode().values())\n        vocab = vocab + [v + \"</w>\" for v in vocab]\n        for merge in merges:\n            vocab.append(\"\".join(merge))\n        vocab.extend([\"<|startoftext|>\", \"<|endoftext|>\"])\n        self.encoder = dict(zip(vocab, range(len(vocab))))\n        self.decoder = {v: k for k, v in self.encoder.items()}\n        self.bpe_ranks = dict(zip(merges, range(len(merges))))\n        self.cache = {\n            \"<|startoftext|>\": \"<|startoftext|>\",\n            \"<|endoftext|>\": \"<|endoftext|>\",\n        }\n        self.pat = re.compile(\n            r\"\"\"<\\|startoftext\\|>|<\\|endoftext\\|>|'s|'t|'re|'ve|'m|'ll|'d|[\\p{L}]+|[\\p{N}]|[^\\s\\p{L}\\p{N}]+\"\"\",\n            re.IGNORECASE,\n        )\n        self.context_length = context_length\n\n    def bpe(self, token):\n        if token in self.cache:\n            return self.cache[token]\n        word = tuple(token[:-1]) + (token[-1] + \"</w>\",)\n        pairs = get_pairs(word)\n\n        if not pairs:\n            return token + \"</w>\"\n\n        while True:\n            bigram = min(pairs, key=lambda pair: self.bpe_ranks.get(pair, float(\"inf\")))\n            if bigram not in self.bpe_ranks:\n                break\n            first, second = bigram\n            new_word = []\n            i = 0\n            while i < len(word):\n                try:\n                    j = word.index(first, i)\n                    new_word.extend(word[i:j])\n                    i = j\n                except:\n                    new_word.extend(word[i:])\n                    break\n\n                if word[i] == first and i < len(word) - 1 and word[i + 1] == second:\n                    new_word.append(first + second)\n                    i += 2\n                else:\n                    new_word.append(word[i])\n                    i += 1\n            new_word = tuple(new_word)\n            word = new_word\n            if len(word) == 1:\n                break\n            else:\n                pairs = get_pairs(word)\n        word = \" \".join(word)\n        self.cache[token] = word\n        return word\n\n    def encode(self, text):\n        bpe_tokens = []\n        text = whitespace_clean(basic_clean(text)).lower()\n        for token in re.findall(self.pat, text):\n            token = \"\".join(self.byte_encoder[b] for b in token.encode(\"utf-8\"))\n            bpe_tokens.extend(\n                self.encoder[bpe_token] for bpe_token in self.bpe(token).split(\" \")\n            )\n        return bpe_tokens\n\n    def decode(self, tokens):\n        text = \"\".join([self.decoder[token] for token in tokens])\n        text = (\n            bytearray([self.byte_decoder[c] for c in text])\n            .decode(\"utf-8\", errors=\"replace\")\n            .replace(\"</w>\", \" \")\n        )\n        return text\n","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.encode","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.encode#L566-L574","kind":"function","name":"encode","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":566,"end_line":574,"context_start_line":546,"context_end_line":594,"code":"                except:\n                    new_word.extend(word[i:])\n                    break\n\n                if word[i] == first and i < len(word) - 1 and word[i + 1] == second:\n                    new_word.append(first + second)\n                    i += 2\n                else:\n                    new_word.append(word[i])\n                    i += 1\n            new_word = tuple(new_word)\n            word = new_word\n            if len(word) == 1:\n                break\n            else:\n                pairs = get_pairs(word)\n        word = \" \".join(word)\n        self.cache[token] = word\n        return word\n\n    def encode(self, text):\n        bpe_tokens = []\n        text = whitespace_clean(basic_clean(text)).lower()\n        for token in re.findall(self.pat, text):\n            token = \"\".join(self.byte_encoder[b] for b in token.encode(\"utf-8\"))\n            bpe_tokens.extend(\n                self.encoder[bpe_token] for bpe_token in self.bpe(token).split(\" \")\n            )\n        return bpe_tokens\n\n    def decode(self, tokens):\n        text = \"\".join([self.decoder[token] for token in tokens])\n        text = (\n            bytearray([self.byte_decoder[c] for c in text])\n            .decode(\"utf-8\", errors=\"replace\")\n            .replace(\"</w>\", \" \")\n        )\n        return text\n\n    def __call__(self, texts, context_length=None):\n        if not context_length:\n            context_length = self.context_length\n\n        if isinstance(texts, str):\n            texts = [texts]\n\n        sot_token = self.encoder[\"<|startoftext|>\"]\n        eot_token = self.encoder[\"<|endoftext|>\"]\n        all_tokens = [[sot_token] + self.encode(text) + [eot_token] for text in texts]","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.decode","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.decode#L576-L583","kind":"function","name":"decode","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":576,"end_line":583,"context_start_line":556,"context_end_line":603,"code":"            new_word = tuple(new_word)\n            word = new_word\n            if len(word) == 1:\n                break\n            else:\n                pairs = get_pairs(word)\n        word = \" \".join(word)\n        self.cache[token] = word\n        return word\n\n    def encode(self, text):\n        bpe_tokens = []\n        text = whitespace_clean(basic_clean(text)).lower()\n        for token in re.findall(self.pat, text):\n            token = \"\".join(self.byte_encoder[b] for b in token.encode(\"utf-8\"))\n            bpe_tokens.extend(\n                self.encoder[bpe_token] for bpe_token in self.bpe(token).split(\" \")\n            )\n        return bpe_tokens\n\n    def decode(self, tokens):\n        text = \"\".join([self.decoder[token] for token in tokens])\n        text = (\n            bytearray([self.byte_decoder[c] for c in text])\n            .decode(\"utf-8\", errors=\"replace\")\n            .replace(\"</w>\", \" \")\n        )\n        return text\n\n    def __call__(self, texts, context_length=None):\n        if not context_length:\n            context_length = self.context_length\n\n        if isinstance(texts, str):\n            texts = [texts]\n\n        sot_token = self.encoder[\"<|startoftext|>\"]\n        eot_token = self.encoder[\"<|endoftext|>\"]\n        all_tokens = [[sot_token] + self.encode(text) + [eot_token] for text in texts]\n        result = torch.zeros(len(all_tokens), context_length, dtype=torch.long)\n\n        for i, tokens in enumerate(all_tokens):\n            tokens = tokens[:context_length]\n            result[i, : len(tokens)] = torch.tensor(tokens)\n\n        if len(result) == 1:\n            return result[0]\n        return result","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.multimodal_preprocessors.__call__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.multimodal_preprocessors.__call__#L585-L603","kind":"function","name":"__call__","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":585,"end_line":603,"context_start_line":565,"context_end_line":623,"code":"\n    def encode(self, text):\n        bpe_tokens = []\n        text = whitespace_clean(basic_clean(text)).lower()\n        for token in re.findall(self.pat, text):\n            token = \"\".join(self.byte_encoder[b] for b in token.encode(\"utf-8\"))\n            bpe_tokens.extend(\n                self.encoder[bpe_token] for bpe_token in self.bpe(token).split(\" \")\n            )\n        return bpe_tokens\n\n    def decode(self, tokens):\n        text = \"\".join([self.decoder[token] for token in tokens])\n        text = (\n            bytearray([self.byte_decoder[c] for c in text])\n            .decode(\"utf-8\", errors=\"replace\")\n            .replace(\"</w>\", \" \")\n        )\n        return text\n\n    def __call__(self, texts, context_length=None):\n        if not context_length:\n            context_length = self.context_length\n\n        if isinstance(texts, str):\n            texts = [texts]\n\n        sot_token = self.encoder[\"<|startoftext|>\"]\n        eot_token = self.encoder[\"<|endoftext|>\"]\n        all_tokens = [[sot_token] + self.encode(text) + [eot_token] for text in texts]\n        result = torch.zeros(len(all_tokens), context_length, dtype=torch.long)\n\n        for i, tokens in enumerate(all_tokens):\n            tokens = tokens[:context_length]\n            result[i, : len(tokens)] = torch.tensor(tokens)\n\n        if len(result) == 1:\n            return result[0]\n        return result\n\n\nclass IMUPreprocessor(VerboseNNModule):\n    def __init__(\n        self,\n        kernel_size: int,\n        imu_stem: PatchEmbedGeneric,\n        embed_dim: int,\n        img_size: Tuple = (6, 2000),\n        num_cls_tokens: int = 1,\n        pos_embed_fn: Optional[Callable] = None,\n        init_param_style: str = \"openclip\",\n    ) -> None:\n        super().__init__()\n        self.imu_stem = imu_stem\n        self.embed_dim = embed_dim\n        self.use_pos_embed = pos_embed_fn is not None\n        self.num_cls_tokens = num_cls_tokens\n        self.kernel_size = kernel_size\n        self.pos_embed = nn.Parameter(","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.imagebind_model","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.models.imagebind_model#L1-L534","kind":"module","name":"imagebind_LLM.ImageBind.models.imagebind_model","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":1,"end_line":534,"context_start_line":1,"context_end_line":534,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\n\nimport os\nfrom functools import partial\nfrom types import SimpleNamespace\nfrom collections import OrderedDict\nfrom .pointbert.point_encoder import PointTransformerBind\nfrom .pointbert.checkpoint import get_missing_parameters_message, get_unexpected_parameters_message\n\nimport torch\nimport torch.nn as nn\n\nfrom .helpers import (EinOpsRearrange, LearnableLogitScaling, Normalize,\n                            SelectElement, SelectEOSAndProject)\nfrom .multimodal_preprocessors import (AudioPreprocessor,\n                                             IMUPreprocessor, PadIm2Video,\n                                             PatchEmbedGeneric,\n                                             RGBDTPreprocessor,\n                                             SpatioTemporalPosEmbeddingHelper,\n                                             TextPreprocessor,\n                                             ThermalPreprocessor)\nfrom .transformer import MultiheadAttention, SimpleTransformer\n\nModalityType = SimpleNamespace(\n    VISION=\"vision\",\n    TEXT=\"text\",\n    AUDIO=\"audio\",\n    THERMAL=\"thermal\",\n    DEPTH=\"depth\",\n    IMU=\"imu\",\n    POINT=\"point\",\n)\n\nclass ImageBindModel(nn.Module):\n    def __init__(\n        self,\n        video_frames=2,\n        kernel_size=(2, 14, 14),\n        audio_kernel_size=16,\n        audio_stride=10,\n        out_embed_dim=768,\n        vision_embed_dim=1024,\n        vision_num_blocks=24,\n        vision_num_heads=16,\n        audio_embed_dim=768,\n        audio_num_blocks=12,\n        audio_num_heads=12,\n        audio_num_mel_bins=128,\n        audio_target_len=204,\n        audio_drop_path=0.1,\n        text_embed_dim=768,\n        text_num_blocks=12,\n        text_num_heads=12,\n        depth_embed_dim=384,\n        depth_kernel_size=16,\n        depth_num_blocks=12,\n        depth_num_heads=8,\n        depth_drop_path=0.0,\n        thermal_embed_dim=768,\n        thermal_kernel_size=16,\n        thermal_num_blocks=12,\n        thermal_num_heads=12,\n        thermal_drop_path=0.0,\n        imu_embed_dim=512,\n        imu_kernel_size=8,\n        imu_num_blocks=6,\n        imu_num_heads=8,\n        imu_drop_path=0.7,\n    ):\n        super().__init__()\n\n        self.modality_preprocessors = self._create_modality_preprocessors(\n            video_frames,\n            vision_embed_dim,\n            kernel_size,\n            text_embed_dim,\n            audio_embed_dim,\n            audio_kernel_size,\n            audio_stride,\n            audio_num_mel_bins,\n            audio_target_len,\n            depth_embed_dim,\n            depth_kernel_size,\n            thermal_embed_dim,\n            thermal_kernel_size,\n            imu_embed_dim,\n        )\n\n        self.modality_trunks = self._create_modality_trunks(\n            vision_embed_dim,\n            vision_num_blocks,\n            vision_num_heads,\n            text_embed_dim,\n            text_num_blocks,\n            text_num_heads,\n            audio_embed_dim,\n            audio_num_blocks,\n            audio_num_heads,\n            audio_drop_path,\n            depth_embed_dim,\n            depth_num_blocks,\n            depth_num_heads,\n            depth_drop_path,\n            thermal_embed_dim,\n            thermal_num_blocks,\n            thermal_num_heads,\n            thermal_drop_path,\n            imu_embed_dim,\n            imu_num_blocks,\n            imu_num_heads,\n            imu_drop_path,\n        )\n\n        self.modality_heads = self._create_modality_heads(\n            out_embed_dim,\n            vision_embed_dim,\n            text_embed_dim,\n            audio_embed_dim,\n            depth_embed_dim,\n            thermal_embed_dim,\n            imu_embed_dim,\n        )\n\n        self.modality_postprocessors = self._create_modality_postprocessors(\n            out_embed_dim\n        )\n\n        self.point_trunk = PointTransformerBind()\n    \n    def _create_modality_preprocessors(\n        self,\n        video_frames=2,\n        vision_embed_dim=1024,\n        kernel_size=(2, 14, 14),\n        text_embed_dim=768,\n        audio_embed_dim=768,\n        audio_kernel_size=16,\n        audio_stride=10,\n        audio_num_mel_bins=128,\n        audio_target_len=204,\n        depth_embed_dim=768,\n        depth_kernel_size=16,\n        thermal_embed_dim=768,\n        thermal_kernel_size=16,\n        imu_embed_dim=512,\n    ):\n        rgbt_stem = PatchEmbedGeneric(\n            proj_stem=[\n                PadIm2Video(pad_type=\"repeat\", ntimes=2),\n                nn.Conv3d(\n                    in_channels=3,\n                    kernel_size=kernel_size,\n                    out_channels=vision_embed_dim,\n                    stride=kernel_size,\n                    bias=False,\n                ),\n            ]\n        )\n        rgbt_preprocessor = RGBDTPreprocessor(\n            img_size=[3, video_frames, 224, 224],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            rgbt_stem=rgbt_stem,\n            depth_stem=None,\n        )\n\n        text_preprocessor = TextPreprocessor(\n            context_length=77,\n            vocab_size=49408,\n            embed_dim=text_embed_dim,\n            causal_masking=True,\n        )\n\n        audio_stem = PatchEmbedGeneric(\n            proj_stem=[\n                nn.Conv2d(\n                    in_channels=1,\n                    kernel_size=audio_kernel_size,\n                    stride=audio_stride,\n                    out_channels=audio_embed_dim,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=audio_embed_dim),\n        )\n        audio_preprocessor = AudioPreprocessor(\n            img_size=[1, audio_num_mel_bins, audio_target_len],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            audio_stem=audio_stem,\n        )\n\n        depth_stem = PatchEmbedGeneric(\n            [\n                nn.Conv2d(\n                    kernel_size=depth_kernel_size,\n                    in_channels=1,\n                    out_channels=depth_embed_dim,\n                    stride=depth_kernel_size,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=depth_embed_dim),\n        )\n\n        depth_preprocessor = RGBDTPreprocessor(\n            img_size=[1, 224, 224],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            rgbt_stem=None,\n            depth_stem=depth_stem,\n        )\n\n        thermal_stem = PatchEmbedGeneric(\n            [\n                nn.Conv2d(\n                    kernel_size=thermal_kernel_size,\n                    in_channels=1,\n                    out_channels=thermal_embed_dim,\n                    stride=thermal_kernel_size,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=thermal_embed_dim),\n        )\n        thermal_preprocessor = ThermalPreprocessor(\n            img_size=[1, 224, 224],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            thermal_stem=thermal_stem,\n        )\n\n        imu_stem = PatchEmbedGeneric(\n            [\n                nn.Linear(\n                    in_features=48,\n                    out_features=imu_embed_dim,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=imu_embed_dim),\n        )\n\n        imu_preprocessor = IMUPreprocessor(\n            img_size=[6, 2000],\n            num_cls_tokens=1,\n            kernel_size=8,\n            embed_dim=imu_embed_dim,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            imu_stem=imu_stem,\n        )\n\n        modality_preprocessors = {\n            ModalityType.VISION: rgbt_preprocessor,\n            ModalityType.TEXT: text_preprocessor,\n            ModalityType.AUDIO: audio_preprocessor,\n            ModalityType.DEPTH: depth_preprocessor,\n            ModalityType.THERMAL: thermal_preprocessor,\n            ModalityType.IMU: imu_preprocessor,\n        }\n\n        return nn.ModuleDict(modality_preprocessors)\n\n    def _create_modality_trunks(\n        self,\n        vision_embed_dim=1024,\n        vision_num_blocks=24,\n        vision_num_heads=16,\n        text_embed_dim=768,\n        text_num_blocks=12,\n        text_num_heads=12,\n        audio_embed_dim=768,\n        audio_num_blocks=12,\n        audio_num_heads=12,\n        audio_drop_path=0.0,\n        depth_embed_dim=768,\n        depth_num_blocks=12,\n        depth_num_heads=12,\n        depth_drop_path=0.0,\n        thermal_embed_dim=768,\n        thermal_num_blocks=12,\n        thermal_num_heads=12,\n        thermal_drop_path=0.0,\n        imu_embed_dim=512,\n        imu_num_blocks=6,\n        imu_num_heads=8,\n        imu_drop_path=0.7,\n    ):\n        def instantiate_trunk(\n            embed_dim, num_blocks, num_heads, pre_transformer_ln, add_bias_kv, drop_path\n        ):\n            return SimpleTransformer(\n                embed_dim=embed_dim,\n                num_blocks=num_blocks,\n                ffn_dropout_rate=0.0,\n                drop_path_rate=drop_path,\n                attn_target=partial(\n                    MultiheadAttention,\n                    embed_dim=embed_dim,\n                    num_heads=num_heads,\n                    bias=True,\n                    add_bias_kv=add_bias_kv,\n                ),\n                pre_transformer_layer=nn.Sequential(\n                    nn.LayerNorm(embed_dim, eps=1e-6)\n                    if pre_transformer_ln\n                    else nn.Identity(),\n                    EinOpsRearrange(\"b l d -> l b d\"),\n                ),\n                post_transformer_layer=EinOpsRearrange(\"l b d -> b l d\"),\n            )\n\n        modality_trunks = {}\n        modality_trunks[ModalityType.VISION] = instantiate_trunk(\n            vision_embed_dim,\n            vision_num_blocks,\n            vision_num_heads,\n            pre_transformer_ln=True,\n            add_bias_kv=False,\n            drop_path=0.0,\n        )\n        modality_trunks[ModalityType.TEXT] = instantiate_trunk(\n            text_embed_dim,\n            text_num_blocks,\n            text_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=False,\n            drop_path=0.0,\n        )\n        modality_trunks[ModalityType.AUDIO] = instantiate_trunk(\n            audio_embed_dim,\n            audio_num_blocks,\n            audio_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=audio_drop_path,\n        )\n        modality_trunks[ModalityType.DEPTH] = instantiate_trunk(\n            depth_embed_dim,\n            depth_num_blocks,\n            depth_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=depth_drop_path,\n        )\n        modality_trunks[ModalityType.THERMAL] = instantiate_trunk(\n            thermal_embed_dim,\n            thermal_num_blocks,\n            thermal_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=thermal_drop_path,\n        )\n        modality_trunks[ModalityType.IMU] = instantiate_trunk(\n            imu_embed_dim,\n            imu_num_blocks,\n            imu_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=imu_drop_path,\n        )\n\n        return nn.ModuleDict(modality_trunks)\n\n    def _create_modality_heads(\n        self,\n        out_embed_dim,\n        vision_embed_dim,\n        text_embed_dim,\n        audio_embed_dim,\n        depth_embed_dim,\n        thermal_embed_dim,\n        imu_embed_dim,\n    ):\n        modality_heads = {}\n\n        modality_heads[ModalityType.VISION] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=vision_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(vision_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.TEXT] = SelectEOSAndProject(\n            proj=nn.Sequential(\n                nn.LayerNorm(normalized_shape=text_embed_dim, eps=1e-6),\n                nn.Linear(text_embed_dim, out_embed_dim, bias=False),\n            )\n        )\n\n        modality_heads[ModalityType.AUDIO] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=audio_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(audio_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.DEPTH] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=depth_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(depth_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.THERMAL] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=thermal_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(thermal_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.IMU] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=imu_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Dropout(p=0.5),\n            nn.Linear(imu_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.POINT] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=512, eps=1e-6),\n            nn.Dropout(p=0.5),\n            nn.Linear(512, out_embed_dim, bias=False),\n        )\n\n        return nn.ModuleDict(modality_heads)\n\n    def _create_modality_postprocessors(self, out_embed_dim):\n        modality_postprocessors = {}\n\n        modality_postprocessors[ModalityType.VISION] = Normalize(dim=-1)\n        modality_postprocessors[ModalityType.TEXT] = nn.Sequential(\n            Normalize(dim=-1), LearnableLogitScaling(learnable=True)\n        )\n        modality_postprocessors[ModalityType.AUDIO] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=20.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.DEPTH] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=5.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.THERMAL] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=10.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.IMU] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=5.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.POINT] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=1.0, learnable=False),\n        )\n        return nn.ModuleDict(modality_postprocessors)\n\n    def forward(self, inputs, prenorm=False):\n        outputs = {}\n        outputs_prenorm = {}\n        for modality_key, modality_value in inputs.items():\n            reduce_list = (\n                modality_value.ndim >= 5\n            )  # Audio and Video inputs consist of multiple clips\n            if reduce_list:\n                B, S = modality_value.shape[:2]\n                modality_value = modality_value.reshape(\n                    B * S, *modality_value.shape[2:]\n                )\n\n            if modality_value is not None:\n                if modality_key == ModalityType.POINT:\n                    modality_value = self.point_trunk(modality_value)\n                    modality_value = self.modality_heads[modality_key](modality_value)\n                else:\n                    modality_value = self.modality_preprocessors[modality_key](\n                        **{modality_key: modality_value}\n                    )\n                    trunk_inputs = modality_value[\"trunk\"]\n                    head_inputs = modality_value[\"head\"]\n                    modality_value = self.modality_trunks[modality_key](**trunk_inputs)\n                    modality_value = self.modality_heads[modality_key](\n                        modality_value, **head_inputs\n                    )\n\n                modality_value_postnorm = self.modality_postprocessors[modality_key](\n                    modality_value\n                )\n\n\n                if reduce_list:\n                    if prenorm:\n                        modality_value = modality_value.reshape(B, S, -1)\n                        modality_value = modality_value.mean(dim=1)\n\n                    modality_value_postnorm = modality_value_postnorm.reshape(B, S, -1)\n                    modality_value_postnorm = modality_value_postnorm.mean(dim=1)\n\n                if prenorm:\n                    outputs_prenorm[modality_key] = modality_value\n                outputs[modality_key] = modality_value_postnorm\n\n        if prenorm:\n            return outputs, outputs_prenorm\n        else:\n            return outputs\n\n\ndef imagebind_huge(pretrained=False):\n    model = ImageBindModel(\n        vision_embed_dim=1280,\n        vision_num_blocks=32,\n        vision_num_heads=16,\n        text_embed_dim=1024,\n        text_num_blocks=24,\n        text_num_heads=16,\n        out_embed_dim=1024,\n        audio_drop_path=0.1,\n        imu_drop_path=0.7,\n    )\n\n    if pretrained:\n        if not os.path.exists(\"./ckpts/imagebind_w3D.pth\"):\n            print(\n                \"Downloading the pre-train weight of ImageBind with 3D encoder to ./ckpts/imagebind_w3D.pth ...\"\n            )\n            os.makedirs(\"./ckpts\", exist_ok=True)\n            torch.hub.download_url_to_file(\n                \"https://huggingface.co/ZiyuG/ImageBind_w3D/resolve/e9206a10f118b0790c730264e7b3aa5c324c35cb/imagebind_w3D.pth\",\n                \"./ckpts/imagebind_w3D.pth\",\n                progress=True,\n            )\n        model.load_state_dict(torch.load(\"./ckpts/imagebind_w3D.pth\"))\n    return model","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.imagebind_model.ImageBindModel","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.imagebind_model.ImageBindModel#L40-L506","kind":"class","name":"ImageBindModel","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":40,"end_line":506,"context_start_line":20,"context_end_line":526,"code":"                            SelectElement, SelectEOSAndProject)\nfrom .multimodal_preprocessors import (AudioPreprocessor,\n                                             IMUPreprocessor, PadIm2Video,\n                                             PatchEmbedGeneric,\n                                             RGBDTPreprocessor,\n                                             SpatioTemporalPosEmbeddingHelper,\n                                             TextPreprocessor,\n                                             ThermalPreprocessor)\nfrom .transformer import MultiheadAttention, SimpleTransformer\n\nModalityType = SimpleNamespace(\n    VISION=\"vision\",\n    TEXT=\"text\",\n    AUDIO=\"audio\",\n    THERMAL=\"thermal\",\n    DEPTH=\"depth\",\n    IMU=\"imu\",\n    POINT=\"point\",\n)\n\nclass ImageBindModel(nn.Module):\n    def __init__(\n        self,\n        video_frames=2,\n        kernel_size=(2, 14, 14),\n        audio_kernel_size=16,\n        audio_stride=10,\n        out_embed_dim=768,\n        vision_embed_dim=1024,\n        vision_num_blocks=24,\n        vision_num_heads=16,\n        audio_embed_dim=768,\n        audio_num_blocks=12,\n        audio_num_heads=12,\n        audio_num_mel_bins=128,\n        audio_target_len=204,\n        audio_drop_path=0.1,\n        text_embed_dim=768,\n        text_num_blocks=12,\n        text_num_heads=12,\n        depth_embed_dim=384,\n        depth_kernel_size=16,\n        depth_num_blocks=12,\n        depth_num_heads=8,\n        depth_drop_path=0.0,\n        thermal_embed_dim=768,\n        thermal_kernel_size=16,\n        thermal_num_blocks=12,\n        thermal_num_heads=12,\n        thermal_drop_path=0.0,\n        imu_embed_dim=512,\n        imu_kernel_size=8,\n        imu_num_blocks=6,\n        imu_num_heads=8,\n        imu_drop_path=0.7,\n    ):\n        super().__init__()\n\n        self.modality_preprocessors = self._create_modality_preprocessors(\n            video_frames,\n            vision_embed_dim,\n            kernel_size,\n            text_embed_dim,\n            audio_embed_dim,\n            audio_kernel_size,\n            audio_stride,\n            audio_num_mel_bins,\n            audio_target_len,\n            depth_embed_dim,\n            depth_kernel_size,\n            thermal_embed_dim,\n            thermal_kernel_size,\n            imu_embed_dim,\n        )\n\n        self.modality_trunks = self._create_modality_trunks(\n            vision_embed_dim,\n            vision_num_blocks,\n            vision_num_heads,\n            text_embed_dim,\n            text_num_blocks,\n            text_num_heads,\n            audio_embed_dim,\n            audio_num_blocks,\n            audio_num_heads,\n            audio_drop_path,\n            depth_embed_dim,\n            depth_num_blocks,\n            depth_num_heads,\n            depth_drop_path,\n            thermal_embed_dim,\n            thermal_num_blocks,\n            thermal_num_heads,\n            thermal_drop_path,\n            imu_embed_dim,\n            imu_num_blocks,\n            imu_num_heads,\n            imu_drop_path,\n        )\n\n        self.modality_heads = self._create_modality_heads(\n            out_embed_dim,\n            vision_embed_dim,\n            text_embed_dim,\n            audio_embed_dim,\n            depth_embed_dim,\n            thermal_embed_dim,\n            imu_embed_dim,\n        )\n\n        self.modality_postprocessors = self._create_modality_postprocessors(\n            out_embed_dim\n        )\n\n        self.point_trunk = PointTransformerBind()\n    \n    def _create_modality_preprocessors(\n        self,\n        video_frames=2,\n        vision_embed_dim=1024,\n        kernel_size=(2, 14, 14),\n        text_embed_dim=768,\n        audio_embed_dim=768,\n        audio_kernel_size=16,\n        audio_stride=10,\n        audio_num_mel_bins=128,\n        audio_target_len=204,\n        depth_embed_dim=768,\n        depth_kernel_size=16,\n        thermal_embed_dim=768,\n        thermal_kernel_size=16,\n        imu_embed_dim=512,\n    ):\n        rgbt_stem = PatchEmbedGeneric(\n            proj_stem=[\n                PadIm2Video(pad_type=\"repeat\", ntimes=2),\n                nn.Conv3d(\n                    in_channels=3,\n                    kernel_size=kernel_size,\n                    out_channels=vision_embed_dim,\n                    stride=kernel_size,\n                    bias=False,\n                ),\n            ]\n        )\n        rgbt_preprocessor = RGBDTPreprocessor(\n            img_size=[3, video_frames, 224, 224],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            rgbt_stem=rgbt_stem,\n            depth_stem=None,\n        )\n\n        text_preprocessor = TextPreprocessor(\n            context_length=77,\n            vocab_size=49408,\n            embed_dim=text_embed_dim,\n            causal_masking=True,\n        )\n\n        audio_stem = PatchEmbedGeneric(\n            proj_stem=[\n                nn.Conv2d(\n                    in_channels=1,\n                    kernel_size=audio_kernel_size,\n                    stride=audio_stride,\n                    out_channels=audio_embed_dim,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=audio_embed_dim),\n        )\n        audio_preprocessor = AudioPreprocessor(\n            img_size=[1, audio_num_mel_bins, audio_target_len],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            audio_stem=audio_stem,\n        )\n\n        depth_stem = PatchEmbedGeneric(\n            [\n                nn.Conv2d(\n                    kernel_size=depth_kernel_size,\n                    in_channels=1,\n                    out_channels=depth_embed_dim,\n                    stride=depth_kernel_size,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=depth_embed_dim),\n        )\n\n        depth_preprocessor = RGBDTPreprocessor(\n            img_size=[1, 224, 224],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            rgbt_stem=None,\n            depth_stem=depth_stem,\n        )\n\n        thermal_stem = PatchEmbedGeneric(\n            [\n                nn.Conv2d(\n                    kernel_size=thermal_kernel_size,\n                    in_channels=1,\n                    out_channels=thermal_embed_dim,\n                    stride=thermal_kernel_size,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=thermal_embed_dim),\n        )\n        thermal_preprocessor = ThermalPreprocessor(\n            img_size=[1, 224, 224],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            thermal_stem=thermal_stem,\n        )\n\n        imu_stem = PatchEmbedGeneric(\n            [\n                nn.Linear(\n                    in_features=48,\n                    out_features=imu_embed_dim,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=imu_embed_dim),\n        )\n\n        imu_preprocessor = IMUPreprocessor(\n            img_size=[6, 2000],\n            num_cls_tokens=1,\n            kernel_size=8,\n            embed_dim=imu_embed_dim,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            imu_stem=imu_stem,\n        )\n\n        modality_preprocessors = {\n            ModalityType.VISION: rgbt_preprocessor,\n            ModalityType.TEXT: text_preprocessor,\n            ModalityType.AUDIO: audio_preprocessor,\n            ModalityType.DEPTH: depth_preprocessor,\n            ModalityType.THERMAL: thermal_preprocessor,\n            ModalityType.IMU: imu_preprocessor,\n        }\n\n        return nn.ModuleDict(modality_preprocessors)\n\n    def _create_modality_trunks(\n        self,\n        vision_embed_dim=1024,\n        vision_num_blocks=24,\n        vision_num_heads=16,\n        text_embed_dim=768,\n        text_num_blocks=12,\n        text_num_heads=12,\n        audio_embed_dim=768,\n        audio_num_blocks=12,\n        audio_num_heads=12,\n        audio_drop_path=0.0,\n        depth_embed_dim=768,\n        depth_num_blocks=12,\n        depth_num_heads=12,\n        depth_drop_path=0.0,\n        thermal_embed_dim=768,\n        thermal_num_blocks=12,\n        thermal_num_heads=12,\n        thermal_drop_path=0.0,\n        imu_embed_dim=512,\n        imu_num_blocks=6,\n        imu_num_heads=8,\n        imu_drop_path=0.7,\n    ):\n        def instantiate_trunk(\n            embed_dim, num_blocks, num_heads, pre_transformer_ln, add_bias_kv, drop_path\n        ):\n            return SimpleTransformer(\n                embed_dim=embed_dim,\n                num_blocks=num_blocks,\n                ffn_dropout_rate=0.0,\n                drop_path_rate=drop_path,\n                attn_target=partial(\n                    MultiheadAttention,\n                    embed_dim=embed_dim,\n                    num_heads=num_heads,\n                    bias=True,\n                    add_bias_kv=add_bias_kv,\n                ),\n                pre_transformer_layer=nn.Sequential(\n                    nn.LayerNorm(embed_dim, eps=1e-6)\n                    if pre_transformer_ln\n                    else nn.Identity(),\n                    EinOpsRearrange(\"b l d -> l b d\"),\n                ),\n                post_transformer_layer=EinOpsRearrange(\"l b d -> b l d\"),\n            )\n\n        modality_trunks = {}\n        modality_trunks[ModalityType.VISION] = instantiate_trunk(\n            vision_embed_dim,\n            vision_num_blocks,\n            vision_num_heads,\n            pre_transformer_ln=True,\n            add_bias_kv=False,\n            drop_path=0.0,\n        )\n        modality_trunks[ModalityType.TEXT] = instantiate_trunk(\n            text_embed_dim,\n            text_num_blocks,\n            text_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=False,\n            drop_path=0.0,\n        )\n        modality_trunks[ModalityType.AUDIO] = instantiate_trunk(\n            audio_embed_dim,\n            audio_num_blocks,\n            audio_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=audio_drop_path,\n        )\n        modality_trunks[ModalityType.DEPTH] = instantiate_trunk(\n            depth_embed_dim,\n            depth_num_blocks,\n            depth_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=depth_drop_path,\n        )\n        modality_trunks[ModalityType.THERMAL] = instantiate_trunk(\n            thermal_embed_dim,\n            thermal_num_blocks,\n            thermal_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=thermal_drop_path,\n        )\n        modality_trunks[ModalityType.IMU] = instantiate_trunk(\n            imu_embed_dim,\n            imu_num_blocks,\n            imu_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=imu_drop_path,\n        )\n\n        return nn.ModuleDict(modality_trunks)\n\n    def _create_modality_heads(\n        self,\n        out_embed_dim,\n        vision_embed_dim,\n        text_embed_dim,\n        audio_embed_dim,\n        depth_embed_dim,\n        thermal_embed_dim,\n        imu_embed_dim,\n    ):\n        modality_heads = {}\n\n        modality_heads[ModalityType.VISION] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=vision_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(vision_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.TEXT] = SelectEOSAndProject(\n            proj=nn.Sequential(\n                nn.LayerNorm(normalized_shape=text_embed_dim, eps=1e-6),\n                nn.Linear(text_embed_dim, out_embed_dim, bias=False),\n            )\n        )\n\n        modality_heads[ModalityType.AUDIO] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=audio_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(audio_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.DEPTH] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=depth_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(depth_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.THERMAL] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=thermal_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(thermal_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.IMU] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=imu_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Dropout(p=0.5),\n            nn.Linear(imu_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.POINT] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=512, eps=1e-6),\n            nn.Dropout(p=0.5),\n            nn.Linear(512, out_embed_dim, bias=False),\n        )\n\n        return nn.ModuleDict(modality_heads)\n\n    def _create_modality_postprocessors(self, out_embed_dim):\n        modality_postprocessors = {}\n\n        modality_postprocessors[ModalityType.VISION] = Normalize(dim=-1)\n        modality_postprocessors[ModalityType.TEXT] = nn.Sequential(\n            Normalize(dim=-1), LearnableLogitScaling(learnable=True)\n        )\n        modality_postprocessors[ModalityType.AUDIO] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=20.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.DEPTH] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=5.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.THERMAL] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=10.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.IMU] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=5.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.POINT] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=1.0, learnable=False),\n        )\n        return nn.ModuleDict(modality_postprocessors)\n\n    def forward(self, inputs, prenorm=False):\n        outputs = {}\n        outputs_prenorm = {}\n        for modality_key, modality_value in inputs.items():\n            reduce_list = (\n                modality_value.ndim >= 5\n            )  # Audio and Video inputs consist of multiple clips\n            if reduce_list:\n                B, S = modality_value.shape[:2]\n                modality_value = modality_value.reshape(\n                    B * S, *modality_value.shape[2:]\n                )\n\n            if modality_value is not None:\n                if modality_key == ModalityType.POINT:\n                    modality_value = self.point_trunk(modality_value)\n                    modality_value = self.modality_heads[modality_key](modality_value)\n                else:\n                    modality_value = self.modality_preprocessors[modality_key](\n                        **{modality_key: modality_value}\n                    )\n                    trunk_inputs = modality_value[\"trunk\"]\n                    head_inputs = modality_value[\"head\"]\n                    modality_value = self.modality_trunks[modality_key](**trunk_inputs)\n                    modality_value = self.modality_heads[modality_key](\n                        modality_value, **head_inputs\n                    )\n\n                modality_value_postnorm = self.modality_postprocessors[modality_key](\n                    modality_value\n                )\n\n\n                if reduce_list:\n                    if prenorm:\n                        modality_value = modality_value.reshape(B, S, -1)\n                        modality_value = modality_value.mean(dim=1)\n\n                    modality_value_postnorm = modality_value_postnorm.reshape(B, S, -1)\n                    modality_value_postnorm = modality_value_postnorm.mean(dim=1)\n\n                if prenorm:\n                    outputs_prenorm[modality_key] = modality_value\n                outputs[modality_key] = modality_value_postnorm\n\n        if prenorm:\n            return outputs, outputs_prenorm\n        else:\n            return outputs\n\n\ndef imagebind_huge(pretrained=False):\n    model = ImageBindModel(\n        vision_embed_dim=1280,\n        vision_num_blocks=32,\n        vision_num_heads=16,\n        text_embed_dim=1024,\n        text_num_blocks=24,\n        text_num_heads=16,\n        out_embed_dim=1024,\n        audio_drop_path=0.1,\n        imu_drop_path=0.7,\n    )\n\n    if pretrained:\n        if not os.path.exists(\"./ckpts/imagebind_w3D.pth\"):\n            print(\n                \"Downloading the pre-train weight of ImageBind with 3D encoder to ./ckpts/imagebind_w3D.pth ...\"\n            )","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.imagebind_model.imagebind_huge","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.imagebind_model.imagebind_huge#L509-L534","kind":"function","name":"imagebind_huge","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":509,"end_line":534,"context_start_line":489,"context_end_line":534,"code":"\n\n                if reduce_list:\n                    if prenorm:\n                        modality_value = modality_value.reshape(B, S, -1)\n                        modality_value = modality_value.mean(dim=1)\n\n                    modality_value_postnorm = modality_value_postnorm.reshape(B, S, -1)\n                    modality_value_postnorm = modality_value_postnorm.mean(dim=1)\n\n                if prenorm:\n                    outputs_prenorm[modality_key] = modality_value\n                outputs[modality_key] = modality_value_postnorm\n\n        if prenorm:\n            return outputs, outputs_prenorm\n        else:\n            return outputs\n\n\ndef imagebind_huge(pretrained=False):\n    model = ImageBindModel(\n        vision_embed_dim=1280,\n        vision_num_blocks=32,\n        vision_num_heads=16,\n        text_embed_dim=1024,\n        text_num_blocks=24,\n        text_num_heads=16,\n        out_embed_dim=1024,\n        audio_drop_path=0.1,\n        imu_drop_path=0.7,\n    )\n\n    if pretrained:\n        if not os.path.exists(\"./ckpts/imagebind_w3D.pth\"):\n            print(\n                \"Downloading the pre-train weight of ImageBind with 3D encoder to ./ckpts/imagebind_w3D.pth ...\"\n            )\n            os.makedirs(\"./ckpts\", exist_ok=True)\n            torch.hub.download_url_to_file(\n                \"https://huggingface.co/ZiyuG/ImageBind_w3D/resolve/e9206a10f118b0790c730264e7b3aa5c324c35cb/imagebind_w3D.pth\",\n                \"./ckpts/imagebind_w3D.pth\",\n                progress=True,\n            )\n        model.load_state_dict(torch.load(\"./ckpts/imagebind_w3D.pth\"))\n    return model","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.imagebind_model.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.imagebind_model.__init__#L41-L134","kind":"function","name":"__init__","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":41,"end_line":134,"context_start_line":21,"context_end_line":154,"code":"from .multimodal_preprocessors import (AudioPreprocessor,\n                                             IMUPreprocessor, PadIm2Video,\n                                             PatchEmbedGeneric,\n                                             RGBDTPreprocessor,\n                                             SpatioTemporalPosEmbeddingHelper,\n                                             TextPreprocessor,\n                                             ThermalPreprocessor)\nfrom .transformer import MultiheadAttention, SimpleTransformer\n\nModalityType = SimpleNamespace(\n    VISION=\"vision\",\n    TEXT=\"text\",\n    AUDIO=\"audio\",\n    THERMAL=\"thermal\",\n    DEPTH=\"depth\",\n    IMU=\"imu\",\n    POINT=\"point\",\n)\n\nclass ImageBindModel(nn.Module):\n    def __init__(\n        self,\n        video_frames=2,\n        kernel_size=(2, 14, 14),\n        audio_kernel_size=16,\n        audio_stride=10,\n        out_embed_dim=768,\n        vision_embed_dim=1024,\n        vision_num_blocks=24,\n        vision_num_heads=16,\n        audio_embed_dim=768,\n        audio_num_blocks=12,\n        audio_num_heads=12,\n        audio_num_mel_bins=128,\n        audio_target_len=204,\n        audio_drop_path=0.1,\n        text_embed_dim=768,\n        text_num_blocks=12,\n        text_num_heads=12,\n        depth_embed_dim=384,\n        depth_kernel_size=16,\n        depth_num_blocks=12,\n        depth_num_heads=8,\n        depth_drop_path=0.0,\n        thermal_embed_dim=768,\n        thermal_kernel_size=16,\n        thermal_num_blocks=12,\n        thermal_num_heads=12,\n        thermal_drop_path=0.0,\n        imu_embed_dim=512,\n        imu_kernel_size=8,\n        imu_num_blocks=6,\n        imu_num_heads=8,\n        imu_drop_path=0.7,\n    ):\n        super().__init__()\n\n        self.modality_preprocessors = self._create_modality_preprocessors(\n            video_frames,\n            vision_embed_dim,\n            kernel_size,\n            text_embed_dim,\n            audio_embed_dim,\n            audio_kernel_size,\n            audio_stride,\n            audio_num_mel_bins,\n            audio_target_len,\n            depth_embed_dim,\n            depth_kernel_size,\n            thermal_embed_dim,\n            thermal_kernel_size,\n            imu_embed_dim,\n        )\n\n        self.modality_trunks = self._create_modality_trunks(\n            vision_embed_dim,\n            vision_num_blocks,\n            vision_num_heads,\n            text_embed_dim,\n            text_num_blocks,\n            text_num_heads,\n            audio_embed_dim,\n            audio_num_blocks,\n            audio_num_heads,\n            audio_drop_path,\n            depth_embed_dim,\n            depth_num_blocks,\n            depth_num_heads,\n            depth_drop_path,\n            thermal_embed_dim,\n            thermal_num_blocks,\n            thermal_num_heads,\n            thermal_drop_path,\n            imu_embed_dim,\n            imu_num_blocks,\n            imu_num_heads,\n            imu_drop_path,\n        )\n\n        self.modality_heads = self._create_modality_heads(\n            out_embed_dim,\n            vision_embed_dim,\n            text_embed_dim,\n            audio_embed_dim,\n            depth_embed_dim,\n            thermal_embed_dim,\n            imu_embed_dim,\n        )\n\n        self.modality_postprocessors = self._create_modality_postprocessors(\n            out_embed_dim\n        )\n\n        self.point_trunk = PointTransformerBind()\n    \n    def _create_modality_preprocessors(\n        self,\n        video_frames=2,\n        vision_embed_dim=1024,\n        kernel_size=(2, 14, 14),\n        text_embed_dim=768,\n        audio_embed_dim=768,\n        audio_kernel_size=16,\n        audio_stride=10,\n        audio_num_mel_bins=128,\n        audio_target_len=204,\n        depth_embed_dim=768,\n        depth_kernel_size=16,\n        thermal_embed_dim=768,\n        thermal_kernel_size=16,\n        imu_embed_dim=512,\n    ):\n        rgbt_stem = PatchEmbedGeneric(\n            proj_stem=[","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.imagebind_model._create_modality_preprocessors","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.imagebind_model._create_modality_preprocessors#L136-L268","kind":"function","name":"_create_modality_preprocessors","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":136,"end_line":268,"context_start_line":116,"context_end_line":288,"code":"            imu_num_heads,\n            imu_drop_path,\n        )\n\n        self.modality_heads = self._create_modality_heads(\n            out_embed_dim,\n            vision_embed_dim,\n            text_embed_dim,\n            audio_embed_dim,\n            depth_embed_dim,\n            thermal_embed_dim,\n            imu_embed_dim,\n        )\n\n        self.modality_postprocessors = self._create_modality_postprocessors(\n            out_embed_dim\n        )\n\n        self.point_trunk = PointTransformerBind()\n    \n    def _create_modality_preprocessors(\n        self,\n        video_frames=2,\n        vision_embed_dim=1024,\n        kernel_size=(2, 14, 14),\n        text_embed_dim=768,\n        audio_embed_dim=768,\n        audio_kernel_size=16,\n        audio_stride=10,\n        audio_num_mel_bins=128,\n        audio_target_len=204,\n        depth_embed_dim=768,\n        depth_kernel_size=16,\n        thermal_embed_dim=768,\n        thermal_kernel_size=16,\n        imu_embed_dim=512,\n    ):\n        rgbt_stem = PatchEmbedGeneric(\n            proj_stem=[\n                PadIm2Video(pad_type=\"repeat\", ntimes=2),\n                nn.Conv3d(\n                    in_channels=3,\n                    kernel_size=kernel_size,\n                    out_channels=vision_embed_dim,\n                    stride=kernel_size,\n                    bias=False,\n                ),\n            ]\n        )\n        rgbt_preprocessor = RGBDTPreprocessor(\n            img_size=[3, video_frames, 224, 224],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            rgbt_stem=rgbt_stem,\n            depth_stem=None,\n        )\n\n        text_preprocessor = TextPreprocessor(\n            context_length=77,\n            vocab_size=49408,\n            embed_dim=text_embed_dim,\n            causal_masking=True,\n        )\n\n        audio_stem = PatchEmbedGeneric(\n            proj_stem=[\n                nn.Conv2d(\n                    in_channels=1,\n                    kernel_size=audio_kernel_size,\n                    stride=audio_stride,\n                    out_channels=audio_embed_dim,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=audio_embed_dim),\n        )\n        audio_preprocessor = AudioPreprocessor(\n            img_size=[1, audio_num_mel_bins, audio_target_len],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            audio_stem=audio_stem,\n        )\n\n        depth_stem = PatchEmbedGeneric(\n            [\n                nn.Conv2d(\n                    kernel_size=depth_kernel_size,\n                    in_channels=1,\n                    out_channels=depth_embed_dim,\n                    stride=depth_kernel_size,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=depth_embed_dim),\n        )\n\n        depth_preprocessor = RGBDTPreprocessor(\n            img_size=[1, 224, 224],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            rgbt_stem=None,\n            depth_stem=depth_stem,\n        )\n\n        thermal_stem = PatchEmbedGeneric(\n            [\n                nn.Conv2d(\n                    kernel_size=thermal_kernel_size,\n                    in_channels=1,\n                    out_channels=thermal_embed_dim,\n                    stride=thermal_kernel_size,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=thermal_embed_dim),\n        )\n        thermal_preprocessor = ThermalPreprocessor(\n            img_size=[1, 224, 224],\n            num_cls_tokens=1,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            thermal_stem=thermal_stem,\n        )\n\n        imu_stem = PatchEmbedGeneric(\n            [\n                nn.Linear(\n                    in_features=48,\n                    out_features=imu_embed_dim,\n                    bias=False,\n                ),\n            ],\n            norm_layer=nn.LayerNorm(normalized_shape=imu_embed_dim),\n        )\n\n        imu_preprocessor = IMUPreprocessor(\n            img_size=[6, 2000],\n            num_cls_tokens=1,\n            kernel_size=8,\n            embed_dim=imu_embed_dim,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            imu_stem=imu_stem,\n        )\n\n        modality_preprocessors = {\n            ModalityType.VISION: rgbt_preprocessor,\n            ModalityType.TEXT: text_preprocessor,\n            ModalityType.AUDIO: audio_preprocessor,\n            ModalityType.DEPTH: depth_preprocessor,\n            ModalityType.THERMAL: thermal_preprocessor,\n            ModalityType.IMU: imu_preprocessor,\n        }\n\n        return nn.ModuleDict(modality_preprocessors)\n\n    def _create_modality_trunks(\n        self,\n        vision_embed_dim=1024,\n        vision_num_blocks=24,\n        vision_num_heads=16,\n        text_embed_dim=768,\n        text_num_blocks=12,\n        text_num_heads=12,\n        audio_embed_dim=768,\n        audio_num_blocks=12,\n        audio_num_heads=12,\n        audio_drop_path=0.0,\n        depth_embed_dim=768,\n        depth_num_blocks=12,\n        depth_num_heads=12,\n        depth_drop_path=0.0,\n        thermal_embed_dim=768,\n        thermal_num_blocks=12,\n        thermal_num_heads=12,","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.imagebind_model._create_modality_trunks","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.imagebind_model._create_modality_trunks#L270-L369","kind":"function","name":"_create_modality_trunks","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":270,"end_line":369,"context_start_line":250,"context_end_line":389,"code":"        imu_preprocessor = IMUPreprocessor(\n            img_size=[6, 2000],\n            num_cls_tokens=1,\n            kernel_size=8,\n            embed_dim=imu_embed_dim,\n            pos_embed_fn=partial(SpatioTemporalPosEmbeddingHelper, learnable=True),\n            imu_stem=imu_stem,\n        )\n\n        modality_preprocessors = {\n            ModalityType.VISION: rgbt_preprocessor,\n            ModalityType.TEXT: text_preprocessor,\n            ModalityType.AUDIO: audio_preprocessor,\n            ModalityType.DEPTH: depth_preprocessor,\n            ModalityType.THERMAL: thermal_preprocessor,\n            ModalityType.IMU: imu_preprocessor,\n        }\n\n        return nn.ModuleDict(modality_preprocessors)\n\n    def _create_modality_trunks(\n        self,\n        vision_embed_dim=1024,\n        vision_num_blocks=24,\n        vision_num_heads=16,\n        text_embed_dim=768,\n        text_num_blocks=12,\n        text_num_heads=12,\n        audio_embed_dim=768,\n        audio_num_blocks=12,\n        audio_num_heads=12,\n        audio_drop_path=0.0,\n        depth_embed_dim=768,\n        depth_num_blocks=12,\n        depth_num_heads=12,\n        depth_drop_path=0.0,\n        thermal_embed_dim=768,\n        thermal_num_blocks=12,\n        thermal_num_heads=12,\n        thermal_drop_path=0.0,\n        imu_embed_dim=512,\n        imu_num_blocks=6,\n        imu_num_heads=8,\n        imu_drop_path=0.7,\n    ):\n        def instantiate_trunk(\n            embed_dim, num_blocks, num_heads, pre_transformer_ln, add_bias_kv, drop_path\n        ):\n            return SimpleTransformer(\n                embed_dim=embed_dim,\n                num_blocks=num_blocks,\n                ffn_dropout_rate=0.0,\n                drop_path_rate=drop_path,\n                attn_target=partial(\n                    MultiheadAttention,\n                    embed_dim=embed_dim,\n                    num_heads=num_heads,\n                    bias=True,\n                    add_bias_kv=add_bias_kv,\n                ),\n                pre_transformer_layer=nn.Sequential(\n                    nn.LayerNorm(embed_dim, eps=1e-6)\n                    if pre_transformer_ln\n                    else nn.Identity(),\n                    EinOpsRearrange(\"b l d -> l b d\"),\n                ),\n                post_transformer_layer=EinOpsRearrange(\"l b d -> b l d\"),\n            )\n\n        modality_trunks = {}\n        modality_trunks[ModalityType.VISION] = instantiate_trunk(\n            vision_embed_dim,\n            vision_num_blocks,\n            vision_num_heads,\n            pre_transformer_ln=True,\n            add_bias_kv=False,\n            drop_path=0.0,\n        )\n        modality_trunks[ModalityType.TEXT] = instantiate_trunk(\n            text_embed_dim,\n            text_num_blocks,\n            text_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=False,\n            drop_path=0.0,\n        )\n        modality_trunks[ModalityType.AUDIO] = instantiate_trunk(\n            audio_embed_dim,\n            audio_num_blocks,\n            audio_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=audio_drop_path,\n        )\n        modality_trunks[ModalityType.DEPTH] = instantiate_trunk(\n            depth_embed_dim,\n            depth_num_blocks,\n            depth_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=depth_drop_path,\n        )\n        modality_trunks[ModalityType.THERMAL] = instantiate_trunk(\n            thermal_embed_dim,\n            thermal_num_blocks,\n            thermal_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=thermal_drop_path,\n        )\n        modality_trunks[ModalityType.IMU] = instantiate_trunk(\n            imu_embed_dim,\n            imu_num_blocks,\n            imu_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=imu_drop_path,\n        )\n\n        return nn.ModuleDict(modality_trunks)\n\n    def _create_modality_heads(\n        self,\n        out_embed_dim,\n        vision_embed_dim,\n        text_embed_dim,\n        audio_embed_dim,\n        depth_embed_dim,\n        thermal_embed_dim,\n        imu_embed_dim,\n    ):\n        modality_heads = {}\n\n        modality_heads[ModalityType.VISION] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=vision_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(vision_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.TEXT] = SelectEOSAndProject(","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.imagebind_model._create_modality_heads","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.imagebind_model._create_modality_heads#L371-L427","kind":"function","name":"_create_modality_heads","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":371,"end_line":427,"context_start_line":351,"context_end_line":447,"code":"        )\n        modality_trunks[ModalityType.THERMAL] = instantiate_trunk(\n            thermal_embed_dim,\n            thermal_num_blocks,\n            thermal_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=thermal_drop_path,\n        )\n        modality_trunks[ModalityType.IMU] = instantiate_trunk(\n            imu_embed_dim,\n            imu_num_blocks,\n            imu_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=True,\n            drop_path=imu_drop_path,\n        )\n\n        return nn.ModuleDict(modality_trunks)\n\n    def _create_modality_heads(\n        self,\n        out_embed_dim,\n        vision_embed_dim,\n        text_embed_dim,\n        audio_embed_dim,\n        depth_embed_dim,\n        thermal_embed_dim,\n        imu_embed_dim,\n    ):\n        modality_heads = {}\n\n        modality_heads[ModalityType.VISION] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=vision_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(vision_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.TEXT] = SelectEOSAndProject(\n            proj=nn.Sequential(\n                nn.LayerNorm(normalized_shape=text_embed_dim, eps=1e-6),\n                nn.Linear(text_embed_dim, out_embed_dim, bias=False),\n            )\n        )\n\n        modality_heads[ModalityType.AUDIO] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=audio_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(audio_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.DEPTH] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=depth_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(depth_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.THERMAL] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=thermal_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(thermal_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.IMU] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=imu_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Dropout(p=0.5),\n            nn.Linear(imu_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.POINT] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=512, eps=1e-6),\n            nn.Dropout(p=0.5),\n            nn.Linear(512, out_embed_dim, bias=False),\n        )\n\n        return nn.ModuleDict(modality_heads)\n\n    def _create_modality_postprocessors(self, out_embed_dim):\n        modality_postprocessors = {}\n\n        modality_postprocessors[ModalityType.VISION] = Normalize(dim=-1)\n        modality_postprocessors[ModalityType.TEXT] = nn.Sequential(\n            Normalize(dim=-1), LearnableLogitScaling(learnable=True)\n        )\n        modality_postprocessors[ModalityType.AUDIO] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=20.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.DEPTH] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=5.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.THERMAL] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=10.0, learnable=False),\n        )","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.imagebind_model._create_modality_postprocessors","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.imagebind_model._create_modality_postprocessors#L429-L456","kind":"function","name":"_create_modality_postprocessors","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":429,"end_line":456,"context_start_line":409,"context_end_line":476,"code":"            nn.LayerNorm(normalized_shape=thermal_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Linear(thermal_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.IMU] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=imu_embed_dim, eps=1e-6),\n            SelectElement(index=0),\n            nn.Dropout(p=0.5),\n            nn.Linear(imu_embed_dim, out_embed_dim, bias=False),\n        )\n\n        modality_heads[ModalityType.POINT] = nn.Sequential(\n            nn.LayerNorm(normalized_shape=512, eps=1e-6),\n            nn.Dropout(p=0.5),\n            nn.Linear(512, out_embed_dim, bias=False),\n        )\n\n        return nn.ModuleDict(modality_heads)\n\n    def _create_modality_postprocessors(self, out_embed_dim):\n        modality_postprocessors = {}\n\n        modality_postprocessors[ModalityType.VISION] = Normalize(dim=-1)\n        modality_postprocessors[ModalityType.TEXT] = nn.Sequential(\n            Normalize(dim=-1), LearnableLogitScaling(learnable=True)\n        )\n        modality_postprocessors[ModalityType.AUDIO] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=20.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.DEPTH] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=5.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.THERMAL] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=10.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.IMU] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=5.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.POINT] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=1.0, learnable=False),\n        )\n        return nn.ModuleDict(modality_postprocessors)\n\n    def forward(self, inputs, prenorm=False):\n        outputs = {}\n        outputs_prenorm = {}\n        for modality_key, modality_value in inputs.items():\n            reduce_list = (\n                modality_value.ndim >= 5\n            )  # Audio and Video inputs consist of multiple clips\n            if reduce_list:\n                B, S = modality_value.shape[:2]\n                modality_value = modality_value.reshape(\n                    B * S, *modality_value.shape[2:]\n                )\n\n            if modality_value is not None:\n                if modality_key == ModalityType.POINT:\n                    modality_value = self.point_trunk(modality_value)\n                    modality_value = self.modality_heads[modality_key](modality_value)\n                else:\n                    modality_value = self.modality_preprocessors[modality_key](","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.imagebind_model.forward","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.imagebind_model.forward#L458-L506","kind":"function","name":"forward","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":458,"end_line":506,"context_start_line":438,"context_end_line":526,"code":"            LearnableLogitScaling(logit_scale_init=20.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.DEPTH] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=5.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.THERMAL] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=10.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.IMU] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=5.0, learnable=False),\n        )\n        modality_postprocessors[ModalityType.POINT] = nn.Sequential(\n            Normalize(dim=-1),\n            LearnableLogitScaling(logit_scale_init=1.0, learnable=False),\n        )\n        return nn.ModuleDict(modality_postprocessors)\n\n    def forward(self, inputs, prenorm=False):\n        outputs = {}\n        outputs_prenorm = {}\n        for modality_key, modality_value in inputs.items():\n            reduce_list = (\n                modality_value.ndim >= 5\n            )  # Audio and Video inputs consist of multiple clips\n            if reduce_list:\n                B, S = modality_value.shape[:2]\n                modality_value = modality_value.reshape(\n                    B * S, *modality_value.shape[2:]\n                )\n\n            if modality_value is not None:\n                if modality_key == ModalityType.POINT:\n                    modality_value = self.point_trunk(modality_value)\n                    modality_value = self.modality_heads[modality_key](modality_value)\n                else:\n                    modality_value = self.modality_preprocessors[modality_key](\n                        **{modality_key: modality_value}\n                    )\n                    trunk_inputs = modality_value[\"trunk\"]\n                    head_inputs = modality_value[\"head\"]\n                    modality_value = self.modality_trunks[modality_key](**trunk_inputs)\n                    modality_value = self.modality_heads[modality_key](\n                        modality_value, **head_inputs\n                    )\n\n                modality_value_postnorm = self.modality_postprocessors[modality_key](\n                    modality_value\n                )\n\n\n                if reduce_list:\n                    if prenorm:\n                        modality_value = modality_value.reshape(B, S, -1)\n                        modality_value = modality_value.mean(dim=1)\n\n                    modality_value_postnorm = modality_value_postnorm.reshape(B, S, -1)\n                    modality_value_postnorm = modality_value_postnorm.mean(dim=1)\n\n                if prenorm:\n                    outputs_prenorm[modality_key] = modality_value\n                outputs[modality_key] = modality_value_postnorm\n\n        if prenorm:\n            return outputs, outputs_prenorm\n        else:\n            return outputs\n\n\ndef imagebind_huge(pretrained=False):\n    model = ImageBindModel(\n        vision_embed_dim=1280,\n        vision_num_blocks=32,\n        vision_num_heads=16,\n        text_embed_dim=1024,\n        text_num_blocks=24,\n        text_num_heads=16,\n        out_embed_dim=1024,\n        audio_drop_path=0.1,\n        imu_drop_path=0.7,\n    )\n\n    if pretrained:\n        if not os.path.exists(\"./ckpts/imagebind_w3D.pth\"):\n            print(\n                \"Downloading the pre-train weight of ImageBind with 3D encoder to ./ckpts/imagebind_w3D.pth ...\"\n            )","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.imagebind_model.instantiate_trunk","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.imagebind_model.instantiate_trunk#L295-L317","kind":"function","name":"instantiate_trunk","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":295,"end_line":317,"context_start_line":275,"context_end_line":337,"code":"        text_embed_dim=768,\n        text_num_blocks=12,\n        text_num_heads=12,\n        audio_embed_dim=768,\n        audio_num_blocks=12,\n        audio_num_heads=12,\n        audio_drop_path=0.0,\n        depth_embed_dim=768,\n        depth_num_blocks=12,\n        depth_num_heads=12,\n        depth_drop_path=0.0,\n        thermal_embed_dim=768,\n        thermal_num_blocks=12,\n        thermal_num_heads=12,\n        thermal_drop_path=0.0,\n        imu_embed_dim=512,\n        imu_num_blocks=6,\n        imu_num_heads=8,\n        imu_drop_path=0.7,\n    ):\n        def instantiate_trunk(\n            embed_dim, num_blocks, num_heads, pre_transformer_ln, add_bias_kv, drop_path\n        ):\n            return SimpleTransformer(\n                embed_dim=embed_dim,\n                num_blocks=num_blocks,\n                ffn_dropout_rate=0.0,\n                drop_path_rate=drop_path,\n                attn_target=partial(\n                    MultiheadAttention,\n                    embed_dim=embed_dim,\n                    num_heads=num_heads,\n                    bias=True,\n                    add_bias_kv=add_bias_kv,\n                ),\n                pre_transformer_layer=nn.Sequential(\n                    nn.LayerNorm(embed_dim, eps=1e-6)\n                    if pre_transformer_ln\n                    else nn.Identity(),\n                    EinOpsRearrange(\"b l d -> l b d\"),\n                ),\n                post_transformer_layer=EinOpsRearrange(\"l b d -> b l d\"),\n            )\n\n        modality_trunks = {}\n        modality_trunks[ModalityType.VISION] = instantiate_trunk(\n            vision_embed_dim,\n            vision_num_blocks,\n            vision_num_heads,\n            pre_transformer_ln=True,\n            add_bias_kv=False,\n            drop_path=0.0,\n        )\n        modality_trunks[ModalityType.TEXT] = instantiate_trunk(\n            text_embed_dim,\n            text_num_blocks,\n            text_num_heads,\n            pre_transformer_ln=False,\n            add_bias_kv=False,\n            drop_path=0.0,\n        )\n        modality_trunks[ModalityType.AUDIO] = instantiate_trunk(\n            audio_embed_dim,","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.models.pointbert.misc#L1-L286","kind":"module","name":"imagebind_LLM.ImageBind.models.pointbert.misc","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":1,"end_line":286,"context_start_line":1,"context_end_line":286,"code":"import numpy as np\nimport matplotlib.pyplot as plt\nfrom mpl_toolkits.mplot3d import Axes3D\nimport random\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport os\nfrom collections import abc\n\n\n# def fps(data, number):\n#     '''\n#         data B N 3\n#         number int\n#     '''\n#     fps_idx = pointnet2_utils.furthest_point_sample(data, number)\n#     fps_data = pointnet2_utils.gather_operation(data.transpose(1, 2).contiguous(), fps_idx).transpose(1,2).contiguous()\n#     return fps_data\n\ndef index_points(points, idx):\n    \"\"\"\n    Input:\n        points: input points data, [B, N, C]\n        idx: sample index data, [B, S]\n    Return:\n        new_points:, indexed points data, [B, S, C]\n    \"\"\"\n    device = points.device\n    B = points.shape[0]\n    view_shape = list(idx.shape)\n    view_shape[1:] = [1] * (len(view_shape) - 1)\n    repeat_shape = list(idx.shape)\n    repeat_shape[0] = 1\n    batch_indices = torch.arange(B, dtype=torch.long).to(device).view(view_shape).repeat(repeat_shape)\n    new_points = points[batch_indices, idx, :]\n    return new_points\n\ndef fps(xyz, npoint):\n    \"\"\"\n    Input:\n        xyz: pointcloud data, [B, N, 3]\n        npoint: number of samples\n    Return:\n        centroids: sampled pointcloud index, [B, npoint]\n    \"\"\"\n    device = xyz.device\n    B, N, C = xyz.shape\n    centroids = torch.zeros(B, npoint, dtype=torch.long).to(device)\n    distance = torch.ones(B, N).to(device) * 1e10\n    farthest = torch.randint(0, N, (B,), dtype=torch.long).to(device)\n    batch_indices = torch.arange(B, dtype=torch.long).to(device)\n    for i in range(npoint):\n        centroids[:, i] = farthest\n        centroid = xyz[batch_indices, farthest, :].view(B, 1, 3)\n        dist = torch.sum((xyz - centroid) ** 2, -1)\n        distance = torch.min(distance, dist)\n        farthest = torch.max(distance, -1)[1]\n    return index_points(xyz, centroids)\n\ndef worker_init_fn(worker_id):\n    np.random.seed(np.random.get_state()[1][0] + worker_id)\n\ndef build_lambda_sche(opti, config):\n    if config.get('decay_step') is not None:\n        lr_lbmd = lambda e: max(config.lr_decay ** (e / config.decay_step), config.lowest_decay)\n        scheduler = torch.optim.lr_scheduler.LambdaLR(opti, lr_lbmd)\n    else:\n        raise NotImplementedError()\n    return scheduler\n\ndef build_lambda_bnsche(model, config):\n    if config.get('decay_step') is not None:\n        bnm_lmbd = lambda e: max(config.bn_momentum * config.bn_decay ** (e / config.decay_step), config.lowest_decay)\n        bnm_scheduler = BNMomentumScheduler(model, bnm_lmbd)\n    else:\n        raise NotImplementedError()\n    return bnm_scheduler\n    \ndef set_random_seed(seed, deterministic=False):\n    \"\"\"Set random seed.\n    Args:\n        seed (int): Seed to be used.\n        deterministic (bool): Whether to set the deterministic option for\n            CUDNN backend, i.e., set `torch.backends.cudnn.deterministic`\n            to True and `torch.backends.cudnn.benchmark` to False.\n            Default: False.\n\n    # Speed-reproducibility tradeoff https://pytorch.org/docs/stable/notes/randomness.html\n    if cuda_deterministic:  # slower, more reproducible\n        cudnn.deterministic = True\n        cudnn.benchmark = False\n    else:  # faster, less reproducible\n        cudnn.deterministic = False\n        cudnn.benchmark = True\n\n    \"\"\"\n    random.seed(seed)\n    np.random.seed(seed)\n    torch.manual_seed(seed)\n    torch.cuda.manual_seed_all(seed)\n    if deterministic:\n        torch.backends.cudnn.deterministic = True\n        torch.backends.cudnn.benchmark = False\n\n\ndef is_seq_of(seq, expected_type, seq_type=None):\n    \"\"\"Check whether it is a sequence of some type.\n    Args:\n        seq (Sequence): The sequence to be checked.\n        expected_type (type): Expected type of sequence items.\n        seq_type (type, optional): Expected sequence type.\n    Returns:\n        bool: Whether the sequence is valid.\n    \"\"\"\n    if seq_type is None:\n        exp_seq_type = abc.Sequence\n    else:\n        assert isinstance(seq_type, type)\n        exp_seq_type = seq_type\n    if not isinstance(seq, exp_seq_type):\n        return False\n    for item in seq:\n        if not isinstance(item, expected_type):\n            return False\n    return True\n\n\ndef set_bn_momentum_default(bn_momentum):\n    def fn(m):\n        if isinstance(m, (nn.BatchNorm1d, nn.BatchNorm2d, nn.BatchNorm3d)):\n            m.momentum = bn_momentum\n    return fn\n\nclass BNMomentumScheduler(object):\n\n    def __init__(\n            self, model, bn_lambda, last_epoch=-1,\n            setter=set_bn_momentum_default\n    ):\n        if not isinstance(model, nn.Module):\n            raise RuntimeError(\n                \"Class '{}' is not a PyTorch nn Module\".format(\n                    type(model).__name__\n                )\n            )\n\n        self.model = model\n        self.setter = setter\n        self.lmbd = bn_lambda\n\n        self.step(last_epoch + 1)\n        self.last_epoch = last_epoch\n\n    def step(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n\n        self.last_epoch = epoch\n        self.model.apply(self.setter(self.lmbd(epoch)))\n\n    def get_momentum(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n        return self.lmbd(epoch)\n\n\n\ndef seprate_point_cloud(xyz, num_points, crop, fixed_points = None, padding_zeros = False):\n    '''\n     seprate point cloud: usage : using to generate the incomplete point cloud with a setted number.\n    '''\n    _,n,c = xyz.shape\n\n    assert n == num_points\n    assert c == 3\n    if crop == num_points:\n        return xyz, None\n        \n    INPUT = []\n    CROP = []\n    for points in xyz:\n        if isinstance(crop,list):\n            num_crop = random.randint(crop[0],crop[1])\n        else:\n            num_crop = crop\n\n        points = points.unsqueeze(0)\n\n        if fixed_points is None:       \n            center = F.normalize(torch.randn(1,1,3),p=2,dim=-1).cuda()\n        else:\n            if isinstance(fixed_points,list):\n                fixed_point = random.sample(fixed_points,1)[0]\n            else:\n                fixed_point = fixed_points\n            center = fixed_point.reshape(1,1,3).cuda()\n\n        distance_matrix = torch.norm(center.unsqueeze(2) - points.unsqueeze(1), p =2 ,dim = -1)  # 1 1 2048\n\n        idx = torch.argsort(distance_matrix,dim=-1, descending=False)[0,0] # 2048\n\n        if padding_zeros:\n            input_data = points.clone()\n            input_data[0, idx[:num_crop]] =  input_data[0,idx[:num_crop]] * 0\n\n        else:\n            input_data = points.clone()[0, idx[num_crop:]].unsqueeze(0) # 1 N 3\n\n        crop_data =  points.clone()[0, idx[:num_crop]].unsqueeze(0)\n\n        if isinstance(crop,list):\n            INPUT.append(fps(input_data,2048))\n            CROP.append(fps(crop_data,2048))\n        else:\n            INPUT.append(input_data)\n            CROP.append(crop_data)\n\n    input_data = torch.cat(INPUT,dim=0)# B N 3\n    crop_data = torch.cat(CROP,dim=0)# B M 3\n\n    return input_data.contiguous(), crop_data.contiguous()\n\ndef get_ptcloud_img(ptcloud):\n    fig = plt.figure(figsize=(8, 8))\n\n    x, z, y = ptcloud.transpose(1, 0)\n    ax = fig.gca(projection=Axes3D.name, adjustable='box')\n    ax.axis('off')\n    # ax.axis('scaled')\n    ax.view_init(30, 45)\n    max, min = np.max(ptcloud), np.min(ptcloud)\n    ax.set_xbound(min, max)\n    ax.set_ybound(min, max)\n    ax.set_zbound(min, max)\n    ax.scatter(x, y, z, zdir='z', c=x, cmap='jet')\n\n    fig.canvas.draw()\n    img = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')\n    img = img.reshape(fig.canvas.get_width_height()[::-1] + (3, ))\n    return img\n\n\n\ndef visualize_KITTI(path, data_list, titles = ['input','pred'], cmap=['bwr','autumn'], zdir='y', \n                         xlim=(-1, 1), ylim=(-1, 1), zlim=(-1, 1) ):\n    fig = plt.figure(figsize=(6*len(data_list),6))\n    cmax = data_list[-1][:,0].max()\n\n    for i in range(len(data_list)):\n        data = data_list[i][:-2048] if i == 1 else data_list[i]\n        color = data[:,0] /cmax\n        ax = fig.add_subplot(1, len(data_list) , i + 1, projection='3d')\n        ax.view_init(30, -120)\n        b = ax.scatter(data[:, 0], data[:, 1], data[:, 2], zdir=zdir, c=color,vmin=-1,vmax=1 ,cmap = cmap[0],s=4,linewidth=0.05, edgecolors = 'black')\n        ax.set_title(titles[i])\n\n        ax.set_axis_off()\n        ax.set_xlim(xlim)\n        ax.set_ylim(ylim)\n        ax.set_zlim(zlim)\n    plt.subplots_adjust(left=0, right=1, bottom=0, top=1, wspace=0.2, hspace=0)\n    if not os.path.exists(path):\n        os.makedirs(path)\n\n    pic_path = path + '.png'\n    fig.savefig(pic_path)\n\n    np.save(os.path.join(path, 'input.npy'), data_list[0].numpy())\n    np.save(os.path.join(path, 'pred.npy'), data_list[1].numpy())\n    plt.close(fig)\n\n\ndef random_dropping(pc, e):\n    up_num = max(64, 768 // (e//50 + 1))\n    pc = pc\n    random_num = torch.randint(1, up_num, (1,1))[0,0]\n    pc = fps(pc, random_num)\n    padding = torch.zeros(pc.size(0), 2048 - pc.size(1), 3).to(pc.device)\n    pc = torch.cat([pc, padding], dim = 1)\n    return pc\n    \n\ndef random_scale(partial, scale_range=[0.8, 1.2]):\n    scale = torch.rand(1).cuda() * (scale_range[1] - scale_range[0]) + scale_range[0]\n    return partial * scale","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.index_points","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.index_points#L21-L37","kind":"function","name":"index_points","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":21,"end_line":37,"context_start_line":1,"context_end_line":57,"code":"import numpy as np\nimport matplotlib.pyplot as plt\nfrom mpl_toolkits.mplot3d import Axes3D\nimport random\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport os\nfrom collections import abc\n\n\n# def fps(data, number):\n#     '''\n#         data B N 3\n#         number int\n#     '''\n#     fps_idx = pointnet2_utils.furthest_point_sample(data, number)\n#     fps_data = pointnet2_utils.gather_operation(data.transpose(1, 2).contiguous(), fps_idx).transpose(1,2).contiguous()\n#     return fps_data\n\ndef index_points(points, idx):\n    \"\"\"\n    Input:\n        points: input points data, [B, N, C]\n        idx: sample index data, [B, S]\n    Return:\n        new_points:, indexed points data, [B, S, C]\n    \"\"\"\n    device = points.device\n    B = points.shape[0]\n    view_shape = list(idx.shape)\n    view_shape[1:] = [1] * (len(view_shape) - 1)\n    repeat_shape = list(idx.shape)\n    repeat_shape[0] = 1\n    batch_indices = torch.arange(B, dtype=torch.long).to(device).view(view_shape).repeat(repeat_shape)\n    new_points = points[batch_indices, idx, :]\n    return new_points\n\ndef fps(xyz, npoint):\n    \"\"\"\n    Input:\n        xyz: pointcloud data, [B, N, 3]\n        npoint: number of samples\n    Return:\n        centroids: sampled pointcloud index, [B, npoint]\n    \"\"\"\n    device = xyz.device\n    B, N, C = xyz.shape\n    centroids = torch.zeros(B, npoint, dtype=torch.long).to(device)\n    distance = torch.ones(B, N).to(device) * 1e10\n    farthest = torch.randint(0, N, (B,), dtype=torch.long).to(device)\n    batch_indices = torch.arange(B, dtype=torch.long).to(device)\n    for i in range(npoint):\n        centroids[:, i] = farthest\n        centroid = xyz[batch_indices, farthest, :].view(B, 1, 3)\n        dist = torch.sum((xyz - centroid) ** 2, -1)\n        distance = torch.min(distance, dist)","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.fps","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.fps#L39-L59","kind":"function","name":"fps","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":39,"end_line":59,"context_start_line":19,"context_end_line":79,"code":"#     return fps_data\n\ndef index_points(points, idx):\n    \"\"\"\n    Input:\n        points: input points data, [B, N, C]\n        idx: sample index data, [B, S]\n    Return:\n        new_points:, indexed points data, [B, S, C]\n    \"\"\"\n    device = points.device\n    B = points.shape[0]\n    view_shape = list(idx.shape)\n    view_shape[1:] = [1] * (len(view_shape) - 1)\n    repeat_shape = list(idx.shape)\n    repeat_shape[0] = 1\n    batch_indices = torch.arange(B, dtype=torch.long).to(device).view(view_shape).repeat(repeat_shape)\n    new_points = points[batch_indices, idx, :]\n    return new_points\n\ndef fps(xyz, npoint):\n    \"\"\"\n    Input:\n        xyz: pointcloud data, [B, N, 3]\n        npoint: number of samples\n    Return:\n        centroids: sampled pointcloud index, [B, npoint]\n    \"\"\"\n    device = xyz.device\n    B, N, C = xyz.shape\n    centroids = torch.zeros(B, npoint, dtype=torch.long).to(device)\n    distance = torch.ones(B, N).to(device) * 1e10\n    farthest = torch.randint(0, N, (B,), dtype=torch.long).to(device)\n    batch_indices = torch.arange(B, dtype=torch.long).to(device)\n    for i in range(npoint):\n        centroids[:, i] = farthest\n        centroid = xyz[batch_indices, farthest, :].view(B, 1, 3)\n        dist = torch.sum((xyz - centroid) ** 2, -1)\n        distance = torch.min(distance, dist)\n        farthest = torch.max(distance, -1)[1]\n    return index_points(xyz, centroids)\n\ndef worker_init_fn(worker_id):\n    np.random.seed(np.random.get_state()[1][0] + worker_id)\n\ndef build_lambda_sche(opti, config):\n    if config.get('decay_step') is not None:\n        lr_lbmd = lambda e: max(config.lr_decay ** (e / config.decay_step), config.lowest_decay)\n        scheduler = torch.optim.lr_scheduler.LambdaLR(opti, lr_lbmd)\n    else:\n        raise NotImplementedError()\n    return scheduler\n\ndef build_lambda_bnsche(model, config):\n    if config.get('decay_step') is not None:\n        bnm_lmbd = lambda e: max(config.bn_momentum * config.bn_decay ** (e / config.decay_step), config.lowest_decay)\n        bnm_scheduler = BNMomentumScheduler(model, bnm_lmbd)\n    else:\n        raise NotImplementedError()\n    return bnm_scheduler\n    ","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.worker_init_fn","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.worker_init_fn#L61-L62","kind":"function","name":"worker_init_fn","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":61,"end_line":62,"context_start_line":41,"context_end_line":82,"code":"    Input:\n        xyz: pointcloud data, [B, N, 3]\n        npoint: number of samples\n    Return:\n        centroids: sampled pointcloud index, [B, npoint]\n    \"\"\"\n    device = xyz.device\n    B, N, C = xyz.shape\n    centroids = torch.zeros(B, npoint, dtype=torch.long).to(device)\n    distance = torch.ones(B, N).to(device) * 1e10\n    farthest = torch.randint(0, N, (B,), dtype=torch.long).to(device)\n    batch_indices = torch.arange(B, dtype=torch.long).to(device)\n    for i in range(npoint):\n        centroids[:, i] = farthest\n        centroid = xyz[batch_indices, farthest, :].view(B, 1, 3)\n        dist = torch.sum((xyz - centroid) ** 2, -1)\n        distance = torch.min(distance, dist)\n        farthest = torch.max(distance, -1)[1]\n    return index_points(xyz, centroids)\n\ndef worker_init_fn(worker_id):\n    np.random.seed(np.random.get_state()[1][0] + worker_id)\n\ndef build_lambda_sche(opti, config):\n    if config.get('decay_step') is not None:\n        lr_lbmd = lambda e: max(config.lr_decay ** (e / config.decay_step), config.lowest_decay)\n        scheduler = torch.optim.lr_scheduler.LambdaLR(opti, lr_lbmd)\n    else:\n        raise NotImplementedError()\n    return scheduler\n\ndef build_lambda_bnsche(model, config):\n    if config.get('decay_step') is not None:\n        bnm_lmbd = lambda e: max(config.bn_momentum * config.bn_decay ** (e / config.decay_step), config.lowest_decay)\n        bnm_scheduler = BNMomentumScheduler(model, bnm_lmbd)\n    else:\n        raise NotImplementedError()\n    return bnm_scheduler\n    \ndef set_random_seed(seed, deterministic=False):\n    \"\"\"Set random seed.\n    Args:","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.build_lambda_sche","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.build_lambda_sche#L64-L70","kind":"function","name":"build_lambda_sche","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":64,"end_line":70,"context_start_line":44,"context_end_line":90,"code":"    Return:\n        centroids: sampled pointcloud index, [B, npoint]\n    \"\"\"\n    device = xyz.device\n    B, N, C = xyz.shape\n    centroids = torch.zeros(B, npoint, dtype=torch.long).to(device)\n    distance = torch.ones(B, N).to(device) * 1e10\n    farthest = torch.randint(0, N, (B,), dtype=torch.long).to(device)\n    batch_indices = torch.arange(B, dtype=torch.long).to(device)\n    for i in range(npoint):\n        centroids[:, i] = farthest\n        centroid = xyz[batch_indices, farthest, :].view(B, 1, 3)\n        dist = torch.sum((xyz - centroid) ** 2, -1)\n        distance = torch.min(distance, dist)\n        farthest = torch.max(distance, -1)[1]\n    return index_points(xyz, centroids)\n\ndef worker_init_fn(worker_id):\n    np.random.seed(np.random.get_state()[1][0] + worker_id)\n\ndef build_lambda_sche(opti, config):\n    if config.get('decay_step') is not None:\n        lr_lbmd = lambda e: max(config.lr_decay ** (e / config.decay_step), config.lowest_decay)\n        scheduler = torch.optim.lr_scheduler.LambdaLR(opti, lr_lbmd)\n    else:\n        raise NotImplementedError()\n    return scheduler\n\ndef build_lambda_bnsche(model, config):\n    if config.get('decay_step') is not None:\n        bnm_lmbd = lambda e: max(config.bn_momentum * config.bn_decay ** (e / config.decay_step), config.lowest_decay)\n        bnm_scheduler = BNMomentumScheduler(model, bnm_lmbd)\n    else:\n        raise NotImplementedError()\n    return bnm_scheduler\n    \ndef set_random_seed(seed, deterministic=False):\n    \"\"\"Set random seed.\n    Args:\n        seed (int): Seed to be used.\n        deterministic (bool): Whether to set the deterministic option for\n            CUDNN backend, i.e., set `torch.backends.cudnn.deterministic`\n            to True and `torch.backends.cudnn.benchmark` to False.\n            Default: False.\n\n    # Speed-reproducibility tradeoff https://pytorch.org/docs/stable/notes/randomness.html\n    if cuda_deterministic:  # slower, more reproducible","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.build_lambda_bnsche","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.build_lambda_bnsche#L72-L78","kind":"function","name":"build_lambda_bnsche","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":72,"end_line":78,"context_start_line":52,"context_end_line":98,"code":"    batch_indices = torch.arange(B, dtype=torch.long).to(device)\n    for i in range(npoint):\n        centroids[:, i] = farthest\n        centroid = xyz[batch_indices, farthest, :].view(B, 1, 3)\n        dist = torch.sum((xyz - centroid) ** 2, -1)\n        distance = torch.min(distance, dist)\n        farthest = torch.max(distance, -1)[1]\n    return index_points(xyz, centroids)\n\ndef worker_init_fn(worker_id):\n    np.random.seed(np.random.get_state()[1][0] + worker_id)\n\ndef build_lambda_sche(opti, config):\n    if config.get('decay_step') is not None:\n        lr_lbmd = lambda e: max(config.lr_decay ** (e / config.decay_step), config.lowest_decay)\n        scheduler = torch.optim.lr_scheduler.LambdaLR(opti, lr_lbmd)\n    else:\n        raise NotImplementedError()\n    return scheduler\n\ndef build_lambda_bnsche(model, config):\n    if config.get('decay_step') is not None:\n        bnm_lmbd = lambda e: max(config.bn_momentum * config.bn_decay ** (e / config.decay_step), config.lowest_decay)\n        bnm_scheduler = BNMomentumScheduler(model, bnm_lmbd)\n    else:\n        raise NotImplementedError()\n    return bnm_scheduler\n    \ndef set_random_seed(seed, deterministic=False):\n    \"\"\"Set random seed.\n    Args:\n        seed (int): Seed to be used.\n        deterministic (bool): Whether to set the deterministic option for\n            CUDNN backend, i.e., set `torch.backends.cudnn.deterministic`\n            to True and `torch.backends.cudnn.benchmark` to False.\n            Default: False.\n\n    # Speed-reproducibility tradeoff https://pytorch.org/docs/stable/notes/randomness.html\n    if cuda_deterministic:  # slower, more reproducible\n        cudnn.deterministic = True\n        cudnn.benchmark = False\n    else:  # faster, less reproducible\n        cudnn.deterministic = False\n        cudnn.benchmark = True\n\n    \"\"\"\n    random.seed(seed)","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.set_random_seed","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.set_random_seed#L80-L104","kind":"function","name":"set_random_seed","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":80,"end_line":104,"context_start_line":60,"context_end_line":124,"code":"\ndef worker_init_fn(worker_id):\n    np.random.seed(np.random.get_state()[1][0] + worker_id)\n\ndef build_lambda_sche(opti, config):\n    if config.get('decay_step') is not None:\n        lr_lbmd = lambda e: max(config.lr_decay ** (e / config.decay_step), config.lowest_decay)\n        scheduler = torch.optim.lr_scheduler.LambdaLR(opti, lr_lbmd)\n    else:\n        raise NotImplementedError()\n    return scheduler\n\ndef build_lambda_bnsche(model, config):\n    if config.get('decay_step') is not None:\n        bnm_lmbd = lambda e: max(config.bn_momentum * config.bn_decay ** (e / config.decay_step), config.lowest_decay)\n        bnm_scheduler = BNMomentumScheduler(model, bnm_lmbd)\n    else:\n        raise NotImplementedError()\n    return bnm_scheduler\n    \ndef set_random_seed(seed, deterministic=False):\n    \"\"\"Set random seed.\n    Args:\n        seed (int): Seed to be used.\n        deterministic (bool): Whether to set the deterministic option for\n            CUDNN backend, i.e., set `torch.backends.cudnn.deterministic`\n            to True and `torch.backends.cudnn.benchmark` to False.\n            Default: False.\n\n    # Speed-reproducibility tradeoff https://pytorch.org/docs/stable/notes/randomness.html\n    if cuda_deterministic:  # slower, more reproducible\n        cudnn.deterministic = True\n        cudnn.benchmark = False\n    else:  # faster, less reproducible\n        cudnn.deterministic = False\n        cudnn.benchmark = True\n\n    \"\"\"\n    random.seed(seed)\n    np.random.seed(seed)\n    torch.manual_seed(seed)\n    torch.cuda.manual_seed_all(seed)\n    if deterministic:\n        torch.backends.cudnn.deterministic = True\n        torch.backends.cudnn.benchmark = False\n\n\ndef is_seq_of(seq, expected_type, seq_type=None):\n    \"\"\"Check whether it is a sequence of some type.\n    Args:\n        seq (Sequence): The sequence to be checked.\n        expected_type (type): Expected type of sequence items.\n        seq_type (type, optional): Expected sequence type.\n    Returns:\n        bool: Whether the sequence is valid.\n    \"\"\"\n    if seq_type is None:\n        exp_seq_type = abc.Sequence\n    else:\n        assert isinstance(seq_type, type)\n        exp_seq_type = seq_type\n    if not isinstance(seq, exp_seq_type):\n        return False\n    for item in seq:\n        if not isinstance(item, expected_type):","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.is_seq_of","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.is_seq_of#L107-L126","kind":"function","name":"is_seq_of","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":107,"end_line":126,"context_start_line":87,"context_end_line":146,"code":"            Default: False.\n\n    # Speed-reproducibility tradeoff https://pytorch.org/docs/stable/notes/randomness.html\n    if cuda_deterministic:  # slower, more reproducible\n        cudnn.deterministic = True\n        cudnn.benchmark = False\n    else:  # faster, less reproducible\n        cudnn.deterministic = False\n        cudnn.benchmark = True\n\n    \"\"\"\n    random.seed(seed)\n    np.random.seed(seed)\n    torch.manual_seed(seed)\n    torch.cuda.manual_seed_all(seed)\n    if deterministic:\n        torch.backends.cudnn.deterministic = True\n        torch.backends.cudnn.benchmark = False\n\n\ndef is_seq_of(seq, expected_type, seq_type=None):\n    \"\"\"Check whether it is a sequence of some type.\n    Args:\n        seq (Sequence): The sequence to be checked.\n        expected_type (type): Expected type of sequence items.\n        seq_type (type, optional): Expected sequence type.\n    Returns:\n        bool: Whether the sequence is valid.\n    \"\"\"\n    if seq_type is None:\n        exp_seq_type = abc.Sequence\n    else:\n        assert isinstance(seq_type, type)\n        exp_seq_type = seq_type\n    if not isinstance(seq, exp_seq_type):\n        return False\n    for item in seq:\n        if not isinstance(item, expected_type):\n            return False\n    return True\n\n\ndef set_bn_momentum_default(bn_momentum):\n    def fn(m):\n        if isinstance(m, (nn.BatchNorm1d, nn.BatchNorm2d, nn.BatchNorm3d)):\n            m.momentum = bn_momentum\n    return fn\n\nclass BNMomentumScheduler(object):\n\n    def __init__(\n            self, model, bn_lambda, last_epoch=-1,\n            setter=set_bn_momentum_default\n    ):\n        if not isinstance(model, nn.Module):\n            raise RuntimeError(\n                \"Class '{}' is not a PyTorch nn Module\".format(\n                    type(model).__name__\n                )\n            )","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.set_bn_momentum_default","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.set_bn_momentum_default#L129-L133","kind":"function","name":"set_bn_momentum_default","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":129,"end_line":133,"context_start_line":109,"context_end_line":153,"code":"    Args:\n        seq (Sequence): The sequence to be checked.\n        expected_type (type): Expected type of sequence items.\n        seq_type (type, optional): Expected sequence type.\n    Returns:\n        bool: Whether the sequence is valid.\n    \"\"\"\n    if seq_type is None:\n        exp_seq_type = abc.Sequence\n    else:\n        assert isinstance(seq_type, type)\n        exp_seq_type = seq_type\n    if not isinstance(seq, exp_seq_type):\n        return False\n    for item in seq:\n        if not isinstance(item, expected_type):\n            return False\n    return True\n\n\ndef set_bn_momentum_default(bn_momentum):\n    def fn(m):\n        if isinstance(m, (nn.BatchNorm1d, nn.BatchNorm2d, nn.BatchNorm3d)):\n            m.momentum = bn_momentum\n    return fn\n\nclass BNMomentumScheduler(object):\n\n    def __init__(\n            self, model, bn_lambda, last_epoch=-1,\n            setter=set_bn_momentum_default\n    ):\n        if not isinstance(model, nn.Module):\n            raise RuntimeError(\n                \"Class '{}' is not a PyTorch nn Module\".format(\n                    type(model).__name__\n                )\n            )\n\n        self.model = model\n        self.setter = setter\n        self.lmbd = bn_lambda\n\n        self.step(last_epoch + 1)\n        self.last_epoch = last_epoch","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.BNMomentumScheduler","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.misc.BNMomentumScheduler#L135-L165","kind":"class","name":"BNMomentumScheduler","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":135,"end_line":165,"context_start_line":115,"context_end_line":185,"code":"    \"\"\"\n    if seq_type is None:\n        exp_seq_type = abc.Sequence\n    else:\n        assert isinstance(seq_type, type)\n        exp_seq_type = seq_type\n    if not isinstance(seq, exp_seq_type):\n        return False\n    for item in seq:\n        if not isinstance(item, expected_type):\n            return False\n    return True\n\n\ndef set_bn_momentum_default(bn_momentum):\n    def fn(m):\n        if isinstance(m, (nn.BatchNorm1d, nn.BatchNorm2d, nn.BatchNorm3d)):\n            m.momentum = bn_momentum\n    return fn\n\nclass BNMomentumScheduler(object):\n\n    def __init__(\n            self, model, bn_lambda, last_epoch=-1,\n            setter=set_bn_momentum_default\n    ):\n        if not isinstance(model, nn.Module):\n            raise RuntimeError(\n                \"Class '{}' is not a PyTorch nn Module\".format(\n                    type(model).__name__\n                )\n            )\n\n        self.model = model\n        self.setter = setter\n        self.lmbd = bn_lambda\n\n        self.step(last_epoch + 1)\n        self.last_epoch = last_epoch\n\n    def step(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n\n        self.last_epoch = epoch\n        self.model.apply(self.setter(self.lmbd(epoch)))\n\n    def get_momentum(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n        return self.lmbd(epoch)\n\n\n\ndef seprate_point_cloud(xyz, num_points, crop, fixed_points = None, padding_zeros = False):\n    '''\n     seprate point cloud: usage : using to generate the incomplete point cloud with a setted number.\n    '''\n    _,n,c = xyz.shape\n\n    assert n == num_points\n    assert c == 3\n    if crop == num_points:\n        return xyz, None\n        \n    INPUT = []\n    CROP = []\n    for points in xyz:\n        if isinstance(crop,list):\n            num_crop = random.randint(crop[0],crop[1])\n        else:","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.seprate_point_cloud","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.seprate_point_cloud#L169-L222","kind":"function","name":"seprate_point_cloud","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":169,"end_line":222,"context_start_line":149,"context_end_line":242,"code":"        self.setter = setter\n        self.lmbd = bn_lambda\n\n        self.step(last_epoch + 1)\n        self.last_epoch = last_epoch\n\n    def step(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n\n        self.last_epoch = epoch\n        self.model.apply(self.setter(self.lmbd(epoch)))\n\n    def get_momentum(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n        return self.lmbd(epoch)\n\n\n\ndef seprate_point_cloud(xyz, num_points, crop, fixed_points = None, padding_zeros = False):\n    '''\n     seprate point cloud: usage : using to generate the incomplete point cloud with a setted number.\n    '''\n    _,n,c = xyz.shape\n\n    assert n == num_points\n    assert c == 3\n    if crop == num_points:\n        return xyz, None\n        \n    INPUT = []\n    CROP = []\n    for points in xyz:\n        if isinstance(crop,list):\n            num_crop = random.randint(crop[0],crop[1])\n        else:\n            num_crop = crop\n\n        points = points.unsqueeze(0)\n\n        if fixed_points is None:       \n            center = F.normalize(torch.randn(1,1,3),p=2,dim=-1).cuda()\n        else:\n            if isinstance(fixed_points,list):\n                fixed_point = random.sample(fixed_points,1)[0]\n            else:\n                fixed_point = fixed_points\n            center = fixed_point.reshape(1,1,3).cuda()\n\n        distance_matrix = torch.norm(center.unsqueeze(2) - points.unsqueeze(1), p =2 ,dim = -1)  # 1 1 2048\n\n        idx = torch.argsort(distance_matrix,dim=-1, descending=False)[0,0] # 2048\n\n        if padding_zeros:\n            input_data = points.clone()\n            input_data[0, idx[:num_crop]] =  input_data[0,idx[:num_crop]] * 0\n\n        else:\n            input_data = points.clone()[0, idx[num_crop:]].unsqueeze(0) # 1 N 3\n\n        crop_data =  points.clone()[0, idx[:num_crop]].unsqueeze(0)\n\n        if isinstance(crop,list):\n            INPUT.append(fps(input_data,2048))\n            CROP.append(fps(crop_data,2048))\n        else:\n            INPUT.append(input_data)\n            CROP.append(crop_data)\n\n    input_data = torch.cat(INPUT,dim=0)# B N 3\n    crop_data = torch.cat(CROP,dim=0)# B M 3\n\n    return input_data.contiguous(), crop_data.contiguous()\n\ndef get_ptcloud_img(ptcloud):\n    fig = plt.figure(figsize=(8, 8))\n\n    x, z, y = ptcloud.transpose(1, 0)\n    ax = fig.gca(projection=Axes3D.name, adjustable='box')\n    ax.axis('off')\n    # ax.axis('scaled')\n    ax.view_init(30, 45)\n    max, min = np.max(ptcloud), np.min(ptcloud)\n    ax.set_xbound(min, max)\n    ax.set_ybound(min, max)\n    ax.set_zbound(min, max)\n    ax.scatter(x, y, z, zdir='z', c=x, cmap='jet')\n\n    fig.canvas.draw()\n    img = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')\n    img = img.reshape(fig.canvas.get_width_height()[::-1] + (3, ))\n    return img\n","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.get_ptcloud_img","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.get_ptcloud_img#L224-L241","kind":"function","name":"get_ptcloud_img","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":224,"end_line":241,"context_start_line":204,"context_end_line":261,"code":"            input_data = points.clone()\n            input_data[0, idx[:num_crop]] =  input_data[0,idx[:num_crop]] * 0\n\n        else:\n            input_data = points.clone()[0, idx[num_crop:]].unsqueeze(0) # 1 N 3\n\n        crop_data =  points.clone()[0, idx[:num_crop]].unsqueeze(0)\n\n        if isinstance(crop,list):\n            INPUT.append(fps(input_data,2048))\n            CROP.append(fps(crop_data,2048))\n        else:\n            INPUT.append(input_data)\n            CROP.append(crop_data)\n\n    input_data = torch.cat(INPUT,dim=0)# B N 3\n    crop_data = torch.cat(CROP,dim=0)# B M 3\n\n    return input_data.contiguous(), crop_data.contiguous()\n\ndef get_ptcloud_img(ptcloud):\n    fig = plt.figure(figsize=(8, 8))\n\n    x, z, y = ptcloud.transpose(1, 0)\n    ax = fig.gca(projection=Axes3D.name, adjustable='box')\n    ax.axis('off')\n    # ax.axis('scaled')\n    ax.view_init(30, 45)\n    max, min = np.max(ptcloud), np.min(ptcloud)\n    ax.set_xbound(min, max)\n    ax.set_ybound(min, max)\n    ax.set_zbound(min, max)\n    ax.scatter(x, y, z, zdir='z', c=x, cmap='jet')\n\n    fig.canvas.draw()\n    img = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')\n    img = img.reshape(fig.canvas.get_width_height()[::-1] + (3, ))\n    return img\n\n\n\ndef visualize_KITTI(path, data_list, titles = ['input','pred'], cmap=['bwr','autumn'], zdir='y', \n                         xlim=(-1, 1), ylim=(-1, 1), zlim=(-1, 1) ):\n    fig = plt.figure(figsize=(6*len(data_list),6))\n    cmax = data_list[-1][:,0].max()\n\n    for i in range(len(data_list)):\n        data = data_list[i][:-2048] if i == 1 else data_list[i]\n        color = data[:,0] /cmax\n        ax = fig.add_subplot(1, len(data_list) , i + 1, projection='3d')\n        ax.view_init(30, -120)\n        b = ax.scatter(data[:, 0], data[:, 1], data[:, 2], zdir=zdir, c=color,vmin=-1,vmax=1 ,cmap = cmap[0],s=4,linewidth=0.05, edgecolors = 'black')\n        ax.set_title(titles[i])\n\n        ax.set_axis_off()\n        ax.set_xlim(xlim)\n        ax.set_ylim(ylim)\n        ax.set_zlim(zlim)","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.visualize_KITTI","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.visualize_KITTI#L245-L271","kind":"function","name":"visualize_KITTI","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":245,"end_line":271,"context_start_line":225,"context_end_line":286,"code":"    fig = plt.figure(figsize=(8, 8))\n\n    x, z, y = ptcloud.transpose(1, 0)\n    ax = fig.gca(projection=Axes3D.name, adjustable='box')\n    ax.axis('off')\n    # ax.axis('scaled')\n    ax.view_init(30, 45)\n    max, min = np.max(ptcloud), np.min(ptcloud)\n    ax.set_xbound(min, max)\n    ax.set_ybound(min, max)\n    ax.set_zbound(min, max)\n    ax.scatter(x, y, z, zdir='z', c=x, cmap='jet')\n\n    fig.canvas.draw()\n    img = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')\n    img = img.reshape(fig.canvas.get_width_height()[::-1] + (3, ))\n    return img\n\n\n\ndef visualize_KITTI(path, data_list, titles = ['input','pred'], cmap=['bwr','autumn'], zdir='y', \n                         xlim=(-1, 1), ylim=(-1, 1), zlim=(-1, 1) ):\n    fig = plt.figure(figsize=(6*len(data_list),6))\n    cmax = data_list[-1][:,0].max()\n\n    for i in range(len(data_list)):\n        data = data_list[i][:-2048] if i == 1 else data_list[i]\n        color = data[:,0] /cmax\n        ax = fig.add_subplot(1, len(data_list) , i + 1, projection='3d')\n        ax.view_init(30, -120)\n        b = ax.scatter(data[:, 0], data[:, 1], data[:, 2], zdir=zdir, c=color,vmin=-1,vmax=1 ,cmap = cmap[0],s=4,linewidth=0.05, edgecolors = 'black')\n        ax.set_title(titles[i])\n\n        ax.set_axis_off()\n        ax.set_xlim(xlim)\n        ax.set_ylim(ylim)\n        ax.set_zlim(zlim)\n    plt.subplots_adjust(left=0, right=1, bottom=0, top=1, wspace=0.2, hspace=0)\n    if not os.path.exists(path):\n        os.makedirs(path)\n\n    pic_path = path + '.png'\n    fig.savefig(pic_path)\n\n    np.save(os.path.join(path, 'input.npy'), data_list[0].numpy())\n    np.save(os.path.join(path, 'pred.npy'), data_list[1].numpy())\n    plt.close(fig)\n\n\ndef random_dropping(pc, e):\n    up_num = max(64, 768 // (e//50 + 1))\n    pc = pc\n    random_num = torch.randint(1, up_num, (1,1))[0,0]\n    pc = fps(pc, random_num)\n    padding = torch.zeros(pc.size(0), 2048 - pc.size(1), 3).to(pc.device)\n    pc = torch.cat([pc, padding], dim = 1)\n    return pc\n    \n\ndef random_scale(partial, scale_range=[0.8, 1.2]):\n    scale = torch.rand(1).cuda() * (scale_range[1] - scale_range[0]) + scale_range[0]\n    return partial * scale","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.random_dropping","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.random_dropping#L274-L281","kind":"function","name":"random_dropping","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":274,"end_line":281,"context_start_line":254,"context_end_line":286,"code":"        ax.view_init(30, -120)\n        b = ax.scatter(data[:, 0], data[:, 1], data[:, 2], zdir=zdir, c=color,vmin=-1,vmax=1 ,cmap = cmap[0],s=4,linewidth=0.05, edgecolors = 'black')\n        ax.set_title(titles[i])\n\n        ax.set_axis_off()\n        ax.set_xlim(xlim)\n        ax.set_ylim(ylim)\n        ax.set_zlim(zlim)\n    plt.subplots_adjust(left=0, right=1, bottom=0, top=1, wspace=0.2, hspace=0)\n    if not os.path.exists(path):\n        os.makedirs(path)\n\n    pic_path = path + '.png'\n    fig.savefig(pic_path)\n\n    np.save(os.path.join(path, 'input.npy'), data_list[0].numpy())\n    np.save(os.path.join(path, 'pred.npy'), data_list[1].numpy())\n    plt.close(fig)\n\n\ndef random_dropping(pc, e):\n    up_num = max(64, 768 // (e//50 + 1))\n    pc = pc\n    random_num = torch.randint(1, up_num, (1,1))[0,0]\n    pc = fps(pc, random_num)\n    padding = torch.zeros(pc.size(0), 2048 - pc.size(1), 3).to(pc.device)\n    pc = torch.cat([pc, padding], dim = 1)\n    return pc\n    \n\ndef random_scale(partial, scale_range=[0.8, 1.2]):\n    scale = torch.rand(1).cuda() * (scale_range[1] - scale_range[0]) + scale_range[0]\n    return partial * scale","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.random_scale","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.random_scale#L284-L286","kind":"function","name":"random_scale","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":284,"end_line":286,"context_start_line":264,"context_end_line":286,"code":"        os.makedirs(path)\n\n    pic_path = path + '.png'\n    fig.savefig(pic_path)\n\n    np.save(os.path.join(path, 'input.npy'), data_list[0].numpy())\n    np.save(os.path.join(path, 'pred.npy'), data_list[1].numpy())\n    plt.close(fig)\n\n\ndef random_dropping(pc, e):\n    up_num = max(64, 768 // (e//50 + 1))\n    pc = pc\n    random_num = torch.randint(1, up_num, (1,1))[0,0]\n    pc = fps(pc, random_num)\n    padding = torch.zeros(pc.size(0), 2048 - pc.size(1), 3).to(pc.device)\n    pc = torch.cat([pc, padding], dim = 1)\n    return pc\n    \n\ndef random_scale(partial, scale_range=[0.8, 1.2]):\n    scale = torch.rand(1).cuda() * (scale_range[1] - scale_range[0]) + scale_range[0]\n    return partial * scale","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.fn","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.fn#L130-L132","kind":"function","name":"fn","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":130,"end_line":132,"context_start_line":110,"context_end_line":152,"code":"        seq (Sequence): The sequence to be checked.\n        expected_type (type): Expected type of sequence items.\n        seq_type (type, optional): Expected sequence type.\n    Returns:\n        bool: Whether the sequence is valid.\n    \"\"\"\n    if seq_type is None:\n        exp_seq_type = abc.Sequence\n    else:\n        assert isinstance(seq_type, type)\n        exp_seq_type = seq_type\n    if not isinstance(seq, exp_seq_type):\n        return False\n    for item in seq:\n        if not isinstance(item, expected_type):\n            return False\n    return True\n\n\ndef set_bn_momentum_default(bn_momentum):\n    def fn(m):\n        if isinstance(m, (nn.BatchNorm1d, nn.BatchNorm2d, nn.BatchNorm3d)):\n            m.momentum = bn_momentum\n    return fn\n\nclass BNMomentumScheduler(object):\n\n    def __init__(\n            self, model, bn_lambda, last_epoch=-1,\n            setter=set_bn_momentum_default\n    ):\n        if not isinstance(model, nn.Module):\n            raise RuntimeError(\n                \"Class '{}' is not a PyTorch nn Module\".format(\n                    type(model).__name__\n                )\n            )\n\n        self.model = model\n        self.setter = setter\n        self.lmbd = bn_lambda\n\n        self.step(last_epoch + 1)","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.__init__#L137-L153","kind":"function","name":"__init__","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":137,"end_line":153,"context_start_line":117,"context_end_line":173,"code":"        exp_seq_type = abc.Sequence\n    else:\n        assert isinstance(seq_type, type)\n        exp_seq_type = seq_type\n    if not isinstance(seq, exp_seq_type):\n        return False\n    for item in seq:\n        if not isinstance(item, expected_type):\n            return False\n    return True\n\n\ndef set_bn_momentum_default(bn_momentum):\n    def fn(m):\n        if isinstance(m, (nn.BatchNorm1d, nn.BatchNorm2d, nn.BatchNorm3d)):\n            m.momentum = bn_momentum\n    return fn\n\nclass BNMomentumScheduler(object):\n\n    def __init__(\n            self, model, bn_lambda, last_epoch=-1,\n            setter=set_bn_momentum_default\n    ):\n        if not isinstance(model, nn.Module):\n            raise RuntimeError(\n                \"Class '{}' is not a PyTorch nn Module\".format(\n                    type(model).__name__\n                )\n            )\n\n        self.model = model\n        self.setter = setter\n        self.lmbd = bn_lambda\n\n        self.step(last_epoch + 1)\n        self.last_epoch = last_epoch\n\n    def step(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n\n        self.last_epoch = epoch\n        self.model.apply(self.setter(self.lmbd(epoch)))\n\n    def get_momentum(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n        return self.lmbd(epoch)\n\n\n\ndef seprate_point_cloud(xyz, num_points, crop, fixed_points = None, padding_zeros = False):\n    '''\n     seprate point cloud: usage : using to generate the incomplete point cloud with a setted number.\n    '''\n    _,n,c = xyz.shape","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.step","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.step#L155-L160","kind":"function","name":"step","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":155,"end_line":160,"context_start_line":135,"context_end_line":180,"code":"class BNMomentumScheduler(object):\n\n    def __init__(\n            self, model, bn_lambda, last_epoch=-1,\n            setter=set_bn_momentum_default\n    ):\n        if not isinstance(model, nn.Module):\n            raise RuntimeError(\n                \"Class '{}' is not a PyTorch nn Module\".format(\n                    type(model).__name__\n                )\n            )\n\n        self.model = model\n        self.setter = setter\n        self.lmbd = bn_lambda\n\n        self.step(last_epoch + 1)\n        self.last_epoch = last_epoch\n\n    def step(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n\n        self.last_epoch = epoch\n        self.model.apply(self.setter(self.lmbd(epoch)))\n\n    def get_momentum(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n        return self.lmbd(epoch)\n\n\n\ndef seprate_point_cloud(xyz, num_points, crop, fixed_points = None, padding_zeros = False):\n    '''\n     seprate point cloud: usage : using to generate the incomplete point cloud with a setted number.\n    '''\n    _,n,c = xyz.shape\n\n    assert n == num_points\n    assert c == 3\n    if crop == num_points:\n        return xyz, None\n        \n    INPUT = []","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.misc.get_momentum","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.misc.get_momentum#L162-L165","kind":"function","name":"get_momentum","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":162,"end_line":165,"context_start_line":142,"context_end_line":185,"code":"            raise RuntimeError(\n                \"Class '{}' is not a PyTorch nn Module\".format(\n                    type(model).__name__\n                )\n            )\n\n        self.model = model\n        self.setter = setter\n        self.lmbd = bn_lambda\n\n        self.step(last_epoch + 1)\n        self.last_epoch = last_epoch\n\n    def step(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n\n        self.last_epoch = epoch\n        self.model.apply(self.setter(self.lmbd(epoch)))\n\n    def get_momentum(self, epoch=None):\n        if epoch is None:\n            epoch = self.last_epoch + 1\n        return self.lmbd(epoch)\n\n\n\ndef seprate_point_cloud(xyz, num_points, crop, fixed_points = None, padding_zeros = False):\n    '''\n     seprate point cloud: usage : using to generate the incomplete point cloud with a setted number.\n    '''\n    _,n,c = xyz.shape\n\n    assert n == num_points\n    assert c == 3\n    if crop == num_points:\n        return xyz, None\n        \n    INPUT = []\n    CROP = []\n    for points in xyz:\n        if isinstance(crop,list):\n            num_crop = random.randint(crop[0],crop[1])\n        else:","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.checkpoint","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.models.pointbert.checkpoint#L1-L126","kind":"module","name":"imagebind_LLM.ImageBind.models.pointbert.checkpoint","path":"imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","language":"python","start_line":1,"end_line":126,"context_start_line":1,"context_end_line":126,"code":"from collections import defaultdict\nimport torch.nn as nn\n\nfrom typing import Any\nfrom typing import Optional, List, Dict, NamedTuple, Tuple, Iterable\n\nfrom termcolor import colored\n\ndef get_missing_parameters_message(keys: List[str]) -> str:\n    \"\"\"\n    Get a logging-friendly message to report parameter names (keys) that are in\n    the model but not found in a checkpoint.\n    Args:\n        keys (list[str]): List of keys that were not found in the checkpoint.\n    Returns:\n        str: message.\n    \"\"\"\n    groups = _group_checkpoint_keys(keys)\n    msg = \"Some model parameters or buffers are not found in the checkpoint:\\n\"\n    msg += \"\\n\".join(\n        \"  \" + colored(k + _group_to_str(v), \"blue\") for k, v in groups.items()\n    )\n    return msg\n\n\ndef get_unexpected_parameters_message(keys: List[str]) -> str:\n    \"\"\"\n    Get a logging-friendly message to report parameter names (keys) that are in\n    the checkpoint but not found in the model.\n    Args:\n        keys (list[str]): List of keys that were not found in the model.\n    Returns:\n        str: message.\n    \"\"\"\n    groups = _group_checkpoint_keys(keys)\n    msg = \"The checkpoint state_dict contains keys that are not used by the model:\\n\"\n    msg += \"\\n\".join(\n        \"  \" + colored(k + _group_to_str(v), \"magenta\") for k, v in groups.items()\n    )\n    return msg\n\n\ndef _strip_prefix_if_present(state_dict: Dict[str, Any], prefix: str) -> None:\n    \"\"\"\n    Strip the prefix in metadata, if any.\n    Args:\n        state_dict (OrderedDict): a state-dict to be loaded to the model.\n        prefix (str): prefix.\n    \"\"\"\n    keys = sorted(state_dict.keys())\n    if not all(len(key) == 0 or key.startswith(prefix) for key in keys):\n        return\n\n    for key in keys:\n        newkey = key[len(prefix):]\n        state_dict[newkey] = state_dict.pop(key)\n\n    # also strip the prefix in metadata, if any..\n    try:\n        metadata = state_dict._metadata  # pyre-ignore\n    except AttributeError:\n        pass\n    else:\n        for key in list(metadata.keys()):\n            # for the metadata dict, the key can be:\n            # '': for the DDP module, which we want to remove.\n            # 'module': for the actual model.\n            # 'module.xx.xx': for the rest.\n\n            if len(key) == 0:\n                continue\n            newkey = key[len(prefix):]\n            metadata[newkey] = metadata.pop(key)\n\n\ndef _group_checkpoint_keys(keys: List[str]) -> Dict[str, List[str]]:\n    \"\"\"\n    Group keys based on common prefixes. A prefix is the string up to the final\n    \".\" in each key.\n    Args:\n        keys (list[str]): list of parameter names, i.e. keys in the model\n            checkpoint dict.\n    Returns:\n        dict[list]: keys with common prefixes are grouped into lists.\n    \"\"\"\n    groups = defaultdict(list)\n    for key in keys:\n        pos = key.rfind(\".\")\n        if pos >= 0:\n            head, tail = key[:pos], [key[pos + 1:]]\n        else:\n            head, tail = key, []\n        groups[head].extend(tail)\n    return groups\n\n\ndef _group_to_str(group: List[str]) -> str:\n    \"\"\"\n    Format a group of parameter name suffixes into a loggable string.\n    Args:\n        group (list[str]): list of parameter name suffixes.\n    Returns:\n        str: formated string.\n    \"\"\"\n    if len(group) == 0:\n        return \"\"\n\n    if len(group) == 1:\n        return \".\" + group[0]\n\n    return \".{\" + \", \".join(group) + \"}\"\n\n\ndef _named_modules_with_dup(\n        model: nn.Module, prefix: str = \"\"\n) -> Iterable[Tuple[str, nn.Module]]:\n    \"\"\"\n    The same as `model.named_modules()`, except that it includes\n    duplicated modules that have more than one name.\n    \"\"\"\n    yield prefix, model\n    for name, module in model._modules.items():  # pyre-ignore\n        if module is None:\n            continue\n        submodule_prefix = prefix + (\".\" if prefix else \"\") + name\n        yield from _named_modules_with_dup(module, submodule_prefix)","source_hash":"bd95d3c00499acc67b69555c7d9022731b3f53633dda6789380f32e41a084946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.checkpoint.get_missing_parameters_message","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.checkpoint.get_missing_parameters_message#L9-L23","kind":"function","name":"get_missing_parameters_message","path":"imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","language":"python","start_line":9,"end_line":23,"context_start_line":1,"context_end_line":43,"code":"from collections import defaultdict\nimport torch.nn as nn\n\nfrom typing import Any\nfrom typing import Optional, List, Dict, NamedTuple, Tuple, Iterable\n\nfrom termcolor import colored\n\ndef get_missing_parameters_message(keys: List[str]) -> str:\n    \"\"\"\n    Get a logging-friendly message to report parameter names (keys) that are in\n    the model but not found in a checkpoint.\n    Args:\n        keys (list[str]): List of keys that were not found in the checkpoint.\n    Returns:\n        str: message.\n    \"\"\"\n    groups = _group_checkpoint_keys(keys)\n    msg = \"Some model parameters or buffers are not found in the checkpoint:\\n\"\n    msg += \"\\n\".join(\n        \"  \" + colored(k + _group_to_str(v), \"blue\") for k, v in groups.items()\n    )\n    return msg\n\n\ndef get_unexpected_parameters_message(keys: List[str]) -> str:\n    \"\"\"\n    Get a logging-friendly message to report parameter names (keys) that are in\n    the checkpoint but not found in the model.\n    Args:\n        keys (list[str]): List of keys that were not found in the model.\n    Returns:\n        str: message.\n    \"\"\"\n    groups = _group_checkpoint_keys(keys)\n    msg = \"The checkpoint state_dict contains keys that are not used by the model:\\n\"\n    msg += \"\\n\".join(\n        \"  \" + colored(k + _group_to_str(v), \"magenta\") for k, v in groups.items()\n    )\n    return msg\n\n\ndef _strip_prefix_if_present(state_dict: Dict[str, Any], prefix: str) -> None:","source_hash":"bd95d3c00499acc67b69555c7d9022731b3f53633dda6789380f32e41a084946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.checkpoint.get_unexpected_parameters_message","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.checkpoint.get_unexpected_parameters_message#L26-L40","kind":"function","name":"get_unexpected_parameters_message","path":"imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","language":"python","start_line":26,"end_line":40,"context_start_line":6,"context_end_line":60,"code":"\nfrom termcolor import colored\n\ndef get_missing_parameters_message(keys: List[str]) -> str:\n    \"\"\"\n    Get a logging-friendly message to report parameter names (keys) that are in\n    the model but not found in a checkpoint.\n    Args:\n        keys (list[str]): List of keys that were not found in the checkpoint.\n    Returns:\n        str: message.\n    \"\"\"\n    groups = _group_checkpoint_keys(keys)\n    msg = \"Some model parameters or buffers are not found in the checkpoint:\\n\"\n    msg += \"\\n\".join(\n        \"  \" + colored(k + _group_to_str(v), \"blue\") for k, v in groups.items()\n    )\n    return msg\n\n\ndef get_unexpected_parameters_message(keys: List[str]) -> str:\n    \"\"\"\n    Get a logging-friendly message to report parameter names (keys) that are in\n    the checkpoint but not found in the model.\n    Args:\n        keys (list[str]): List of keys that were not found in the model.\n    Returns:\n        str: message.\n    \"\"\"\n    groups = _group_checkpoint_keys(keys)\n    msg = \"The checkpoint state_dict contains keys that are not used by the model:\\n\"\n    msg += \"\\n\".join(\n        \"  \" + colored(k + _group_to_str(v), \"magenta\") for k, v in groups.items()\n    )\n    return msg\n\n\ndef _strip_prefix_if_present(state_dict: Dict[str, Any], prefix: str) -> None:\n    \"\"\"\n    Strip the prefix in metadata, if any.\n    Args:\n        state_dict (OrderedDict): a state-dict to be loaded to the model.\n        prefix (str): prefix.\n    \"\"\"\n    keys = sorted(state_dict.keys())\n    if not all(len(key) == 0 or key.startswith(prefix) for key in keys):\n        return\n\n    for key in keys:\n        newkey = key[len(prefix):]\n        state_dict[newkey] = state_dict.pop(key)\n\n    # also strip the prefix in metadata, if any..\n    try:\n        metadata = state_dict._metadata  # pyre-ignore","source_hash":"bd95d3c00499acc67b69555c7d9022731b3f53633dda6789380f32e41a084946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.checkpoint._strip_prefix_if_present","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.checkpoint._strip_prefix_if_present#L43-L73","kind":"function","name":"_strip_prefix_if_present","path":"imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","language":"python","start_line":43,"end_line":73,"context_start_line":23,"context_end_line":93,"code":"    return msg\n\n\ndef get_unexpected_parameters_message(keys: List[str]) -> str:\n    \"\"\"\n    Get a logging-friendly message to report parameter names (keys) that are in\n    the checkpoint but not found in the model.\n    Args:\n        keys (list[str]): List of keys that were not found in the model.\n    Returns:\n        str: message.\n    \"\"\"\n    groups = _group_checkpoint_keys(keys)\n    msg = \"The checkpoint state_dict contains keys that are not used by the model:\\n\"\n    msg += \"\\n\".join(\n        \"  \" + colored(k + _group_to_str(v), \"magenta\") for k, v in groups.items()\n    )\n    return msg\n\n\ndef _strip_prefix_if_present(state_dict: Dict[str, Any], prefix: str) -> None:\n    \"\"\"\n    Strip the prefix in metadata, if any.\n    Args:\n        state_dict (OrderedDict): a state-dict to be loaded to the model.\n        prefix (str): prefix.\n    \"\"\"\n    keys = sorted(state_dict.keys())\n    if not all(len(key) == 0 or key.startswith(prefix) for key in keys):\n        return\n\n    for key in keys:\n        newkey = key[len(prefix):]\n        state_dict[newkey] = state_dict.pop(key)\n\n    # also strip the prefix in metadata, if any..\n    try:\n        metadata = state_dict._metadata  # pyre-ignore\n    except AttributeError:\n        pass\n    else:\n        for key in list(metadata.keys()):\n            # for the metadata dict, the key can be:\n            # '': for the DDP module, which we want to remove.\n            # 'module': for the actual model.\n            # 'module.xx.xx': for the rest.\n\n            if len(key) == 0:\n                continue\n            newkey = key[len(prefix):]\n            metadata[newkey] = metadata.pop(key)\n\n\ndef _group_checkpoint_keys(keys: List[str]) -> Dict[str, List[str]]:\n    \"\"\"\n    Group keys based on common prefixes. A prefix is the string up to the final\n    \".\" in each key.\n    Args:\n        keys (list[str]): list of parameter names, i.e. keys in the model\n            checkpoint dict.\n    Returns:\n        dict[list]: keys with common prefixes are grouped into lists.\n    \"\"\"\n    groups = defaultdict(list)\n    for key in keys:\n        pos = key.rfind(\".\")\n        if pos >= 0:\n            head, tail = key[:pos], [key[pos + 1:]]\n        else:\n            head, tail = key, []\n        groups[head].extend(tail)","source_hash":"bd95d3c00499acc67b69555c7d9022731b3f53633dda6789380f32e41a084946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.checkpoint._group_checkpoint_keys","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.checkpoint._group_checkpoint_keys#L76-L94","kind":"function","name":"_group_checkpoint_keys","path":"imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","language":"python","start_line":76,"end_line":94,"context_start_line":56,"context_end_line":114,"code":"        state_dict[newkey] = state_dict.pop(key)\n\n    # also strip the prefix in metadata, if any..\n    try:\n        metadata = state_dict._metadata  # pyre-ignore\n    except AttributeError:\n        pass\n    else:\n        for key in list(metadata.keys()):\n            # for the metadata dict, the key can be:\n            # '': for the DDP module, which we want to remove.\n            # 'module': for the actual model.\n            # 'module.xx.xx': for the rest.\n\n            if len(key) == 0:\n                continue\n            newkey = key[len(prefix):]\n            metadata[newkey] = metadata.pop(key)\n\n\ndef _group_checkpoint_keys(keys: List[str]) -> Dict[str, List[str]]:\n    \"\"\"\n    Group keys based on common prefixes. A prefix is the string up to the final\n    \".\" in each key.\n    Args:\n        keys (list[str]): list of parameter names, i.e. keys in the model\n            checkpoint dict.\n    Returns:\n        dict[list]: keys with common prefixes are grouped into lists.\n    \"\"\"\n    groups = defaultdict(list)\n    for key in keys:\n        pos = key.rfind(\".\")\n        if pos >= 0:\n            head, tail = key[:pos], [key[pos + 1:]]\n        else:\n            head, tail = key, []\n        groups[head].extend(tail)\n    return groups\n\n\ndef _group_to_str(group: List[str]) -> str:\n    \"\"\"\n    Format a group of parameter name suffixes into a loggable string.\n    Args:\n        group (list[str]): list of parameter name suffixes.\n    Returns:\n        str: formated string.\n    \"\"\"\n    if len(group) == 0:\n        return \"\"\n\n    if len(group) == 1:\n        return \".\" + group[0]\n\n    return \".{\" + \", \".join(group) + \"}\"\n\n\ndef _named_modules_with_dup(","source_hash":"bd95d3c00499acc67b69555c7d9022731b3f53633dda6789380f32e41a084946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.checkpoint._group_to_str","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.checkpoint._group_to_str#L97-L111","kind":"function","name":"_group_to_str","path":"imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","language":"python","start_line":97,"end_line":111,"context_start_line":77,"context_end_line":126,"code":"    \"\"\"\n    Group keys based on common prefixes. A prefix is the string up to the final\n    \".\" in each key.\n    Args:\n        keys (list[str]): list of parameter names, i.e. keys in the model\n            checkpoint dict.\n    Returns:\n        dict[list]: keys with common prefixes are grouped into lists.\n    \"\"\"\n    groups = defaultdict(list)\n    for key in keys:\n        pos = key.rfind(\".\")\n        if pos >= 0:\n            head, tail = key[:pos], [key[pos + 1:]]\n        else:\n            head, tail = key, []\n        groups[head].extend(tail)\n    return groups\n\n\ndef _group_to_str(group: List[str]) -> str:\n    \"\"\"\n    Format a group of parameter name suffixes into a loggable string.\n    Args:\n        group (list[str]): list of parameter name suffixes.\n    Returns:\n        str: formated string.\n    \"\"\"\n    if len(group) == 0:\n        return \"\"\n\n    if len(group) == 1:\n        return \".\" + group[0]\n\n    return \".{\" + \", \".join(group) + \"}\"\n\n\ndef _named_modules_with_dup(\n        model: nn.Module, prefix: str = \"\"\n) -> Iterable[Tuple[str, nn.Module]]:\n    \"\"\"\n    The same as `model.named_modules()`, except that it includes\n    duplicated modules that have more than one name.\n    \"\"\"\n    yield prefix, model\n    for name, module in model._modules.items():  # pyre-ignore\n        if module is None:\n            continue\n        submodule_prefix = prefix + (\".\" if prefix else \"\") + name\n        yield from _named_modules_with_dup(module, submodule_prefix)","source_hash":"bd95d3c00499acc67b69555c7d9022731b3f53633dda6789380f32e41a084946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.checkpoint._named_modules_with_dup","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.checkpoint._named_modules_with_dup#L114-L126","kind":"function","name":"_named_modules_with_dup","path":"imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","language":"python","start_line":114,"end_line":126,"context_start_line":94,"context_end_line":126,"code":"    return groups\n\n\ndef _group_to_str(group: List[str]) -> str:\n    \"\"\"\n    Format a group of parameter name suffixes into a loggable string.\n    Args:\n        group (list[str]): list of parameter name suffixes.\n    Returns:\n        str: formated string.\n    \"\"\"\n    if len(group) == 0:\n        return \"\"\n\n    if len(group) == 1:\n        return \".\" + group[0]\n\n    return \".{\" + \", \".join(group) + \"}\"\n\n\ndef _named_modules_with_dup(\n        model: nn.Module, prefix: str = \"\"\n) -> Iterable[Tuple[str, nn.Module]]:\n    \"\"\"\n    The same as `model.named_modules()`, except that it includes\n    duplicated modules that have more than one name.\n    \"\"\"\n    yield prefix, model\n    for name, module in model._modules.items():  # pyre-ignore\n        if module is None:\n            continue\n        submodule_prefix = prefix + (\".\" if prefix else \"\") + name\n        yield from _named_modules_with_dup(module, submodule_prefix)","source_hash":"bd95d3c00499acc67b69555c7d9022731b3f53633dda6789380f32e41a084946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.models.pointbert.point_encoder#L1-L251","kind":"module","name":"imagebind_LLM.ImageBind.models.pointbert.point_encoder","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":1,"end_line":251,"context_start_line":1,"context_end_line":251,"code":"import torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nfrom timm.models.layers import DropPath\nfrom .dvae import Group\nfrom .dvae import Encoder\nfrom .logger import print_log\nimport yaml\nfrom easydict import EasyDict\nfrom .checkpoint import get_missing_parameters_message, get_unexpected_parameters_message\n\ndef merge_new_config(config, new_config):\n    for key, val in new_config.items():\n        if not isinstance(val, dict):\n            if key == '_base_':\n                with open(new_config['_base_'], 'r') as f:\n                    try:\n                        val = yaml.load(f, Loader=yaml.FullLoader)\n                    except:\n                        val = yaml.load(f)\n                config[key] = EasyDict()\n                merge_new_config(config[key], val)\n            else:\n                config[key] = val\n                continue\n        if key not in config:\n            config[key] = EasyDict()\n        merge_new_config(config[key], val)\n    return config\n\ndef cfg_from_yaml_file(cfg_file):\n    config = EasyDict()\n    with open(cfg_file, 'r') as f:\n        new_config = yaml.load(f, Loader=yaml.FullLoader)\n    merge_new_config(config=config, new_config=new_config)\n    return config\n\nclass Mlp(nn.Module):\n    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):\n        super().__init__()\n        out_features = out_features or in_features\n        hidden_features = hidden_features or in_features\n        self.fc1 = nn.Linear(in_features, hidden_features)\n        self.act = act_layer()\n        self.fc2 = nn.Linear(hidden_features, out_features)\n        self.drop = nn.Dropout(drop)\n\n    def forward(self, x):\n        x = self.fc1(x)\n        x = self.act(x)\n        x = self.drop(x)\n        x = self.fc2(x)\n        x = self.drop(x)\n        return x\n\n\nclass Attention(nn.Module):\n    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0.):\n        super().__init__()\n        self.num_heads = num_heads\n        head_dim = dim // num_heads\n        # NOTE scale factor was wrong in my original version, can set manually to be compat with prev weights\n        self.scale = qk_scale or head_dim ** -0.5\n\n        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)\n        self.attn_drop = nn.Dropout(attn_drop)\n        self.proj = nn.Linear(dim, dim)\n        self.proj_drop = nn.Dropout(proj_drop)\n\n    def forward(self, x):\n        B, N, C = x.shape\n        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)\n        q, k, v = qkv[0], qkv[1], qkv[2]  # make torchscript happy (cannot use tensor as tuple)\n\n        attn = (q @ k.transpose(-2, -1)) * self.scale\n        attn = attn.softmax(dim=-1)\n        attn = self.attn_drop(attn)\n\n        x = (attn @ v).transpose(1, 2).reshape(B, N, C)\n        x = self.proj(x)\n        x = self.proj_drop(x)\n        return x\n\n\nclass Block(nn.Module):\n    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,\n                 drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm):\n        super().__init__()\n        self.norm1 = norm_layer(dim)\n\n        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here\n        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()\n        self.norm2 = norm_layer(dim)\n        mlp_hidden_dim = int(dim * mlp_ratio)\n        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)\n\n        self.attn = Attention(\n            dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop)\n\n    def forward(self, x):\n        x = x + self.drop_path(self.attn(self.norm1(x)))\n        x = x + self.drop_path(self.mlp(self.norm2(x)))\n        return x\n\n\nclass TransformerEncoder(nn.Module):\n    \"\"\" Transformer Encoder without hierarchical structure\n    \"\"\"\n\n    def __init__(self, embed_dim=768, depth=4, num_heads=12, mlp_ratio=4., qkv_bias=False, qk_scale=None,\n                 drop_rate=0., attn_drop_rate=0., drop_path_rate=0.):\n        super().__init__()\n\n        self.blocks = nn.ModuleList([\n            Block(\n                dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale,\n                drop=drop_rate, attn_drop=attn_drop_rate,\n                drop_path=drop_path_rate[i] if isinstance(drop_path_rate, list) else drop_path_rate\n            )\n            for i in range(depth)])\n\n    def forward(self, x, pos):\n        for _, block in enumerate(self.blocks):\n            x = block(x + pos)\n        return x\n\nclass PointTransformer(nn.Module):\n    def __init__(self, config):\n        super().__init__()\n        self.config = config\n\n        self.trans_dim = config.trans_dim\n        self.depth = config.depth\n        self.drop_path_rate = config.drop_path_rate\n        self.cls_dim = config.cls_dim\n        self.num_heads = config.num_heads\n\n        self.group_size = config.group_size\n        self.num_group = config.num_group\n        # grouper\n        self.group_divider = Group(num_group=self.num_group, group_size=self.group_size)\n        # define the encoder\n        self.encoder_dims = config.encoder_dims\n        self.encoder = Encoder(encoder_channel=self.encoder_dims)\n        # bridge encoder and transformer\n        self.reduce_dim = nn.Linear(self.encoder_dims, self.trans_dim)\n\n        self.cls_token = nn.Parameter(torch.zeros(1, 1, self.trans_dim))\n        self.cls_pos = nn.Parameter(torch.randn(1, 1, self.trans_dim))\n\n        self.pos_embed = nn.Sequential(\n            nn.Linear(3, 128),\n            nn.GELU(),\n            nn.Linear(128, self.trans_dim)\n        )\n\n        dpr = [x.item() for x in torch.linspace(0, self.drop_path_rate, self.depth)]\n        self.blocks = TransformerEncoder(\n            embed_dim=self.trans_dim,\n            depth=self.depth,\n            drop_path_rate=dpr,\n            num_heads=self.num_heads\n        )\n\n        self.norm = nn.LayerNorm(self.trans_dim)\n\n    def build_loss_func(self):\n        self.loss_ce = nn.CrossEntropyLoss()\n\n    def get_loss_acc(self, pred, gt, smoothing=True):\n        # import pdb; pdb.set_trace()\n        gt = gt.contiguous().view(-1).long()\n\n        if smoothing:\n            eps = 0.2\n            n_class = pred.size(1)\n\n            one_hot = torch.zeros_like(pred).scatter(1, gt.view(-1, 1), 1)\n            one_hot = one_hot * (1 - eps) + (1 - one_hot) * eps / (n_class - 1)\n            log_prb = F.log_softmax(pred, dim=1)\n\n            loss = -(one_hot * log_prb).sum(dim=1).mean()\n        else:\n            loss = self.loss_ce(pred, gt.long())\n\n        pred = pred.argmax(-1)\n        acc = (pred == gt).sum() / float(gt.size(0))\n\n        return loss, acc * 100\n\n    def load_model_from_ckpt(self, bert_ckpt_path):\n        ckpt = torch.load(bert_ckpt_path)\n        base_ckpt = {k.replace(\"module.\", \"\"): v for k, v in ckpt['base_model'].items()}\n        for k in list(base_ckpt.keys()):\n            if k.startswith('transformer_q') and not k.startswith('transformer_q.cls_head'):\n                base_ckpt[k[len('transformer_q.'):]] = base_ckpt[k]\n            elif k.startswith('base_model'):\n                base_ckpt[k[len('base_model.'):]] = base_ckpt[k]\n            del base_ckpt[k]\n\n        incompatible = self.load_state_dict(base_ckpt, strict=False)\n\n        if incompatible.missing_keys:\n            print_log('missing_keys', logger='Transformer')\n            print_log(\n                get_missing_parameters_message(incompatible.missing_keys),\n                logger='Transformer'\n            )\n        if incompatible.unexpected_keys:\n            print_log('unexpected_keys', logger='Transformer')\n            print_log(\n                get_unexpected_parameters_message(incompatible.unexpected_keys),\n                logger='Transformer'\n            )\n\n        print_log(f'[Transformer] Successful Loading the ckpt from {bert_ckpt_path}', logger='Transformer')\n\n    def forward(self, pts):\n        # divide the point cloud in the same form. This is important\n        neighborhood, center = self.group_divider(pts)\n        # encoder the input cloud blocks\n        group_input_tokens = self.encoder(neighborhood)  # B G N\n        group_input_tokens = self.reduce_dim(group_input_tokens)\n        # prepare cls\n        cls_tokens = self.cls_token.expand(group_input_tokens.size(0), -1, -1)\n        cls_pos = self.cls_pos.expand(group_input_tokens.size(0), -1, -1)\n        # add pos embedding\n        pos = self.pos_embed(center)\n        # final input\n        x = torch.cat((cls_tokens, group_input_tokens), dim=1)\n        pos = torch.cat((cls_pos, pos), dim=1)\n        # transformer\n        x = self.blocks(x, pos)\n        x = self.norm(x)\n        concat_f = torch.cat([x[:, 0], x[:, 1:].max(1)[0]], dim=-1)\n        return concat_f\n    \nclass PointTransformerBind(nn.Module):\n    def __init__(self):\n        super().__init__()\n        config_addr = './ImageBind/models/pointbert/PointTransformer_8192point.yaml'\n        config = cfg_from_yaml_file(config_addr)\n\n        self.point_encoder = PointTransformer(config.model)\n        self.pc_projection = nn.Parameter(torch.empty(768, 512))\n        nn.init.normal_(self.pc_projection, std=512 ** -0.5)\n\n    def forward(self, pc):\n        pc_feat = self.point_encoder(pc)\n        pc_embed = pc_feat @ self.pc_projection\n        return pc_embed","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.merge_new_config","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.point_encoder.merge_new_config#L12-L29","kind":"function","name":"merge_new_config","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":12,"end_line":29,"context_start_line":1,"context_end_line":49,"code":"import torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nfrom timm.models.layers import DropPath\nfrom .dvae import Group\nfrom .dvae import Encoder\nfrom .logger import print_log\nimport yaml\nfrom easydict import EasyDict\nfrom .checkpoint import get_missing_parameters_message, get_unexpected_parameters_message\n\ndef merge_new_config(config, new_config):\n    for key, val in new_config.items():\n        if not isinstance(val, dict):\n            if key == '_base_':\n                with open(new_config['_base_'], 'r') as f:\n                    try:\n                        val = yaml.load(f, Loader=yaml.FullLoader)\n                    except:\n                        val = yaml.load(f)\n                config[key] = EasyDict()\n                merge_new_config(config[key], val)\n            else:\n                config[key] = val\n                continue\n        if key not in config:\n            config[key] = EasyDict()\n        merge_new_config(config[key], val)\n    return config\n\ndef cfg_from_yaml_file(cfg_file):\n    config = EasyDict()\n    with open(cfg_file, 'r') as f:\n        new_config = yaml.load(f, Loader=yaml.FullLoader)\n    merge_new_config(config=config, new_config=new_config)\n    return config\n\nclass Mlp(nn.Module):\n    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):\n        super().__init__()\n        out_features = out_features or in_features\n        hidden_features = hidden_features or in_features\n        self.fc1 = nn.Linear(in_features, hidden_features)\n        self.act = act_layer()\n        self.fc2 = nn.Linear(hidden_features, out_features)\n        self.drop = nn.Dropout(drop)\n\n    def forward(self, x):\n        x = self.fc1(x)","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.cfg_from_yaml_file","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.point_encoder.cfg_from_yaml_file#L31-L36","kind":"function","name":"cfg_from_yaml_file","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":31,"end_line":36,"context_start_line":11,"context_end_line":56,"code":"\ndef merge_new_config(config, new_config):\n    for key, val in new_config.items():\n        if not isinstance(val, dict):\n            if key == '_base_':\n                with open(new_config['_base_'], 'r') as f:\n                    try:\n                        val = yaml.load(f, Loader=yaml.FullLoader)\n                    except:\n                        val = yaml.load(f)\n                config[key] = EasyDict()\n                merge_new_config(config[key], val)\n            else:\n                config[key] = val\n                continue\n        if key not in config:\n            config[key] = EasyDict()\n        merge_new_config(config[key], val)\n    return config\n\ndef cfg_from_yaml_file(cfg_file):\n    config = EasyDict()\n    with open(cfg_file, 'r') as f:\n        new_config = yaml.load(f, Loader=yaml.FullLoader)\n    merge_new_config(config=config, new_config=new_config)\n    return config\n\nclass Mlp(nn.Module):\n    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):\n        super().__init__()\n        out_features = out_features or in_features\n        hidden_features = hidden_features or in_features\n        self.fc1 = nn.Linear(in_features, hidden_features)\n        self.act = act_layer()\n        self.fc2 = nn.Linear(hidden_features, out_features)\n        self.drop = nn.Dropout(drop)\n\n    def forward(self, x):\n        x = self.fc1(x)\n        x = self.act(x)\n        x = self.drop(x)\n        x = self.fc2(x)\n        x = self.drop(x)\n        return x\n\n","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.Mlp","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.point_encoder.Mlp#L38-L54","kind":"class","name":"Mlp","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":38,"end_line":54,"context_start_line":18,"context_end_line":74,"code":"                        val = yaml.load(f, Loader=yaml.FullLoader)\n                    except:\n                        val = yaml.load(f)\n                config[key] = EasyDict()\n                merge_new_config(config[key], val)\n            else:\n                config[key] = val\n                continue\n        if key not in config:\n            config[key] = EasyDict()\n        merge_new_config(config[key], val)\n    return config\n\ndef cfg_from_yaml_file(cfg_file):\n    config = EasyDict()\n    with open(cfg_file, 'r') as f:\n        new_config = yaml.load(f, Loader=yaml.FullLoader)\n    merge_new_config(config=config, new_config=new_config)\n    return config\n\nclass Mlp(nn.Module):\n    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):\n        super().__init__()\n        out_features = out_features or in_features\n        hidden_features = hidden_features or in_features\n        self.fc1 = nn.Linear(in_features, hidden_features)\n        self.act = act_layer()\n        self.fc2 = nn.Linear(hidden_features, out_features)\n        self.drop = nn.Dropout(drop)\n\n    def forward(self, x):\n        x = self.fc1(x)\n        x = self.act(x)\n        x = self.drop(x)\n        x = self.fc2(x)\n        x = self.drop(x)\n        return x\n\n\nclass Attention(nn.Module):\n    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0.):\n        super().__init__()\n        self.num_heads = num_heads\n        head_dim = dim // num_heads\n        # NOTE scale factor was wrong in my original version, can set manually to be compat with prev weights\n        self.scale = qk_scale or head_dim ** -0.5\n\n        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)\n        self.attn_drop = nn.Dropout(attn_drop)\n        self.proj = nn.Linear(dim, dim)\n        self.proj_drop = nn.Dropout(proj_drop)\n\n    def forward(self, x):\n        B, N, C = x.shape\n        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)\n        q, k, v = qkv[0], qkv[1], qkv[2]  # make torchscript happy (cannot use tensor as tuple)\n","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.Attention","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.point_encoder.Attention#L57-L82","kind":"class","name":"Attention","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":57,"end_line":82,"context_start_line":37,"context_end_line":102,"code":"\nclass Mlp(nn.Module):\n    def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.):\n        super().__init__()\n        out_features = out_features or in_features\n        hidden_features = hidden_features or in_features\n        self.fc1 = nn.Linear(in_features, hidden_features)\n        self.act = act_layer()\n        self.fc2 = nn.Linear(hidden_features, out_features)\n        self.drop = nn.Dropout(drop)\n\n    def forward(self, x):\n        x = self.fc1(x)\n        x = self.act(x)\n        x = self.drop(x)\n        x = self.fc2(x)\n        x = self.drop(x)\n        return x\n\n\nclass Attention(nn.Module):\n    def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0.):\n        super().__init__()\n        self.num_heads = num_heads\n        head_dim = dim // num_heads\n        # NOTE scale factor was wrong in my original version, can set manually to be compat with prev weights\n        self.scale = qk_scale or head_dim ** -0.5\n\n        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)\n        self.attn_drop = nn.Dropout(attn_drop)\n        self.proj = nn.Linear(dim, dim)\n        self.proj_drop = nn.Dropout(proj_drop)\n\n    def forward(self, x):\n        B, N, C = x.shape\n        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)\n        q, k, v = qkv[0], qkv[1], qkv[2]  # make torchscript happy (cannot use tensor as tuple)\n\n        attn = (q @ k.transpose(-2, -1)) * self.scale\n        attn = attn.softmax(dim=-1)\n        attn = self.attn_drop(attn)\n\n        x = (attn @ v).transpose(1, 2).reshape(B, N, C)\n        x = self.proj(x)\n        x = self.proj_drop(x)\n        return x\n\n\nclass Block(nn.Module):\n    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,\n                 drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm):\n        super().__init__()\n        self.norm1 = norm_layer(dim)\n\n        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here\n        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()\n        self.norm2 = norm_layer(dim)\n        mlp_hidden_dim = int(dim * mlp_ratio)\n        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)\n\n        self.attn = Attention(\n            dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop)\n\n    def forward(self, x):\n        x = x + self.drop_path(self.attn(self.norm1(x)))\n        x = x + self.drop_path(self.mlp(self.norm2(x)))","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.Block","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.point_encoder.Block#L85-L103","kind":"class","name":"Block","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":85,"end_line":103,"context_start_line":65,"context_end_line":123,"code":"        self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)\n        self.attn_drop = nn.Dropout(attn_drop)\n        self.proj = nn.Linear(dim, dim)\n        self.proj_drop = nn.Dropout(proj_drop)\n\n    def forward(self, x):\n        B, N, C = x.shape\n        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)\n        q, k, v = qkv[0], qkv[1], qkv[2]  # make torchscript happy (cannot use tensor as tuple)\n\n        attn = (q @ k.transpose(-2, -1)) * self.scale\n        attn = attn.softmax(dim=-1)\n        attn = self.attn_drop(attn)\n\n        x = (attn @ v).transpose(1, 2).reshape(B, N, C)\n        x = self.proj(x)\n        x = self.proj_drop(x)\n        return x\n\n\nclass Block(nn.Module):\n    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,\n                 drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm):\n        super().__init__()\n        self.norm1 = norm_layer(dim)\n\n        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here\n        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()\n        self.norm2 = norm_layer(dim)\n        mlp_hidden_dim = int(dim * mlp_ratio)\n        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)\n\n        self.attn = Attention(\n            dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop)\n\n    def forward(self, x):\n        x = x + self.drop_path(self.attn(self.norm1(x)))\n        x = x + self.drop_path(self.mlp(self.norm2(x)))\n        return x\n\n\nclass TransformerEncoder(nn.Module):\n    \"\"\" Transformer Encoder without hierarchical structure\n    \"\"\"\n\n    def __init__(self, embed_dim=768, depth=4, num_heads=12, mlp_ratio=4., qkv_bias=False, qk_scale=None,\n                 drop_rate=0., attn_drop_rate=0., drop_path_rate=0.):\n        super().__init__()\n\n        self.blocks = nn.ModuleList([\n            Block(\n                dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale,\n                drop=drop_rate, attn_drop=attn_drop_rate,\n                drop_path=drop_path_rate[i] if isinstance(drop_path_rate, list) else drop_path_rate\n            )\n            for i in range(depth)])\n\n    def forward(self, x, pos):\n        for _, block in enumerate(self.blocks):","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.TransformerEncoder","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.point_encoder.TransformerEncoder#L106-L125","kind":"class","name":"TransformerEncoder","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":106,"end_line":125,"context_start_line":86,"context_end_line":145,"code":"    def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0.,\n                 drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm):\n        super().__init__()\n        self.norm1 = norm_layer(dim)\n\n        # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here\n        self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity()\n        self.norm2 = norm_layer(dim)\n        mlp_hidden_dim = int(dim * mlp_ratio)\n        self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop)\n\n        self.attn = Attention(\n            dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop)\n\n    def forward(self, x):\n        x = x + self.drop_path(self.attn(self.norm1(x)))\n        x = x + self.drop_path(self.mlp(self.norm2(x)))\n        return x\n\n\nclass TransformerEncoder(nn.Module):\n    \"\"\" Transformer Encoder without hierarchical structure\n    \"\"\"\n\n    def __init__(self, embed_dim=768, depth=4, num_heads=12, mlp_ratio=4., qkv_bias=False, qk_scale=None,\n                 drop_rate=0., attn_drop_rate=0., drop_path_rate=0.):\n        super().__init__()\n\n        self.blocks = nn.ModuleList([\n            Block(\n                dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale,\n                drop=drop_rate, attn_drop=attn_drop_rate,\n                drop_path=drop_path_rate[i] if isinstance(drop_path_rate, list) else drop_path_rate\n            )\n            for i in range(depth)])\n\n    def forward(self, x, pos):\n        for _, block in enumerate(self.blocks):\n            x = block(x + pos)\n        return x\n\nclass PointTransformer(nn.Module):\n    def __init__(self, config):\n        super().__init__()\n        self.config = config\n\n        self.trans_dim = config.trans_dim\n        self.depth = config.depth\n        self.drop_path_rate = config.drop_path_rate\n        self.cls_dim = config.cls_dim\n        self.num_heads = config.num_heads\n\n        self.group_size = config.group_size\n        self.num_group = config.num_group\n        # grouper\n        self.group_divider = Group(num_group=self.num_group, group_size=self.group_size)\n        # define the encoder\n        self.encoder_dims = config.encoder_dims\n        self.encoder = Encoder(encoder_channel=self.encoder_dims)\n        # bridge encoder and transformer","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.PointTransformer","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.point_encoder.PointTransformer#L127-L236","kind":"class","name":"PointTransformer","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":127,"end_line":236,"context_start_line":107,"context_end_line":251,"code":"    \"\"\" Transformer Encoder without hierarchical structure\n    \"\"\"\n\n    def __init__(self, embed_dim=768, depth=4, num_heads=12, mlp_ratio=4., qkv_bias=False, qk_scale=None,\n                 drop_rate=0., attn_drop_rate=0., drop_path_rate=0.):\n        super().__init__()\n\n        self.blocks = nn.ModuleList([\n            Block(\n                dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale,\n                drop=drop_rate, attn_drop=attn_drop_rate,\n                drop_path=drop_path_rate[i] if isinstance(drop_path_rate, list) else drop_path_rate\n            )\n            for i in range(depth)])\n\n    def forward(self, x, pos):\n        for _, block in enumerate(self.blocks):\n            x = block(x + pos)\n        return x\n\nclass PointTransformer(nn.Module):\n    def __init__(self, config):\n        super().__init__()\n        self.config = config\n\n        self.trans_dim = config.trans_dim\n        self.depth = config.depth\n        self.drop_path_rate = config.drop_path_rate\n        self.cls_dim = config.cls_dim\n        self.num_heads = config.num_heads\n\n        self.group_size = config.group_size\n        self.num_group = config.num_group\n        # grouper\n        self.group_divider = Group(num_group=self.num_group, group_size=self.group_size)\n        # define the encoder\n        self.encoder_dims = config.encoder_dims\n        self.encoder = Encoder(encoder_channel=self.encoder_dims)\n        # bridge encoder and transformer\n        self.reduce_dim = nn.Linear(self.encoder_dims, self.trans_dim)\n\n        self.cls_token = nn.Parameter(torch.zeros(1, 1, self.trans_dim))\n        self.cls_pos = nn.Parameter(torch.randn(1, 1, self.trans_dim))\n\n        self.pos_embed = nn.Sequential(\n            nn.Linear(3, 128),\n            nn.GELU(),\n            nn.Linear(128, self.trans_dim)\n        )\n\n        dpr = [x.item() for x in torch.linspace(0, self.drop_path_rate, self.depth)]\n        self.blocks = TransformerEncoder(\n            embed_dim=self.trans_dim,\n            depth=self.depth,\n            drop_path_rate=dpr,\n            num_heads=self.num_heads\n        )\n\n        self.norm = nn.LayerNorm(self.trans_dim)\n\n    def build_loss_func(self):\n        self.loss_ce = nn.CrossEntropyLoss()\n\n    def get_loss_acc(self, pred, gt, smoothing=True):\n        # import pdb; pdb.set_trace()\n        gt = gt.contiguous().view(-1).long()\n\n        if smoothing:\n            eps = 0.2\n            n_class = pred.size(1)\n\n            one_hot = torch.zeros_like(pred).scatter(1, gt.view(-1, 1), 1)\n            one_hot = one_hot * (1 - eps) + (1 - one_hot) * eps / (n_class - 1)\n            log_prb = F.log_softmax(pred, dim=1)\n\n            loss = -(one_hot * log_prb).sum(dim=1).mean()\n        else:\n            loss = self.loss_ce(pred, gt.long())\n\n        pred = pred.argmax(-1)\n        acc = (pred == gt).sum() / float(gt.size(0))\n\n        return loss, acc * 100\n\n    def load_model_from_ckpt(self, bert_ckpt_path):\n        ckpt = torch.load(bert_ckpt_path)\n        base_ckpt = {k.replace(\"module.\", \"\"): v for k, v in ckpt['base_model'].items()}\n        for k in list(base_ckpt.keys()):\n            if k.startswith('transformer_q') and not k.startswith('transformer_q.cls_head'):\n                base_ckpt[k[len('transformer_q.'):]] = base_ckpt[k]\n            elif k.startswith('base_model'):\n                base_ckpt[k[len('base_model.'):]] = base_ckpt[k]\n            del base_ckpt[k]\n\n        incompatible = self.load_state_dict(base_ckpt, strict=False)\n\n        if incompatible.missing_keys:\n            print_log('missing_keys', logger='Transformer')\n            print_log(\n                get_missing_parameters_message(incompatible.missing_keys),\n                logger='Transformer'\n            )\n        if incompatible.unexpected_keys:\n            print_log('unexpected_keys', logger='Transformer')\n            print_log(\n                get_unexpected_parameters_message(incompatible.unexpected_keys),\n                logger='Transformer'\n            )\n\n        print_log(f'[Transformer] Successful Loading the ckpt from {bert_ckpt_path}', logger='Transformer')\n\n    def forward(self, pts):\n        # divide the point cloud in the same form. This is important\n        neighborhood, center = self.group_divider(pts)\n        # encoder the input cloud blocks\n        group_input_tokens = self.encoder(neighborhood)  # B G N\n        group_input_tokens = self.reduce_dim(group_input_tokens)\n        # prepare cls\n        cls_tokens = self.cls_token.expand(group_input_tokens.size(0), -1, -1)\n        cls_pos = self.cls_pos.expand(group_input_tokens.size(0), -1, -1)\n        # add pos embedding\n        pos = self.pos_embed(center)\n        # final input\n        x = torch.cat((cls_tokens, group_input_tokens), dim=1)\n        pos = torch.cat((cls_pos, pos), dim=1)\n        # transformer\n        x = self.blocks(x, pos)\n        x = self.norm(x)\n        concat_f = torch.cat([x[:, 0], x[:, 1:].max(1)[0]], dim=-1)\n        return concat_f\n    \nclass PointTransformerBind(nn.Module):\n    def __init__(self):\n        super().__init__()\n        config_addr = './ImageBind/models/pointbert/PointTransformer_8192point.yaml'\n        config = cfg_from_yaml_file(config_addr)\n\n        self.point_encoder = PointTransformer(config.model)\n        self.pc_projection = nn.Parameter(torch.empty(768, 512))\n        nn.init.normal_(self.pc_projection, std=512 ** -0.5)\n\n    def forward(self, pc):\n        pc_feat = self.point_encoder(pc)\n        pc_embed = pc_feat @ self.pc_projection\n        return pc_embed","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.PointTransformerBind","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.point_encoder.PointTransformerBind#L238-L251","kind":"class","name":"PointTransformerBind","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":238,"end_line":251,"context_start_line":218,"context_end_line":251,"code":"    def forward(self, pts):\n        # divide the point cloud in the same form. This is important\n        neighborhood, center = self.group_divider(pts)\n        # encoder the input cloud blocks\n        group_input_tokens = self.encoder(neighborhood)  # B G N\n        group_input_tokens = self.reduce_dim(group_input_tokens)\n        # prepare cls\n        cls_tokens = self.cls_token.expand(group_input_tokens.size(0), -1, -1)\n        cls_pos = self.cls_pos.expand(group_input_tokens.size(0), -1, -1)\n        # add pos embedding\n        pos = self.pos_embed(center)\n        # final input\n        x = torch.cat((cls_tokens, group_input_tokens), dim=1)\n        pos = torch.cat((cls_pos, pos), dim=1)\n        # transformer\n        x = self.blocks(x, pos)\n        x = self.norm(x)\n        concat_f = torch.cat([x[:, 0], x[:, 1:].max(1)[0]], dim=-1)\n        return concat_f\n    \nclass PointTransformerBind(nn.Module):\n    def __init__(self):\n        super().__init__()\n        config_addr = './ImageBind/models/pointbert/PointTransformer_8192point.yaml'\n        config = cfg_from_yaml_file(config_addr)\n\n        self.point_encoder = PointTransformer(config.model)\n        self.pc_projection = nn.Parameter(torch.empty(768, 512))\n        nn.init.normal_(self.pc_projection, std=512 ** -0.5)\n\n    def forward(self, pc):\n        pc_feat = self.point_encoder(pc)\n        pc_embed = pc_feat @ self.pc_projection\n        return pc_embed","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.point_encoder.__init__#L239-L246","kind":"function","name":"__init__","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":239,"end_line":246,"context_start_line":219,"context_end_line":251,"code":"        # divide the point cloud in the same form. This is important\n        neighborhood, center = self.group_divider(pts)\n        # encoder the input cloud blocks\n        group_input_tokens = self.encoder(neighborhood)  # B G N\n        group_input_tokens = self.reduce_dim(group_input_tokens)\n        # prepare cls\n        cls_tokens = self.cls_token.expand(group_input_tokens.size(0), -1, -1)\n        cls_pos = self.cls_pos.expand(group_input_tokens.size(0), -1, -1)\n        # add pos embedding\n        pos = self.pos_embed(center)\n        # final input\n        x = torch.cat((cls_tokens, group_input_tokens), dim=1)\n        pos = torch.cat((cls_pos, pos), dim=1)\n        # transformer\n        x = self.blocks(x, pos)\n        x = self.norm(x)\n        concat_f = torch.cat([x[:, 0], x[:, 1:].max(1)[0]], dim=-1)\n        return concat_f\n    \nclass PointTransformerBind(nn.Module):\n    def __init__(self):\n        super().__init__()\n        config_addr = './ImageBind/models/pointbert/PointTransformer_8192point.yaml'\n        config = cfg_from_yaml_file(config_addr)\n\n        self.point_encoder = PointTransformer(config.model)\n        self.pc_projection = nn.Parameter(torch.empty(768, 512))\n        nn.init.normal_(self.pc_projection, std=512 ** -0.5)\n\n    def forward(self, pc):\n        pc_feat = self.point_encoder(pc)\n        pc_embed = pc_feat @ self.pc_projection\n        return pc_embed","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.forward","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.point_encoder.forward#L248-L251","kind":"function","name":"forward","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":248,"end_line":251,"context_start_line":228,"context_end_line":251,"code":"        pos = self.pos_embed(center)\n        # final input\n        x = torch.cat((cls_tokens, group_input_tokens), dim=1)\n        pos = torch.cat((cls_pos, pos), dim=1)\n        # transformer\n        x = self.blocks(x, pos)\n        x = self.norm(x)\n        concat_f = torch.cat([x[:, 0], x[:, 1:].max(1)[0]], dim=-1)\n        return concat_f\n    \nclass PointTransformerBind(nn.Module):\n    def __init__(self):\n        super().__init__()\n        config_addr = './ImageBind/models/pointbert/PointTransformer_8192point.yaml'\n        config = cfg_from_yaml_file(config_addr)\n\n        self.point_encoder = PointTransformer(config.model)\n        self.pc_projection = nn.Parameter(torch.empty(768, 512))\n        nn.init.normal_(self.pc_projection, std=512 ** -0.5)\n\n    def forward(self, pc):\n        pc_feat = self.point_encoder(pc)\n        pc_embed = pc_feat @ self.pc_projection\n        return pc_embed","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.build_loss_func","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.point_encoder.build_loss_func#L167-L168","kind":"function","name":"build_loss_func","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":167,"end_line":168,"context_start_line":147,"context_end_line":188,"code":"\n        self.cls_token = nn.Parameter(torch.zeros(1, 1, self.trans_dim))\n        self.cls_pos = nn.Parameter(torch.randn(1, 1, self.trans_dim))\n\n        self.pos_embed = nn.Sequential(\n            nn.Linear(3, 128),\n            nn.GELU(),\n            nn.Linear(128, self.trans_dim)\n        )\n\n        dpr = [x.item() for x in torch.linspace(0, self.drop_path_rate, self.depth)]\n        self.blocks = TransformerEncoder(\n            embed_dim=self.trans_dim,\n            depth=self.depth,\n            drop_path_rate=dpr,\n            num_heads=self.num_heads\n        )\n\n        self.norm = nn.LayerNorm(self.trans_dim)\n\n    def build_loss_func(self):\n        self.loss_ce = nn.CrossEntropyLoss()\n\n    def get_loss_acc(self, pred, gt, smoothing=True):\n        # import pdb; pdb.set_trace()\n        gt = gt.contiguous().view(-1).long()\n\n        if smoothing:\n            eps = 0.2\n            n_class = pred.size(1)\n\n            one_hot = torch.zeros_like(pred).scatter(1, gt.view(-1, 1), 1)\n            one_hot = one_hot * (1 - eps) + (1 - one_hot) * eps / (n_class - 1)\n            log_prb = F.log_softmax(pred, dim=1)\n\n            loss = -(one_hot * log_prb).sum(dim=1).mean()\n        else:\n            loss = self.loss_ce(pred, gt.long())\n\n        pred = pred.argmax(-1)\n        acc = (pred == gt).sum() / float(gt.size(0))\n","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.get_loss_acc","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.point_encoder.get_loss_acc#L170-L189","kind":"function","name":"get_loss_acc","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":170,"end_line":189,"context_start_line":150,"context_end_line":209,"code":"\n        self.pos_embed = nn.Sequential(\n            nn.Linear(3, 128),\n            nn.GELU(),\n            nn.Linear(128, self.trans_dim)\n        )\n\n        dpr = [x.item() for x in torch.linspace(0, self.drop_path_rate, self.depth)]\n        self.blocks = TransformerEncoder(\n            embed_dim=self.trans_dim,\n            depth=self.depth,\n            drop_path_rate=dpr,\n            num_heads=self.num_heads\n        )\n\n        self.norm = nn.LayerNorm(self.trans_dim)\n\n    def build_loss_func(self):\n        self.loss_ce = nn.CrossEntropyLoss()\n\n    def get_loss_acc(self, pred, gt, smoothing=True):\n        # import pdb; pdb.set_trace()\n        gt = gt.contiguous().view(-1).long()\n\n        if smoothing:\n            eps = 0.2\n            n_class = pred.size(1)\n\n            one_hot = torch.zeros_like(pred).scatter(1, gt.view(-1, 1), 1)\n            one_hot = one_hot * (1 - eps) + (1 - one_hot) * eps / (n_class - 1)\n            log_prb = F.log_softmax(pred, dim=1)\n\n            loss = -(one_hot * log_prb).sum(dim=1).mean()\n        else:\n            loss = self.loss_ce(pred, gt.long())\n\n        pred = pred.argmax(-1)\n        acc = (pred == gt).sum() / float(gt.size(0))\n\n        return loss, acc * 100\n\n    def load_model_from_ckpt(self, bert_ckpt_path):\n        ckpt = torch.load(bert_ckpt_path)\n        base_ckpt = {k.replace(\"module.\", \"\"): v for k, v in ckpt['base_model'].items()}\n        for k in list(base_ckpt.keys()):\n            if k.startswith('transformer_q') and not k.startswith('transformer_q.cls_head'):\n                base_ckpt[k[len('transformer_q.'):]] = base_ckpt[k]\n            elif k.startswith('base_model'):\n                base_ckpt[k[len('base_model.'):]] = base_ckpt[k]\n            del base_ckpt[k]\n\n        incompatible = self.load_state_dict(base_ckpt, strict=False)\n\n        if incompatible.missing_keys:\n            print_log('missing_keys', logger='Transformer')\n            print_log(\n                get_missing_parameters_message(incompatible.missing_keys),\n                logger='Transformer'\n            )\n        if incompatible.unexpected_keys:","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.point_encoder.load_model_from_ckpt","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.point_encoder.load_model_from_ckpt#L191-L216","kind":"function","name":"load_model_from_ckpt","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":191,"end_line":216,"context_start_line":171,"context_end_line":236,"code":"        # import pdb; pdb.set_trace()\n        gt = gt.contiguous().view(-1).long()\n\n        if smoothing:\n            eps = 0.2\n            n_class = pred.size(1)\n\n            one_hot = torch.zeros_like(pred).scatter(1, gt.view(-1, 1), 1)\n            one_hot = one_hot * (1 - eps) + (1 - one_hot) * eps / (n_class - 1)\n            log_prb = F.log_softmax(pred, dim=1)\n\n            loss = -(one_hot * log_prb).sum(dim=1).mean()\n        else:\n            loss = self.loss_ce(pred, gt.long())\n\n        pred = pred.argmax(-1)\n        acc = (pred == gt).sum() / float(gt.size(0))\n\n        return loss, acc * 100\n\n    def load_model_from_ckpt(self, bert_ckpt_path):\n        ckpt = torch.load(bert_ckpt_path)\n        base_ckpt = {k.replace(\"module.\", \"\"): v for k, v in ckpt['base_model'].items()}\n        for k in list(base_ckpt.keys()):\n            if k.startswith('transformer_q') and not k.startswith('transformer_q.cls_head'):\n                base_ckpt[k[len('transformer_q.'):]] = base_ckpt[k]\n            elif k.startswith('base_model'):\n                base_ckpt[k[len('base_model.'):]] = base_ckpt[k]\n            del base_ckpt[k]\n\n        incompatible = self.load_state_dict(base_ckpt, strict=False)\n\n        if incompatible.missing_keys:\n            print_log('missing_keys', logger='Transformer')\n            print_log(\n                get_missing_parameters_message(incompatible.missing_keys),\n                logger='Transformer'\n            )\n        if incompatible.unexpected_keys:\n            print_log('unexpected_keys', logger='Transformer')\n            print_log(\n                get_unexpected_parameters_message(incompatible.unexpected_keys),\n                logger='Transformer'\n            )\n\n        print_log(f'[Transformer] Successful Loading the ckpt from {bert_ckpt_path}', logger='Transformer')\n\n    def forward(self, pts):\n        # divide the point cloud in the same form. This is important\n        neighborhood, center = self.group_divider(pts)\n        # encoder the input cloud blocks\n        group_input_tokens = self.encoder(neighborhood)  # B G N\n        group_input_tokens = self.reduce_dim(group_input_tokens)\n        # prepare cls\n        cls_tokens = self.cls_token.expand(group_input_tokens.size(0), -1, -1)\n        cls_pos = self.cls_pos.expand(group_input_tokens.size(0), -1, -1)\n        # add pos embedding\n        pos = self.pos_embed(center)\n        # final input\n        x = torch.cat((cls_tokens, group_input_tokens), dim=1)\n        pos = torch.cat((cls_pos, pos), dim=1)\n        # transformer\n        x = self.blocks(x, pos)\n        x = self.norm(x)\n        concat_f = torch.cat([x[:, 0], x[:, 1:].max(1)[0]], dim=-1)\n        return concat_f","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.logger","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.models.pointbert.logger#L1-L127","kind":"module","name":"imagebind_LLM.ImageBind.models.pointbert.logger","path":"imagebind_LLM/ImageBind/models/pointbert/logger.py","language":"python","start_line":1,"end_line":127,"context_start_line":1,"context_end_line":127,"code":"import logging\nimport torch.distributed as dist\n\nlogger_initialized = {}\n\ndef get_root_logger(log_file=None, log_level=logging.INFO, name='main'):\n    \"\"\"Get root logger and add a keyword filter to it.\n    The logger will be initialized if it has not been initialized. By default a\n    StreamHandler will be added. If `log_file` is specified, a FileHandler will\n    also be added. The name of the root logger is the top-level package name,\n    e.g., \"mmdet3d\".\n    Args:\n        log_file (str, optional): File path of log. Defaults to None.\n        log_level (int, optional): The level of logger.\n            Defaults to logging.INFO.\n        name (str, optional): The name of the root logger, also used as a\n            filter keyword. Defaults to 'mmdet3d'.\n    Returns:\n        :obj:`logging.Logger`: The obtained logger\n    \"\"\"\n    logger = get_logger(name=name, log_file=log_file, log_level=log_level)\n    # add a logging filter\n    logging_filter = logging.Filter(name)\n    logging_filter.filter = lambda record: record.find(name) != -1\n\n    return logger\n\n\ndef get_logger(name, log_file=None, log_level=logging.INFO, file_mode='w'):\n    \"\"\"Initialize and get a logger by name.\n    If the logger has not been initialized, this method will initialize the\n    logger by adding one or two handlers, otherwise the initialized logger will\n    be directly returned. During initialization, a StreamHandler will always be\n    added. If `log_file` is specified and the process rank is 0, a FileHandler\n    will also be added.\n    Args:\n        name (str): Logger name.\n        log_file (str | None): The log filename. If specified, a FileHandler\n            will be added to the logger.\n        log_level (int): The logger level. Note that only the process of\n            rank 0 is affected, and other processes will set the level to\n            \"Error\" thus be silent most of the time.\n        file_mode (str): The file mode used in opening log file.\n            Defaults to 'w'.\n    Returns:\n        logging.Logger: The expected logger.\n    \"\"\"\n    logger = logging.getLogger(name)\n    if name in logger_initialized:\n        return logger\n    # handle hierarchical names\n    # e.g., logger \"a\" is initialized, then logger \"a.b\" will skip the\n    # initialization since it is a child of \"a\".\n    for logger_name in logger_initialized:\n        if name.startswith(logger_name):\n            return logger\n\n    # handle duplicate logs to the console\n    # Starting in 1.8.0, PyTorch DDP attaches a StreamHandler <stderr> (NOTSET)\n    # to the root logger. As logger.propagate is True by default, this root\n    # level handler causes logging messages from rank>0 processes to\n    # unexpectedly show up on the console, creating much unwanted clutter.\n    # To fix this issue, we set the root logger's StreamHandler, if any, to log\n    # at the ERROR level.\n    for handler in logger.root.handlers:\n        if type(handler) is logging.StreamHandler:\n            handler.setLevel(logging.ERROR)\n\n    stream_handler = logging.StreamHandler()\n    handlers = [stream_handler]\n\n    if dist.is_available() and dist.is_initialized():\n        rank = dist.get_rank()\n    else:\n        rank = 0\n\n    # only rank 0 will add a FileHandler\n    if rank == 0 and log_file is not None:\n        # Here, the default behaviour of the official logger is 'a'. Thus, we\n        # provide an interface to change the file mode to the default\n        # behaviour.\n        file_handler = logging.FileHandler(log_file, file_mode)\n        handlers.append(file_handler)\n\n    formatter = logging.Formatter(\n        '%(asctime)s - %(name)s - %(levelname)s - %(message)s')\n    for handler in handlers:\n        handler.setFormatter(formatter)\n        handler.setLevel(log_level)\n        logger.addHandler(handler)\n\n    if rank == 0:\n        logger.setLevel(log_level)\n    else:\n        logger.setLevel(logging.ERROR)\n\n    logger_initialized[name] = True\n\n\n    return logger\n\n\ndef print_log(msg, logger=None, level=logging.INFO):\n    \"\"\"Print a log message.\n    Args:\n        msg (str): The message to be logged.\n        logger (logging.Logger | str | None): The logger to be used.\n            Some special loggers are:\n            - \"silent\": no message will be printed.\n            - other str: the logger obtained with `get_root_logger(logger)`.\n            - None: The `print()` method will be used to print log messages.\n        level (int): Logging level. Only available when `logger` is a Logger\n            object or \"root\".\n    \"\"\"\n    if logger is None:\n        print(msg)\n    elif isinstance(logger, logging.Logger):\n        logger.log(level, msg)\n    elif logger == 'silent':\n        pass\n    elif isinstance(logger, str):\n        _logger = get_logger(logger)\n        _logger.log(level, msg)\n    else:\n        raise TypeError(\n            'logger should be either a logging.Logger object, str, '\n            f'\"silent\" or None, but got {type(logger)}')","source_hash":"d5ce96284512f59bfe386c857c9e3587fbd9afd605d4140f3383bfe7d0a03f69","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.logger.get_root_logger","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.logger.get_root_logger#L6-L26","kind":"function","name":"get_root_logger","path":"imagebind_LLM/ImageBind/models/pointbert/logger.py","language":"python","start_line":6,"end_line":26,"context_start_line":1,"context_end_line":46,"code":"import logging\nimport torch.distributed as dist\n\nlogger_initialized = {}\n\ndef get_root_logger(log_file=None, log_level=logging.INFO, name='main'):\n    \"\"\"Get root logger and add a keyword filter to it.\n    The logger will be initialized if it has not been initialized. By default a\n    StreamHandler will be added. If `log_file` is specified, a FileHandler will\n    also be added. The name of the root logger is the top-level package name,\n    e.g., \"mmdet3d\".\n    Args:\n        log_file (str, optional): File path of log. Defaults to None.\n        log_level (int, optional): The level of logger.\n            Defaults to logging.INFO.\n        name (str, optional): The name of the root logger, also used as a\n            filter keyword. Defaults to 'mmdet3d'.\n    Returns:\n        :obj:`logging.Logger`: The obtained logger\n    \"\"\"\n    logger = get_logger(name=name, log_file=log_file, log_level=log_level)\n    # add a logging filter\n    logging_filter = logging.Filter(name)\n    logging_filter.filter = lambda record: record.find(name) != -1\n\n    return logger\n\n\ndef get_logger(name, log_file=None, log_level=logging.INFO, file_mode='w'):\n    \"\"\"Initialize and get a logger by name.\n    If the logger has not been initialized, this method will initialize the\n    logger by adding one or two handlers, otherwise the initialized logger will\n    be directly returned. During initialization, a StreamHandler will always be\n    added. If `log_file` is specified and the process rank is 0, a FileHandler\n    will also be added.\n    Args:\n        name (str): Logger name.\n        log_file (str | None): The log filename. If specified, a FileHandler\n            will be added to the logger.\n        log_level (int): The logger level. Note that only the process of\n            rank 0 is affected, and other processes will set the level to\n            \"Error\" thus be silent most of the time.\n        file_mode (str): The file mode used in opening log file.\n            Defaults to 'w'.\n    Returns:\n        logging.Logger: The expected logger.","source_hash":"d5ce96284512f59bfe386c857c9e3587fbd9afd605d4140f3383bfe7d0a03f69","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.logger.get_logger","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.logger.get_logger#L29-L100","kind":"function","name":"get_logger","path":"imagebind_LLM/ImageBind/models/pointbert/logger.py","language":"python","start_line":29,"end_line":100,"context_start_line":9,"context_end_line":120,"code":"    StreamHandler will be added. If `log_file` is specified, a FileHandler will\n    also be added. The name of the root logger is the top-level package name,\n    e.g., \"mmdet3d\".\n    Args:\n        log_file (str, optional): File path of log. Defaults to None.\n        log_level (int, optional): The level of logger.\n            Defaults to logging.INFO.\n        name (str, optional): The name of the root logger, also used as a\n            filter keyword. Defaults to 'mmdet3d'.\n    Returns:\n        :obj:`logging.Logger`: The obtained logger\n    \"\"\"\n    logger = get_logger(name=name, log_file=log_file, log_level=log_level)\n    # add a logging filter\n    logging_filter = logging.Filter(name)\n    logging_filter.filter = lambda record: record.find(name) != -1\n\n    return logger\n\n\ndef get_logger(name, log_file=None, log_level=logging.INFO, file_mode='w'):\n    \"\"\"Initialize and get a logger by name.\n    If the logger has not been initialized, this method will initialize the\n    logger by adding one or two handlers, otherwise the initialized logger will\n    be directly returned. During initialization, a StreamHandler will always be\n    added. If `log_file` is specified and the process rank is 0, a FileHandler\n    will also be added.\n    Args:\n        name (str): Logger name.\n        log_file (str | None): The log filename. If specified, a FileHandler\n            will be added to the logger.\n        log_level (int): The logger level. Note that only the process of\n            rank 0 is affected, and other processes will set the level to\n            \"Error\" thus be silent most of the time.\n        file_mode (str): The file mode used in opening log file.\n            Defaults to 'w'.\n    Returns:\n        logging.Logger: The expected logger.\n    \"\"\"\n    logger = logging.getLogger(name)\n    if name in logger_initialized:\n        return logger\n    # handle hierarchical names\n    # e.g., logger \"a\" is initialized, then logger \"a.b\" will skip the\n    # initialization since it is a child of \"a\".\n    for logger_name in logger_initialized:\n        if name.startswith(logger_name):\n            return logger\n\n    # handle duplicate logs to the console\n    # Starting in 1.8.0, PyTorch DDP attaches a StreamHandler <stderr> (NOTSET)\n    # to the root logger. As logger.propagate is True by default, this root\n    # level handler causes logging messages from rank>0 processes to\n    # unexpectedly show up on the console, creating much unwanted clutter.\n    # To fix this issue, we set the root logger's StreamHandler, if any, to log\n    # at the ERROR level.\n    for handler in logger.root.handlers:\n        if type(handler) is logging.StreamHandler:\n            handler.setLevel(logging.ERROR)\n\n    stream_handler = logging.StreamHandler()\n    handlers = [stream_handler]\n\n    if dist.is_available() and dist.is_initialized():\n        rank = dist.get_rank()\n    else:\n        rank = 0\n\n    # only rank 0 will add a FileHandler\n    if rank == 0 and log_file is not None:\n        # Here, the default behaviour of the official logger is 'a'. Thus, we\n        # provide an interface to change the file mode to the default\n        # behaviour.\n        file_handler = logging.FileHandler(log_file, file_mode)\n        handlers.append(file_handler)\n\n    formatter = logging.Formatter(\n        '%(asctime)s - %(name)s - %(levelname)s - %(message)s')\n    for handler in handlers:\n        handler.setFormatter(formatter)\n        handler.setLevel(log_level)\n        logger.addHandler(handler)\n\n    if rank == 0:\n        logger.setLevel(log_level)\n    else:\n        logger.setLevel(logging.ERROR)\n\n    logger_initialized[name] = True\n\n\n    return logger\n\n\ndef print_log(msg, logger=None, level=logging.INFO):\n    \"\"\"Print a log message.\n    Args:\n        msg (str): The message to be logged.\n        logger (logging.Logger | str | None): The logger to be used.\n            Some special loggers are:\n            - \"silent\": no message will be printed.\n            - other str: the logger obtained with `get_root_logger(logger)`.\n            - None: The `print()` method will be used to print log messages.\n        level (int): Logging level. Only available when `logger` is a Logger\n            object or \"root\".\n    \"\"\"\n    if logger is None:\n        print(msg)\n    elif isinstance(logger, logging.Logger):\n        logger.log(level, msg)\n    elif logger == 'silent':\n        pass","source_hash":"d5ce96284512f59bfe386c857c9e3587fbd9afd605d4140f3383bfe7d0a03f69","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.logger.print_log","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.logger.print_log#L103-L127","kind":"function","name":"print_log","path":"imagebind_LLM/ImageBind/models/pointbert/logger.py","language":"python","start_line":103,"end_line":127,"context_start_line":83,"context_end_line":127,"code":"        handlers.append(file_handler)\n\n    formatter = logging.Formatter(\n        '%(asctime)s - %(name)s - %(levelname)s - %(message)s')\n    for handler in handlers:\n        handler.setFormatter(formatter)\n        handler.setLevel(log_level)\n        logger.addHandler(handler)\n\n    if rank == 0:\n        logger.setLevel(log_level)\n    else:\n        logger.setLevel(logging.ERROR)\n\n    logger_initialized[name] = True\n\n\n    return logger\n\n\ndef print_log(msg, logger=None, level=logging.INFO):\n    \"\"\"Print a log message.\n    Args:\n        msg (str): The message to be logged.\n        logger (logging.Logger | str | None): The logger to be used.\n            Some special loggers are:\n            - \"silent\": no message will be printed.\n            - other str: the logger obtained with `get_root_logger(logger)`.\n            - None: The `print()` method will be used to print log messages.\n        level (int): Logging level. Only available when `logger` is a Logger\n            object or \"root\".\n    \"\"\"\n    if logger is None:\n        print(msg)\n    elif isinstance(logger, logging.Logger):\n        logger.log(level, msg)\n    elif logger == 'silent':\n        pass\n    elif isinstance(logger, str):\n        _logger = get_logger(logger)\n        _logger.log(level, msg)\n    else:\n        raise TypeError(\n            'logger should be either a logging.Logger object, str, '\n            f'\"silent\" or None, but got {type(logger)}')","source_hash":"d5ce96284512f59bfe386c857c9e3587fbd9afd605d4140f3383bfe7d0a03f69","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae","uri":"program://LLaMA-Adapter/module/imagebind_LLM.ImageBind.models.pointbert.dvae#L1-L340","kind":"module","name":"imagebind_LLM.ImageBind.models.pointbert.dvae","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":1,"end_line":340,"context_start_line":1,"context_end_line":340,"code":"import torch.nn as nn\nimport torch\nimport torch.nn.functional as F\nfrom . import misc\n\nfrom knn_cuda import KNN\n\nknn = KNN(k=4, transpose_mode=False)\n\n\nclass DGCNN(nn.Module):\n    def __init__(self, encoder_channel, output_channel):\n        super().__init__()\n        '''\n        K has to be 16\n        '''\n        self.input_trans = nn.Conv1d(encoder_channel, 128, 1)\n\n        self.layer1 = nn.Sequential(nn.Conv2d(256, 256, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 256),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer2 = nn.Sequential(nn.Conv2d(512, 512, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 512),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer3 = nn.Sequential(nn.Conv2d(1024, 512, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 512),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer4 = nn.Sequential(nn.Conv2d(1024, 1024, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 1024),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer5 = nn.Sequential(nn.Conv1d(2304, output_channel, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, output_channel),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n    @staticmethod\n    def get_graph_feature(coor_q, x_q, coor_k, x_k):\n        # coor: bs, 3, np, x: bs, c, np\n\n        k = 4\n        batch_size = x_k.size(0)\n        num_points_k = x_k.size(2)\n        num_points_q = x_q.size(2)\n\n        with torch.no_grad():\n            _, idx = knn(coor_k, coor_q)  # bs k np\n            assert idx.shape[1] == k\n            idx_base = torch.arange(0, batch_size, device=x_q.device).view(-1, 1, 1) * num_points_k\n            idx = idx + idx_base\n            idx = idx.view(-1)\n        num_dims = x_k.size(1)\n        x_k = x_k.transpose(2, 1).contiguous()\n        feature = x_k.view(batch_size * num_points_k, -1)[idx, :]\n        feature = feature.view(batch_size, k, num_points_q, num_dims).permute(0, 3, 2, 1).contiguous()\n        x_q = x_q.view(batch_size, num_dims, num_points_q, 1).expand(-1, -1, -1, k)\n        feature = torch.cat((feature - x_q, x_q), dim=1)\n        return feature\n\n    def forward(self, f, coor):\n        # f: B G C\n        # coor: B G 3\n\n        # bs 3 N   bs C N\n        feature_list = []\n        coor = coor.transpose(1, 2).contiguous()  # B 3 N\n        f = f.transpose(1, 2).contiguous()  # B C N\n        f = self.input_trans(f)  # B 128 N\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 256 N k\n        f = self.layer1(f)  # B 256 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 256 N\n        feature_list.append(f)\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 512 N k\n        f = self.layer2(f)  # B 512 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 512 N\n        feature_list.append(f)\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 1024 N k\n        f = self.layer3(f)  # B 512 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 512 N\n        feature_list.append(f)\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 1024 N k\n        f = self.layer4(f)  # B 1024 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 1024 N\n        feature_list.append(f)\n\n        f = torch.cat(feature_list, dim=1)  # B 2304 N\n\n        f = self.layer5(f)  # B C' N\n\n        f = f.transpose(-1, -2)\n\n        return f\n\n\n### ref https://github.com/Strawberry-Eat-Mango/PCT_Pytorch/blob/main/util.py ###\ndef knn_point(nsample, xyz, new_xyz):\n    \"\"\"\n    Input:\n        nsample: max sample number in local region\n        xyz: all points, [B, N, C]\n        new_xyz: query points, [B, S, C]\n    Return:\n        group_idx: grouped points index, [B, S, nsample]\n    \"\"\"\n    sqrdists = square_distance(new_xyz, xyz)\n    _, group_idx = torch.topk(sqrdists, nsample, dim=-1, largest=False, sorted=False)\n    return group_idx\n\n\ndef square_distance(src, dst):\n    \"\"\"\n    Calculate Euclid distance between each two points.\n    src^T * dst = xn * xm + yn * ym + zn * zm；\n    sum(src^2, dim=-1) = xn*xn + yn*yn + zn*zn;\n    sum(dst^2, dim=-1) = xm*xm + ym*ym + zm*zm;\n    dist = (xn - xm)^2 + (yn - ym)^2 + (zn - zm)^2\n         = sum(src**2,dim=-1)+sum(dst**2,dim=-1)-2*src^T*dst\n    Input:\n        src: source points, [B, N, C]\n        dst: target points, [B, M, C]\n    Output:\n        dist: per-point square distance, [B, N, M]\n    \"\"\"\n    B, N, _ = src.shape\n    _, M, _ = dst.shape\n    dist = -2 * torch.matmul(src, dst.permute(0, 2, 1))\n    dist += torch.sum(src ** 2, -1).view(B, N, 1)\n    dist += torch.sum(dst ** 2, -1).view(B, 1, M)\n    return dist\n\n\nclass Group(nn.Module):\n    def __init__(self, num_group, group_size):\n        super().__init__()\n        self.num_group = num_group\n        self.group_size = group_size\n        # self.knn = KNN(k=self.group_size, transpose_mode=True)\n\n    def forward(self, xyz):\n        '''\n            input: B N 3\n            ---------------------------\n            output: B G M 3\n            center : B G 3\n        '''\n        batch_size, num_points, _ = xyz.shape\n        # fps the centers out\n        center = misc.fps(xyz, self.num_group)  # B G 3\n        # knn to get the neighborhood\n        # _, idx = self.knn(xyz, center) # B G M\n        idx = knn_point(self.group_size, xyz, center)  # B G M\n        assert idx.size(1) == self.num_group\n        assert idx.size(2) == self.group_size\n        idx_base = torch.arange(0, batch_size, device=xyz.device).view(-1, 1, 1) * num_points\n        idx = idx + idx_base\n        idx = idx.view(-1)\n        neighborhood = xyz.view(batch_size * num_points, -1)[idx, :]\n        neighborhood = neighborhood.view(batch_size, self.num_group, self.group_size, 3).contiguous()\n        # normalize\n        neighborhood = neighborhood - center.unsqueeze(2)\n        return neighborhood, center\n\n\nclass Encoder(nn.Module):\n    def __init__(self, encoder_channel):\n        super().__init__()\n        self.encoder_channel = encoder_channel\n        self.first_conv = nn.Sequential(\n            nn.Conv1d(3, 128, 1),\n            nn.BatchNorm1d(128),\n            nn.ReLU(inplace=True),\n            nn.Conv1d(128, 256, 1)\n        )\n        self.second_conv = nn.Sequential(\n            nn.Conv1d(512, 512, 1),\n            nn.BatchNorm1d(512),\n            nn.ReLU(inplace=True),\n            nn.Conv1d(512, self.encoder_channel, 1)\n        )\n\n    def forward(self, point_groups):\n        '''\n            point_groups : B G N 3\n            -----------------\n            feature_global : B G C\n        '''\n        bs, g, n, _ = point_groups.shape\n        point_groups = point_groups.reshape(bs * g, n, 3)\n        # encoder\n        feature = self.first_conv(point_groups.transpose(2, 1))  # BG 256 n\n        feature_global = torch.max(feature, dim=2, keepdim=True)[0]  # BG 256 1\n        feature = torch.cat([feature_global.expand(-1, -1, n), feature], dim=1)  # BG 512 n\n        feature = self.second_conv(feature)  # BG 1024 n\n        feature_global = torch.max(feature, dim=2, keepdim=False)[0]  # BG 1024\n        return feature_global.reshape(bs, g, self.encoder_channel)\n\n\nclass Decoder(nn.Module):\n    def __init__(self, encoder_channel, num_fine):\n        super().__init__()\n        self.num_fine = num_fine\n        self.grid_size = 2\n        self.num_coarse = self.num_fine // 4\n        assert num_fine % 4 == 0\n\n        self.mlp = nn.Sequential(\n            nn.Linear(encoder_channel, 1024),\n            nn.ReLU(inplace=True),\n            nn.Linear(1024, 1024),\n            nn.ReLU(inplace=True),\n            nn.Linear(1024, 3 * self.num_coarse)\n        )\n        self.final_conv = nn.Sequential(\n            nn.Conv1d(encoder_channel + 3 + 2, 512, 1),\n            nn.BatchNorm1d(512),\n            nn.ReLU(inplace=True),\n            nn.Conv1d(512, 512, 1),\n            nn.BatchNorm1d(512),\n            nn.ReLU(inplace=True),\n            nn.Conv1d(512, 3, 1)\n        )\n        a = torch.linspace(-0.05, 0.05, steps=self.grid_size, dtype=torch.float).view(1, self.grid_size).expand(\n            self.grid_size, self.grid_size).reshape(1, -1)\n        b = torch.linspace(-0.05, 0.05, steps=self.grid_size, dtype=torch.float).view(self.grid_size, 1).expand(\n            self.grid_size, self.grid_size).reshape(1, -1)\n        self.folding_seed = torch.cat([a, b], dim=0).view(1, 2, self.grid_size ** 2)  # 1 2 S\n\n    def forward(self, feature_global):\n        '''\n            feature_global : B G C\n            -------\n            coarse : B G M 3\n            fine : B G N 3\n\n        '''\n        bs, g, c = feature_global.shape\n        feature_global = feature_global.reshape(bs * g, c)\n\n        coarse = self.mlp(feature_global).reshape(bs * g, self.num_coarse, 3)  # BG M 3\n\n        point_feat = coarse.unsqueeze(2).expand(-1, -1, self.grid_size ** 2, -1)  # BG (M) S 3\n        point_feat = point_feat.reshape(bs * g, self.num_fine, 3).transpose(2, 1)  # BG 3 N\n\n        seed = self.folding_seed.unsqueeze(2).expand(bs * g, -1, self.num_coarse, -1)  # BG 2 M (S)\n        seed = seed.reshape(bs * g, -1, self.num_fine).to(feature_global.device)  # BG 2 N\n\n        feature_global = feature_global.unsqueeze(2).expand(-1, -1, self.num_fine)  # BG 1024 N\n        feat = torch.cat([feature_global, seed, point_feat], dim=1)  # BG C N\n\n        center = coarse.unsqueeze(2).expand(-1, -1, self.grid_size ** 2, -1)  # BG (M) S 3\n        center = center.reshape(bs * g, self.num_fine, 3).transpose(2, 1)  # BG 3 N\n\n        fine = self.final_conv(feat) + center  # BG 3 N\n        fine = fine.reshape(bs, g, 3, self.num_fine).transpose(-1, -2)\n        coarse = coarse.reshape(bs, g, self.num_coarse, 3)\n        return coarse, fine\n\n\nclass DiscreteVAE(nn.Module):\n    def __init__(self, config, **kwargs):\n        super().__init__()\n        self.group_size = config.group_size\n        self.num_group = config.num_group\n        self.encoder_dims = config.encoder_dims\n        self.tokens_dims = config.tokens_dims\n\n        self.decoder_dims = config.decoder_dims\n        self.num_tokens = config.num_tokens\n\n        self.group_divider = Group(num_group=self.num_group, group_size=self.group_size)\n        self.encoder = Encoder(encoder_channel=self.encoder_dims)\n        self.dgcnn_1 = DGCNN(encoder_channel=self.encoder_dims, output_channel=self.num_tokens)\n        self.codebook = nn.Parameter(torch.randn(self.num_tokens, self.tokens_dims))\n\n        self.dgcnn_2 = DGCNN(encoder_channel=self.tokens_dims, output_channel=self.decoder_dims)\n        self.decoder = Decoder(encoder_channel=self.decoder_dims, num_fine=self.group_size)\n        # self.build_loss_func()\n\n    # def build_loss_func(self):\n    #     self.loss_func_cdl1 = ChamferDistanceL1().cuda()\n    #     self.loss_func_cdl2 = ChamferDistanceL2().cuda()\n    #     self.loss_func_emd = emd().cuda()\n\n    def recon_loss(self, ret, gt):\n        whole_coarse, whole_fine, coarse, fine, group_gt, _ = ret\n\n        bs, g, _, _ = coarse.shape\n\n        coarse = coarse.reshape(bs * g, -1, 3).contiguous()\n        fine = fine.reshape(bs * g, -1, 3).contiguous()\n        group_gt = group_gt.reshape(bs * g, -1, 3).contiguous()\n\n        loss_coarse_block = self.loss_func_cdl1(coarse, group_gt)\n        loss_fine_block = self.loss_func_cdl1(fine, group_gt)\n\n        loss_recon = loss_coarse_block + loss_fine_block\n\n        return loss_recon\n\n    def get_loss(self, ret, gt):\n        # reconstruction loss\n        loss_recon = self.recon_loss(ret, gt)\n        # kl divergence\n        logits = ret[-1]  # B G N\n        softmax = F.softmax(logits, dim=-1)\n        mean_softmax = softmax.mean(dim=1)\n        log_qy = torch.log(mean_softmax)\n        log_uniform = torch.log(torch.tensor([1. / self.num_tokens], device=gt.device))\n        loss_klv = F.kl_div(log_qy, log_uniform.expand(log_qy.size(0), log_qy.size(1)), None, None, 'batchmean',\n                            log_target=True)\n\n        return loss_recon, loss_klv\n\n    def forward(self, inp, temperature=1., hard=False, **kwargs):\n        neighborhood, center = self.group_divider(inp)\n        logits = self.encoder(neighborhood)  # B G C\n        logits = self.dgcnn_1(logits, center)  # B G N\n        soft_one_hot = F.gumbel_softmax(logits, tau=temperature, dim=2, hard=hard)  # B G N\n        sampled = torch.einsum('b g n, n c -> b g c', soft_one_hot, self.codebook)  # B G C\n        feature = self.dgcnn_2(sampled, center)\n        coarse, fine = self.decoder(feature)\n\n        with torch.no_grad():\n            whole_fine = (fine + center.unsqueeze(2)).reshape(inp.size(0), -1, 3)\n            whole_coarse = (coarse + center.unsqueeze(2)).reshape(inp.size(0), -1, 3)\n\n        assert fine.size(2) == self.group_size\n        ret = (whole_coarse, whole_fine, coarse, fine, neighborhood, logits)\n        return ret","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.DGCNN","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.dvae.DGCNN#L11-L103","kind":"class","name":"DGCNN","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":11,"end_line":103,"context_start_line":1,"context_end_line":123,"code":"import torch.nn as nn\nimport torch\nimport torch.nn.functional as F\nfrom . import misc\n\nfrom knn_cuda import KNN\n\nknn = KNN(k=4, transpose_mode=False)\n\n\nclass DGCNN(nn.Module):\n    def __init__(self, encoder_channel, output_channel):\n        super().__init__()\n        '''\n        K has to be 16\n        '''\n        self.input_trans = nn.Conv1d(encoder_channel, 128, 1)\n\n        self.layer1 = nn.Sequential(nn.Conv2d(256, 256, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 256),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer2 = nn.Sequential(nn.Conv2d(512, 512, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 512),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer3 = nn.Sequential(nn.Conv2d(1024, 512, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 512),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer4 = nn.Sequential(nn.Conv2d(1024, 1024, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 1024),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer5 = nn.Sequential(nn.Conv1d(2304, output_channel, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, output_channel),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n    @staticmethod\n    def get_graph_feature(coor_q, x_q, coor_k, x_k):\n        # coor: bs, 3, np, x: bs, c, np\n\n        k = 4\n        batch_size = x_k.size(0)\n        num_points_k = x_k.size(2)\n        num_points_q = x_q.size(2)\n\n        with torch.no_grad():\n            _, idx = knn(coor_k, coor_q)  # bs k np\n            assert idx.shape[1] == k\n            idx_base = torch.arange(0, batch_size, device=x_q.device).view(-1, 1, 1) * num_points_k\n            idx = idx + idx_base\n            idx = idx.view(-1)\n        num_dims = x_k.size(1)\n        x_k = x_k.transpose(2, 1).contiguous()\n        feature = x_k.view(batch_size * num_points_k, -1)[idx, :]\n        feature = feature.view(batch_size, k, num_points_q, num_dims).permute(0, 3, 2, 1).contiguous()\n        x_q = x_q.view(batch_size, num_dims, num_points_q, 1).expand(-1, -1, -1, k)\n        feature = torch.cat((feature - x_q, x_q), dim=1)\n        return feature\n\n    def forward(self, f, coor):\n        # f: B G C\n        # coor: B G 3\n\n        # bs 3 N   bs C N\n        feature_list = []\n        coor = coor.transpose(1, 2).contiguous()  # B 3 N\n        f = f.transpose(1, 2).contiguous()  # B C N\n        f = self.input_trans(f)  # B 128 N\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 256 N k\n        f = self.layer1(f)  # B 256 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 256 N\n        feature_list.append(f)\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 512 N k\n        f = self.layer2(f)  # B 512 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 512 N\n        feature_list.append(f)\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 1024 N k\n        f = self.layer3(f)  # B 512 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 512 N\n        feature_list.append(f)\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 1024 N k\n        f = self.layer4(f)  # B 1024 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 1024 N\n        feature_list.append(f)\n\n        f = torch.cat(feature_list, dim=1)  # B 2304 N\n\n        f = self.layer5(f)  # B C' N\n\n        f = f.transpose(-1, -2)\n\n        return f\n\n\n### ref https://github.com/Strawberry-Eat-Mango/PCT_Pytorch/blob/main/util.py ###\ndef knn_point(nsample, xyz, new_xyz):\n    \"\"\"\n    Input:\n        nsample: max sample number in local region\n        xyz: all points, [B, N, C]\n        new_xyz: query points, [B, S, C]\n    Return:\n        group_idx: grouped points index, [B, S, nsample]\n    \"\"\"\n    sqrdists = square_distance(new_xyz, xyz)\n    _, group_idx = torch.topk(sqrdists, nsample, dim=-1, largest=False, sorted=False)\n    return group_idx\n\n\ndef square_distance(src, dst):\n    \"\"\"\n    Calculate Euclid distance between each two points.","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.knn_point","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.dvae.knn_point#L107-L118","kind":"function","name":"knn_point","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":107,"end_line":118,"context_start_line":87,"context_end_line":138,"code":"        f = self.get_graph_feature(coor, f, coor, f)  # B 1024 N k\n        f = self.layer3(f)  # B 512 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 512 N\n        feature_list.append(f)\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 1024 N k\n        f = self.layer4(f)  # B 1024 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 1024 N\n        feature_list.append(f)\n\n        f = torch.cat(feature_list, dim=1)  # B 2304 N\n\n        f = self.layer5(f)  # B C' N\n\n        f = f.transpose(-1, -2)\n\n        return f\n\n\n### ref https://github.com/Strawberry-Eat-Mango/PCT_Pytorch/blob/main/util.py ###\ndef knn_point(nsample, xyz, new_xyz):\n    \"\"\"\n    Input:\n        nsample: max sample number in local region\n        xyz: all points, [B, N, C]\n        new_xyz: query points, [B, S, C]\n    Return:\n        group_idx: grouped points index, [B, S, nsample]\n    \"\"\"\n    sqrdists = square_distance(new_xyz, xyz)\n    _, group_idx = torch.topk(sqrdists, nsample, dim=-1, largest=False, sorted=False)\n    return group_idx\n\n\ndef square_distance(src, dst):\n    \"\"\"\n    Calculate Euclid distance between each two points.\n    src^T * dst = xn * xm + yn * ym + zn * zm；\n    sum(src^2, dim=-1) = xn*xn + yn*yn + zn*zn;\n    sum(dst^2, dim=-1) = xm*xm + ym*ym + zm*zm;\n    dist = (xn - xm)^2 + (yn - ym)^2 + (zn - zm)^2\n         = sum(src**2,dim=-1)+sum(dst**2,dim=-1)-2*src^T*dst\n    Input:\n        src: source points, [B, N, C]\n        dst: target points, [B, M, C]\n    Output:\n        dist: per-point square distance, [B, N, M]\n    \"\"\"\n    B, N, _ = src.shape\n    _, M, _ = dst.shape\n    dist = -2 * torch.matmul(src, dst.permute(0, 2, 1))\n    dist += torch.sum(src ** 2, -1).view(B, N, 1)","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.square_distance","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.dvae.square_distance#L121-L140","kind":"function","name":"square_distance","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":121,"end_line":140,"context_start_line":101,"context_end_line":160,"code":"        f = f.transpose(-1, -2)\n\n        return f\n\n\n### ref https://github.com/Strawberry-Eat-Mango/PCT_Pytorch/blob/main/util.py ###\ndef knn_point(nsample, xyz, new_xyz):\n    \"\"\"\n    Input:\n        nsample: max sample number in local region\n        xyz: all points, [B, N, C]\n        new_xyz: query points, [B, S, C]\n    Return:\n        group_idx: grouped points index, [B, S, nsample]\n    \"\"\"\n    sqrdists = square_distance(new_xyz, xyz)\n    _, group_idx = torch.topk(sqrdists, nsample, dim=-1, largest=False, sorted=False)\n    return group_idx\n\n\ndef square_distance(src, dst):\n    \"\"\"\n    Calculate Euclid distance between each two points.\n    src^T * dst = xn * xm + yn * ym + zn * zm；\n    sum(src^2, dim=-1) = xn*xn + yn*yn + zn*zn;\n    sum(dst^2, dim=-1) = xm*xm + ym*ym + zm*zm;\n    dist = (xn - xm)^2 + (yn - ym)^2 + (zn - zm)^2\n         = sum(src**2,dim=-1)+sum(dst**2,dim=-1)-2*src^T*dst\n    Input:\n        src: source points, [B, N, C]\n        dst: target points, [B, M, C]\n    Output:\n        dist: per-point square distance, [B, N, M]\n    \"\"\"\n    B, N, _ = src.shape\n    _, M, _ = dst.shape\n    dist = -2 * torch.matmul(src, dst.permute(0, 2, 1))\n    dist += torch.sum(src ** 2, -1).view(B, N, 1)\n    dist += torch.sum(dst ** 2, -1).view(B, 1, M)\n    return dist\n\n\nclass Group(nn.Module):\n    def __init__(self, num_group, group_size):\n        super().__init__()\n        self.num_group = num_group\n        self.group_size = group_size\n        # self.knn = KNN(k=self.group_size, transpose_mode=True)\n\n    def forward(self, xyz):\n        '''\n            input: B N 3\n            ---------------------------\n            output: B G M 3\n            center : B G 3\n        '''\n        batch_size, num_points, _ = xyz.shape\n        # fps the centers out\n        center = misc.fps(xyz, self.num_group)  # B G 3\n        # knn to get the neighborhood","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.Group","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.dvae.Group#L143-L172","kind":"class","name":"Group","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":143,"end_line":172,"context_start_line":123,"context_end_line":192,"code":"    Calculate Euclid distance between each two points.\n    src^T * dst = xn * xm + yn * ym + zn * zm；\n    sum(src^2, dim=-1) = xn*xn + yn*yn + zn*zn;\n    sum(dst^2, dim=-1) = xm*xm + ym*ym + zm*zm;\n    dist = (xn - xm)^2 + (yn - ym)^2 + (zn - zm)^2\n         = sum(src**2,dim=-1)+sum(dst**2,dim=-1)-2*src^T*dst\n    Input:\n        src: source points, [B, N, C]\n        dst: target points, [B, M, C]\n    Output:\n        dist: per-point square distance, [B, N, M]\n    \"\"\"\n    B, N, _ = src.shape\n    _, M, _ = dst.shape\n    dist = -2 * torch.matmul(src, dst.permute(0, 2, 1))\n    dist += torch.sum(src ** 2, -1).view(B, N, 1)\n    dist += torch.sum(dst ** 2, -1).view(B, 1, M)\n    return dist\n\n\nclass Group(nn.Module):\n    def __init__(self, num_group, group_size):\n        super().__init__()\n        self.num_group = num_group\n        self.group_size = group_size\n        # self.knn = KNN(k=self.group_size, transpose_mode=True)\n\n    def forward(self, xyz):\n        '''\n            input: B N 3\n            ---------------------------\n            output: B G M 3\n            center : B G 3\n        '''\n        batch_size, num_points, _ = xyz.shape\n        # fps the centers out\n        center = misc.fps(xyz, self.num_group)  # B G 3\n        # knn to get the neighborhood\n        # _, idx = self.knn(xyz, center) # B G M\n        idx = knn_point(self.group_size, xyz, center)  # B G M\n        assert idx.size(1) == self.num_group\n        assert idx.size(2) == self.group_size\n        idx_base = torch.arange(0, batch_size, device=xyz.device).view(-1, 1, 1) * num_points\n        idx = idx + idx_base\n        idx = idx.view(-1)\n        neighborhood = xyz.view(batch_size * num_points, -1)[idx, :]\n        neighborhood = neighborhood.view(batch_size, self.num_group, self.group_size, 3).contiguous()\n        # normalize\n        neighborhood = neighborhood - center.unsqueeze(2)\n        return neighborhood, center\n\n\nclass Encoder(nn.Module):\n    def __init__(self, encoder_channel):\n        super().__init__()\n        self.encoder_channel = encoder_channel\n        self.first_conv = nn.Sequential(\n            nn.Conv1d(3, 128, 1),\n            nn.BatchNorm1d(128),\n            nn.ReLU(inplace=True),\n            nn.Conv1d(128, 256, 1)\n        )\n        self.second_conv = nn.Sequential(\n            nn.Conv1d(512, 512, 1),\n            nn.BatchNorm1d(512),\n            nn.ReLU(inplace=True),\n            nn.Conv1d(512, self.encoder_channel, 1)\n        )\n\n    def forward(self, point_groups):","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.Encoder","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.dvae.Encoder#L175-L206","kind":"class","name":"Encoder","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":175,"end_line":206,"context_start_line":155,"context_end_line":226,"code":"            center : B G 3\n        '''\n        batch_size, num_points, _ = xyz.shape\n        # fps the centers out\n        center = misc.fps(xyz, self.num_group)  # B G 3\n        # knn to get the neighborhood\n        # _, idx = self.knn(xyz, center) # B G M\n        idx = knn_point(self.group_size, xyz, center)  # B G M\n        assert idx.size(1) == self.num_group\n        assert idx.size(2) == self.group_size\n        idx_base = torch.arange(0, batch_size, device=xyz.device).view(-1, 1, 1) * num_points\n        idx = idx + idx_base\n        idx = idx.view(-1)\n        neighborhood = xyz.view(batch_size * num_points, -1)[idx, :]\n        neighborhood = neighborhood.view(batch_size, self.num_group, self.group_size, 3).contiguous()\n        # normalize\n        neighborhood = neighborhood - center.unsqueeze(2)\n        return neighborhood, center\n\n\nclass Encoder(nn.Module):\n    def __init__(self, encoder_channel):\n        super().__init__()\n        self.encoder_channel = encoder_channel\n        self.first_conv = nn.Sequential(\n            nn.Conv1d(3, 128, 1),\n            nn.BatchNorm1d(128),\n            nn.ReLU(inplace=True),\n            nn.Conv1d(128, 256, 1)\n        )\n        self.second_conv = nn.Sequential(\n            nn.Conv1d(512, 512, 1),\n            nn.BatchNorm1d(512),\n            nn.ReLU(inplace=True),\n            nn.Conv1d(512, self.encoder_channel, 1)\n        )\n\n    def forward(self, point_groups):\n        '''\n            point_groups : B G N 3\n            -----------------\n            feature_global : B G C\n        '''\n        bs, g, n, _ = point_groups.shape\n        point_groups = point_groups.reshape(bs * g, n, 3)\n        # encoder\n        feature = self.first_conv(point_groups.transpose(2, 1))  # BG 256 n\n        feature_global = torch.max(feature, dim=2, keepdim=True)[0]  # BG 256 1\n        feature = torch.cat([feature_global.expand(-1, -1, n), feature], dim=1)  # BG 512 n\n        feature = self.second_conv(feature)  # BG 1024 n\n        feature_global = torch.max(feature, dim=2, keepdim=False)[0]  # BG 1024\n        return feature_global.reshape(bs, g, self.encoder_channel)\n\n\nclass Decoder(nn.Module):\n    def __init__(self, encoder_channel, num_fine):\n        super().__init__()\n        self.num_fine = num_fine\n        self.grid_size = 2\n        self.num_coarse = self.num_fine // 4\n        assert num_fine % 4 == 0\n\n        self.mlp = nn.Sequential(\n            nn.Linear(encoder_channel, 1024),\n            nn.ReLU(inplace=True),\n            nn.Linear(1024, 1024),\n            nn.ReLU(inplace=True),\n            nn.Linear(1024, 3 * self.num_coarse)\n        )\n        self.final_conv = nn.Sequential(\n            nn.Conv1d(encoder_channel + 3 + 2, 512, 1),\n            nn.BatchNorm1d(512),","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.Decoder","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.dvae.Decoder#L209-L267","kind":"class","name":"Decoder","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":209,"end_line":267,"context_start_line":189,"context_end_line":287,"code":"            nn.Conv1d(512, self.encoder_channel, 1)\n        )\n\n    def forward(self, point_groups):\n        '''\n            point_groups : B G N 3\n            -----------------\n            feature_global : B G C\n        '''\n        bs, g, n, _ = point_groups.shape\n        point_groups = point_groups.reshape(bs * g, n, 3)\n        # encoder\n        feature = self.first_conv(point_groups.transpose(2, 1))  # BG 256 n\n        feature_global = torch.max(feature, dim=2, keepdim=True)[0]  # BG 256 1\n        feature = torch.cat([feature_global.expand(-1, -1, n), feature], dim=1)  # BG 512 n\n        feature = self.second_conv(feature)  # BG 1024 n\n        feature_global = torch.max(feature, dim=2, keepdim=False)[0]  # BG 1024\n        return feature_global.reshape(bs, g, self.encoder_channel)\n\n\nclass Decoder(nn.Module):\n    def __init__(self, encoder_channel, num_fine):\n        super().__init__()\n        self.num_fine = num_fine\n        self.grid_size = 2\n        self.num_coarse = self.num_fine // 4\n        assert num_fine % 4 == 0\n\n        self.mlp = nn.Sequential(\n            nn.Linear(encoder_channel, 1024),\n            nn.ReLU(inplace=True),\n            nn.Linear(1024, 1024),\n            nn.ReLU(inplace=True),\n            nn.Linear(1024, 3 * self.num_coarse)\n        )\n        self.final_conv = nn.Sequential(\n            nn.Conv1d(encoder_channel + 3 + 2, 512, 1),\n            nn.BatchNorm1d(512),\n            nn.ReLU(inplace=True),\n            nn.Conv1d(512, 512, 1),\n            nn.BatchNorm1d(512),\n            nn.ReLU(inplace=True),\n            nn.Conv1d(512, 3, 1)\n        )\n        a = torch.linspace(-0.05, 0.05, steps=self.grid_size, dtype=torch.float).view(1, self.grid_size).expand(\n            self.grid_size, self.grid_size).reshape(1, -1)\n        b = torch.linspace(-0.05, 0.05, steps=self.grid_size, dtype=torch.float).view(self.grid_size, 1).expand(\n            self.grid_size, self.grid_size).reshape(1, -1)\n        self.folding_seed = torch.cat([a, b], dim=0).view(1, 2, self.grid_size ** 2)  # 1 2 S\n\n    def forward(self, feature_global):\n        '''\n            feature_global : B G C\n            -------\n            coarse : B G M 3\n            fine : B G N 3\n\n        '''\n        bs, g, c = feature_global.shape\n        feature_global = feature_global.reshape(bs * g, c)\n\n        coarse = self.mlp(feature_global).reshape(bs * g, self.num_coarse, 3)  # BG M 3\n\n        point_feat = coarse.unsqueeze(2).expand(-1, -1, self.grid_size ** 2, -1)  # BG (M) S 3\n        point_feat = point_feat.reshape(bs * g, self.num_fine, 3).transpose(2, 1)  # BG 3 N\n\n        seed = self.folding_seed.unsqueeze(2).expand(bs * g, -1, self.num_coarse, -1)  # BG 2 M (S)\n        seed = seed.reshape(bs * g, -1, self.num_fine).to(feature_global.device)  # BG 2 N\n\n        feature_global = feature_global.unsqueeze(2).expand(-1, -1, self.num_fine)  # BG 1024 N\n        feat = torch.cat([feature_global, seed, point_feat], dim=1)  # BG C N\n\n        center = coarse.unsqueeze(2).expand(-1, -1, self.grid_size ** 2, -1)  # BG (M) S 3\n        center = center.reshape(bs * g, self.num_fine, 3).transpose(2, 1)  # BG 3 N\n\n        fine = self.final_conv(feat) + center  # BG 3 N\n        fine = fine.reshape(bs, g, 3, self.num_fine).transpose(-1, -2)\n        coarse = coarse.reshape(bs, g, self.num_coarse, 3)\n        return coarse, fine\n\n\nclass DiscreteVAE(nn.Module):\n    def __init__(self, config, **kwargs):\n        super().__init__()\n        self.group_size = config.group_size\n        self.num_group = config.num_group\n        self.encoder_dims = config.encoder_dims\n        self.tokens_dims = config.tokens_dims\n\n        self.decoder_dims = config.decoder_dims\n        self.num_tokens = config.num_tokens\n\n        self.group_divider = Group(num_group=self.num_group, group_size=self.group_size)\n        self.encoder = Encoder(encoder_channel=self.encoder_dims)\n        self.dgcnn_1 = DGCNN(encoder_channel=self.encoder_dims, output_channel=self.num_tokens)\n        self.codebook = nn.Parameter(torch.randn(self.num_tokens, self.tokens_dims))\n\n        self.dgcnn_2 = DGCNN(encoder_channel=self.tokens_dims, output_channel=self.decoder_dims)\n        self.decoder = Decoder(encoder_channel=self.decoder_dims, num_fine=self.group_size)","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.DiscreteVAE","uri":"program://LLaMA-Adapter/class/imagebind_LLM.ImageBind.models.pointbert.dvae.DiscreteVAE#L270-L340","kind":"class","name":"DiscreteVAE","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":270,"end_line":340,"context_start_line":250,"context_end_line":340,"code":"        coarse = self.mlp(feature_global).reshape(bs * g, self.num_coarse, 3)  # BG M 3\n\n        point_feat = coarse.unsqueeze(2).expand(-1, -1, self.grid_size ** 2, -1)  # BG (M) S 3\n        point_feat = point_feat.reshape(bs * g, self.num_fine, 3).transpose(2, 1)  # BG 3 N\n\n        seed = self.folding_seed.unsqueeze(2).expand(bs * g, -1, self.num_coarse, -1)  # BG 2 M (S)\n        seed = seed.reshape(bs * g, -1, self.num_fine).to(feature_global.device)  # BG 2 N\n\n        feature_global = feature_global.unsqueeze(2).expand(-1, -1, self.num_fine)  # BG 1024 N\n        feat = torch.cat([feature_global, seed, point_feat], dim=1)  # BG C N\n\n        center = coarse.unsqueeze(2).expand(-1, -1, self.grid_size ** 2, -1)  # BG (M) S 3\n        center = center.reshape(bs * g, self.num_fine, 3).transpose(2, 1)  # BG 3 N\n\n        fine = self.final_conv(feat) + center  # BG 3 N\n        fine = fine.reshape(bs, g, 3, self.num_fine).transpose(-1, -2)\n        coarse = coarse.reshape(bs, g, self.num_coarse, 3)\n        return coarse, fine\n\n\nclass DiscreteVAE(nn.Module):\n    def __init__(self, config, **kwargs):\n        super().__init__()\n        self.group_size = config.group_size\n        self.num_group = config.num_group\n        self.encoder_dims = config.encoder_dims\n        self.tokens_dims = config.tokens_dims\n\n        self.decoder_dims = config.decoder_dims\n        self.num_tokens = config.num_tokens\n\n        self.group_divider = Group(num_group=self.num_group, group_size=self.group_size)\n        self.encoder = Encoder(encoder_channel=self.encoder_dims)\n        self.dgcnn_1 = DGCNN(encoder_channel=self.encoder_dims, output_channel=self.num_tokens)\n        self.codebook = nn.Parameter(torch.randn(self.num_tokens, self.tokens_dims))\n\n        self.dgcnn_2 = DGCNN(encoder_channel=self.tokens_dims, output_channel=self.decoder_dims)\n        self.decoder = Decoder(encoder_channel=self.decoder_dims, num_fine=self.group_size)\n        # self.build_loss_func()\n\n    # def build_loss_func(self):\n    #     self.loss_func_cdl1 = ChamferDistanceL1().cuda()\n    #     self.loss_func_cdl2 = ChamferDistanceL2().cuda()\n    #     self.loss_func_emd = emd().cuda()\n\n    def recon_loss(self, ret, gt):\n        whole_coarse, whole_fine, coarse, fine, group_gt, _ = ret\n\n        bs, g, _, _ = coarse.shape\n\n        coarse = coarse.reshape(bs * g, -1, 3).contiguous()\n        fine = fine.reshape(bs * g, -1, 3).contiguous()\n        group_gt = group_gt.reshape(bs * g, -1, 3).contiguous()\n\n        loss_coarse_block = self.loss_func_cdl1(coarse, group_gt)\n        loss_fine_block = self.loss_func_cdl1(fine, group_gt)\n\n        loss_recon = loss_coarse_block + loss_fine_block\n\n        return loss_recon\n\n    def get_loss(self, ret, gt):\n        # reconstruction loss\n        loss_recon = self.recon_loss(ret, gt)\n        # kl divergence\n        logits = ret[-1]  # B G N\n        softmax = F.softmax(logits, dim=-1)\n        mean_softmax = softmax.mean(dim=1)\n        log_qy = torch.log(mean_softmax)\n        log_uniform = torch.log(torch.tensor([1. / self.num_tokens], device=gt.device))\n        loss_klv = F.kl_div(log_qy, log_uniform.expand(log_qy.size(0), log_qy.size(1)), None, None, 'batchmean',\n                            log_target=True)\n\n        return loss_recon, loss_klv\n\n    def forward(self, inp, temperature=1., hard=False, **kwargs):\n        neighborhood, center = self.group_divider(inp)\n        logits = self.encoder(neighborhood)  # B G C\n        logits = self.dgcnn_1(logits, center)  # B G N\n        soft_one_hot = F.gumbel_softmax(logits, tau=temperature, dim=2, hard=hard)  # B G N\n        sampled = torch.einsum('b g n, n c -> b g c', soft_one_hot, self.codebook)  # B G C\n        feature = self.dgcnn_2(sampled, center)\n        coarse, fine = self.decoder(feature)\n\n        with torch.no_grad():\n            whole_fine = (fine + center.unsqueeze(2)).reshape(inp.size(0), -1, 3)\n            whole_coarse = (coarse + center.unsqueeze(2)).reshape(inp.size(0), -1, 3)\n\n        assert fine.size(2) == self.group_size\n        ret = (whole_coarse, whole_fine, coarse, fine, neighborhood, logits)\n        return ret","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.dvae.__init__#L271-L287","kind":"function","name":"__init__","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":271,"end_line":287,"context_start_line":251,"context_end_line":307,"code":"\n        point_feat = coarse.unsqueeze(2).expand(-1, -1, self.grid_size ** 2, -1)  # BG (M) S 3\n        point_feat = point_feat.reshape(bs * g, self.num_fine, 3).transpose(2, 1)  # BG 3 N\n\n        seed = self.folding_seed.unsqueeze(2).expand(bs * g, -1, self.num_coarse, -1)  # BG 2 M (S)\n        seed = seed.reshape(bs * g, -1, self.num_fine).to(feature_global.device)  # BG 2 N\n\n        feature_global = feature_global.unsqueeze(2).expand(-1, -1, self.num_fine)  # BG 1024 N\n        feat = torch.cat([feature_global, seed, point_feat], dim=1)  # BG C N\n\n        center = coarse.unsqueeze(2).expand(-1, -1, self.grid_size ** 2, -1)  # BG (M) S 3\n        center = center.reshape(bs * g, self.num_fine, 3).transpose(2, 1)  # BG 3 N\n\n        fine = self.final_conv(feat) + center  # BG 3 N\n        fine = fine.reshape(bs, g, 3, self.num_fine).transpose(-1, -2)\n        coarse = coarse.reshape(bs, g, self.num_coarse, 3)\n        return coarse, fine\n\n\nclass DiscreteVAE(nn.Module):\n    def __init__(self, config, **kwargs):\n        super().__init__()\n        self.group_size = config.group_size\n        self.num_group = config.num_group\n        self.encoder_dims = config.encoder_dims\n        self.tokens_dims = config.tokens_dims\n\n        self.decoder_dims = config.decoder_dims\n        self.num_tokens = config.num_tokens\n\n        self.group_divider = Group(num_group=self.num_group, group_size=self.group_size)\n        self.encoder = Encoder(encoder_channel=self.encoder_dims)\n        self.dgcnn_1 = DGCNN(encoder_channel=self.encoder_dims, output_channel=self.num_tokens)\n        self.codebook = nn.Parameter(torch.randn(self.num_tokens, self.tokens_dims))\n\n        self.dgcnn_2 = DGCNN(encoder_channel=self.tokens_dims, output_channel=self.decoder_dims)\n        self.decoder = Decoder(encoder_channel=self.decoder_dims, num_fine=self.group_size)\n        # self.build_loss_func()\n\n    # def build_loss_func(self):\n    #     self.loss_func_cdl1 = ChamferDistanceL1().cuda()\n    #     self.loss_func_cdl2 = ChamferDistanceL2().cuda()\n    #     self.loss_func_emd = emd().cuda()\n\n    def recon_loss(self, ret, gt):\n        whole_coarse, whole_fine, coarse, fine, group_gt, _ = ret\n\n        bs, g, _, _ = coarse.shape\n\n        coarse = coarse.reshape(bs * g, -1, 3).contiguous()\n        fine = fine.reshape(bs * g, -1, 3).contiguous()\n        group_gt = group_gt.reshape(bs * g, -1, 3).contiguous()\n\n        loss_coarse_block = self.loss_func_cdl1(coarse, group_gt)\n        loss_fine_block = self.loss_func_cdl1(fine, group_gt)\n\n        loss_recon = loss_coarse_block + loss_fine_block","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.get_graph_feature","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.dvae.get_graph_feature#L45-L65","kind":"function","name":"get_graph_feature","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":45,"end_line":65,"context_start_line":25,"context_end_line":85,"code":"                                    nn.GroupNorm(4, 512),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer3 = nn.Sequential(nn.Conv2d(1024, 512, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 512),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer4 = nn.Sequential(nn.Conv2d(1024, 1024, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 1024),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n        self.layer5 = nn.Sequential(nn.Conv1d(2304, output_channel, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, output_channel),\n                                    nn.LeakyReLU(negative_slope=0.2)\n                                    )\n\n    @staticmethod\n    def get_graph_feature(coor_q, x_q, coor_k, x_k):\n        # coor: bs, 3, np, x: bs, c, np\n\n        k = 4\n        batch_size = x_k.size(0)\n        num_points_k = x_k.size(2)\n        num_points_q = x_q.size(2)\n\n        with torch.no_grad():\n            _, idx = knn(coor_k, coor_q)  # bs k np\n            assert idx.shape[1] == k\n            idx_base = torch.arange(0, batch_size, device=x_q.device).view(-1, 1, 1) * num_points_k\n            idx = idx + idx_base\n            idx = idx.view(-1)\n        num_dims = x_k.size(1)\n        x_k = x_k.transpose(2, 1).contiguous()\n        feature = x_k.view(batch_size * num_points_k, -1)[idx, :]\n        feature = feature.view(batch_size, k, num_points_q, num_dims).permute(0, 3, 2, 1).contiguous()\n        x_q = x_q.view(batch_size, num_dims, num_points_q, 1).expand(-1, -1, -1, k)\n        feature = torch.cat((feature - x_q, x_q), dim=1)\n        return feature\n\n    def forward(self, f, coor):\n        # f: B G C\n        # coor: B G 3\n\n        # bs 3 N   bs C N\n        feature_list = []\n        coor = coor.transpose(1, 2).contiguous()  # B 3 N\n        f = f.transpose(1, 2).contiguous()  # B C N\n        f = self.input_trans(f)  # B 128 N\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 256 N k\n        f = self.layer1(f)  # B 256 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 256 N\n        feature_list.append(f)\n\n        f = self.get_graph_feature(coor, f, coor, f)  # B 512 N k\n        f = self.layer2(f)  # B 512 N k\n        f = f.max(dim=-1, keepdim=False)[0]  # B 512 N\n        feature_list.append(f)","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.forward","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.dvae.forward#L325-L340","kind":"function","name":"forward","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":325,"end_line":340,"context_start_line":305,"context_end_line":340,"code":"        loss_fine_block = self.loss_func_cdl1(fine, group_gt)\n\n        loss_recon = loss_coarse_block + loss_fine_block\n\n        return loss_recon\n\n    def get_loss(self, ret, gt):\n        # reconstruction loss\n        loss_recon = self.recon_loss(ret, gt)\n        # kl divergence\n        logits = ret[-1]  # B G N\n        softmax = F.softmax(logits, dim=-1)\n        mean_softmax = softmax.mean(dim=1)\n        log_qy = torch.log(mean_softmax)\n        log_uniform = torch.log(torch.tensor([1. / self.num_tokens], device=gt.device))\n        loss_klv = F.kl_div(log_qy, log_uniform.expand(log_qy.size(0), log_qy.size(1)), None, None, 'batchmean',\n                            log_target=True)\n\n        return loss_recon, loss_klv\n\n    def forward(self, inp, temperature=1., hard=False, **kwargs):\n        neighborhood, center = self.group_divider(inp)\n        logits = self.encoder(neighborhood)  # B G C\n        logits = self.dgcnn_1(logits, center)  # B G N\n        soft_one_hot = F.gumbel_softmax(logits, tau=temperature, dim=2, hard=hard)  # B G N\n        sampled = torch.einsum('b g n, n c -> b g c', soft_one_hot, self.codebook)  # B G C\n        feature = self.dgcnn_2(sampled, center)\n        coarse, fine = self.decoder(feature)\n\n        with torch.no_grad():\n            whole_fine = (fine + center.unsqueeze(2)).reshape(inp.size(0), -1, 3)\n            whole_coarse = (coarse + center.unsqueeze(2)).reshape(inp.size(0), -1, 3)\n\n        assert fine.size(2) == self.group_size\n        ret = (whole_coarse, whole_fine, coarse, fine, neighborhood, logits)\n        return ret","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.recon_loss","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.dvae.recon_loss#L295-L309","kind":"function","name":"recon_loss","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":295,"end_line":309,"context_start_line":275,"context_end_line":329,"code":"        self.encoder_dims = config.encoder_dims\n        self.tokens_dims = config.tokens_dims\n\n        self.decoder_dims = config.decoder_dims\n        self.num_tokens = config.num_tokens\n\n        self.group_divider = Group(num_group=self.num_group, group_size=self.group_size)\n        self.encoder = Encoder(encoder_channel=self.encoder_dims)\n        self.dgcnn_1 = DGCNN(encoder_channel=self.encoder_dims, output_channel=self.num_tokens)\n        self.codebook = nn.Parameter(torch.randn(self.num_tokens, self.tokens_dims))\n\n        self.dgcnn_2 = DGCNN(encoder_channel=self.tokens_dims, output_channel=self.decoder_dims)\n        self.decoder = Decoder(encoder_channel=self.decoder_dims, num_fine=self.group_size)\n        # self.build_loss_func()\n\n    # def build_loss_func(self):\n    #     self.loss_func_cdl1 = ChamferDistanceL1().cuda()\n    #     self.loss_func_cdl2 = ChamferDistanceL2().cuda()\n    #     self.loss_func_emd = emd().cuda()\n\n    def recon_loss(self, ret, gt):\n        whole_coarse, whole_fine, coarse, fine, group_gt, _ = ret\n\n        bs, g, _, _ = coarse.shape\n\n        coarse = coarse.reshape(bs * g, -1, 3).contiguous()\n        fine = fine.reshape(bs * g, -1, 3).contiguous()\n        group_gt = group_gt.reshape(bs * g, -1, 3).contiguous()\n\n        loss_coarse_block = self.loss_func_cdl1(coarse, group_gt)\n        loss_fine_block = self.loss_func_cdl1(fine, group_gt)\n\n        loss_recon = loss_coarse_block + loss_fine_block\n\n        return loss_recon\n\n    def get_loss(self, ret, gt):\n        # reconstruction loss\n        loss_recon = self.recon_loss(ret, gt)\n        # kl divergence\n        logits = ret[-1]  # B G N\n        softmax = F.softmax(logits, dim=-1)\n        mean_softmax = softmax.mean(dim=1)\n        log_qy = torch.log(mean_softmax)\n        log_uniform = torch.log(torch.tensor([1. / self.num_tokens], device=gt.device))\n        loss_klv = F.kl_div(log_qy, log_uniform.expand(log_qy.size(0), log_qy.size(1)), None, None, 'batchmean',\n                            log_target=True)\n\n        return loss_recon, loss_klv\n\n    def forward(self, inp, temperature=1., hard=False, **kwargs):\n        neighborhood, center = self.group_divider(inp)\n        logits = self.encoder(neighborhood)  # B G C\n        logits = self.dgcnn_1(logits, center)  # B G N\n        soft_one_hot = F.gumbel_softmax(logits, tau=temperature, dim=2, hard=hard)  # B G N","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.ImageBind.models.pointbert.dvae.get_loss","uri":"program://LLaMA-Adapter/function/imagebind_LLM.ImageBind.models.pointbert.dvae.get_loss#L311-L323","kind":"function","name":"get_loss","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":311,"end_line":323,"context_start_line":291,"context_end_line":340,"code":"    #     self.loss_func_cdl1 = ChamferDistanceL1().cuda()\n    #     self.loss_func_cdl2 = ChamferDistanceL2().cuda()\n    #     self.loss_func_emd = emd().cuda()\n\n    def recon_loss(self, ret, gt):\n        whole_coarse, whole_fine, coarse, fine, group_gt, _ = ret\n\n        bs, g, _, _ = coarse.shape\n\n        coarse = coarse.reshape(bs * g, -1, 3).contiguous()\n        fine = fine.reshape(bs * g, -1, 3).contiguous()\n        group_gt = group_gt.reshape(bs * g, -1, 3).contiguous()\n\n        loss_coarse_block = self.loss_func_cdl1(coarse, group_gt)\n        loss_fine_block = self.loss_func_cdl1(fine, group_gt)\n\n        loss_recon = loss_coarse_block + loss_fine_block\n\n        return loss_recon\n\n    def get_loss(self, ret, gt):\n        # reconstruction loss\n        loss_recon = self.recon_loss(ret, gt)\n        # kl divergence\n        logits = ret[-1]  # B G N\n        softmax = F.softmax(logits, dim=-1)\n        mean_softmax = softmax.mean(dim=1)\n        log_qy = torch.log(mean_softmax)\n        log_uniform = torch.log(torch.tensor([1. / self.num_tokens], device=gt.device))\n        loss_klv = F.kl_div(log_qy, log_uniform.expand(log_qy.size(0), log_qy.size(1)), None, None, 'batchmean',\n                            log_target=True)\n\n        return loss_recon, loss_klv\n\n    def forward(self, inp, temperature=1., hard=False, **kwargs):\n        neighborhood, center = self.group_divider(inp)\n        logits = self.encoder(neighborhood)  # B G C\n        logits = self.dgcnn_1(logits, center)  # B G N\n        soft_one_hot = F.gumbel_softmax(logits, tau=temperature, dim=2, hard=hard)  # B G N\n        sampled = torch.einsum('b g n, n c -> b g c', soft_one_hot, self.codebook)  # B G C\n        feature = self.dgcnn_2(sampled, center)\n        coarse, fine = self.decoder(feature)\n\n        with torch.no_grad():\n            whole_fine = (fine + center.unsqueeze(2)).reshape(inp.size(0), -1, 3)\n            whole_coarse = (coarse + center.unsqueeze(2)).reshape(inp.size(0), -1, 3)\n\n        assert fine.size(2) == self.group_size\n        ret = (whole_coarse, whole_fine, coarse, fine, neighborhood, logits)\n        return ret","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter","uri":"program://LLaMA-Adapter/module/imagebind_LLM.llama.llama_adapter#L1-L319","kind":"module","name":"imagebind_LLM.llama.llama_adapter","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":1,"end_line":319,"context_start_line":1,"context_end_line":319,"code":"import json\nimport os\nfrom pathlib import Path\nimport numpy as np\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\n\nfrom .llama import Transformer, ModelArgs, RMSNorm\nfrom .tokenizer import Tokenizer\nfrom util.misc import download\nfrom .utils import sample_top_p\n\nfrom ImageBind.models import imagebind_model\n\n\nclass LLaMA_adapter(nn.Module):\n    \"\"\" Masked Autoencoder with VisionTransformer backbone\n    \"\"\"\n    def __init__(self, llama_ckpt_dir, llama_tokenizer, knn=False, phase=\"finetune\", legacy_bridge=False):\n        super().__init__()\n\n        # 1. imagebind and imagebind projector\n        self.image_bind = imagebind_model.imagebind_huge(pretrained=True)\n\n        self.image_bind_proj = nn.Linear(1024, 4096)\n\n        if legacy_bridge:\n            bridge_norm_layer = nn.LayerNorm\n            bridge_bias = True\n        else:\n            bridge_norm_layer = RMSNorm\n            bridge_bias = False\n\n\n        self.image_bind_norm_1 = bridge_norm_layer(4096)\n        self.image_bind_f1_1 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n        self.image_bind_f2_1 = nn.Linear(4096 * 4, 4096, bias=bridge_bias)\n        self.image_bind_f3_1 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n\n        self.image_bind_norm_2 = bridge_norm_layer(4096)\n        self.image_bind_f1_2 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n        self.image_bind_f2_2 = nn.Linear(4096 * 4, 4096, bias=bridge_bias)\n        self.image_bind_f3_2 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n\n        self.image_bind_norm_3 = bridge_norm_layer(4096)\n        self.image_bind_f1_3 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n        self.image_bind_f2_3 = nn.Linear(4096 * 4, 4096, bias=bridge_bias)\n        self.image_bind_f3_3 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n\n        # 2. tokenizer\n        self.tokenizer = Tokenizer(model_path=llama_tokenizer)\n\n        # 3. llama\n        with open(os.path.join(llama_ckpt_dir, \"params.json\"), \"r\") as f:\n            params = json.loads(f.read())\n        bias_lora = phase == \"finetune\"\n        model_args: ModelArgs = ModelArgs(\n            max_seq_len=512, max_batch_size=1, w_bias=bias_lora, w_lora=bias_lora,  **params\n        ) # max_batch_size only affects inference\n        print(f\"model args: {model_args}\")\n        model_args.vocab_size = self.tokenizer.n_words\n        torch.set_default_tensor_type(torch.cuda.HalfTensor)\n        self.llama = Transformer(model_args)\n        torch.set_default_tensor_type(torch.FloatTensor)\n\n        ckpts = sorted(Path(llama_ckpt_dir).glob(\"*.pth\"))\n        for ckpt in ckpts:\n            ckpt = torch.load(ckpt, map_location='cpu')\n            self.llama.load_state_dict(ckpt, strict=False)\n\n        # 4. prefix\n        self.query_layer = 32\n        self.query_len = 1\n        self.prefix_query = nn.Embedding(self.query_layer * self.query_len, model_args.dim)\n\n        # 5. knn\n        self.knn = knn\n        if knn:\n            import faiss\n            self.index = faiss.read_index(download(\"https://huggingface.co/csuhan/knn/resolve/main/knn.index\", \"ckpts\"))\n\n        # 6. training criterion\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n\n        self.phase = phase\n        self.set_default_trainability(self.phase)\n\n    def get_trainable_params(self, phase='finetune'):\n        trainable = {}\n        if phase == 'finetune':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'norm' in name or 'bias' in name or 'lora' in name:\n                        trainable[name] = para\n        elif phase == 'pretrain':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'gate' in name:\n                        trainable[name] = para\n                elif name.startswith(\"image_bind_\"):  # not 'image_bind.' so image_bind won't be trained.\n                    trainable[name] = para\n                elif name.startswith(\"prefix_query.\"):\n                    trainable[name] = para\n        else:\n            raise ValueError(f\"Unknown model phase: {phase}\")\n        return trainable\n\n    def set_default_trainability(self, phase='finetune'):\n        for key, value in self.named_parameters():\n            value.requires_grad = False\n        for key, value in self.get_trainable_params(phase).items():\n            value.data = value.data.float()\n            value.requires_grad = True\n\n    def forward_visual(self, inputs, cache_size=10, cache_t=20, cache_weight=0.5):\n        outputs = []\n        outputs_weights = []\n        for input_type, (input, input_weight) in inputs.items():\n            if input_type in ['Image', 'Video']:\n                type = 'vision'\n            else:\n                type = input_type.lower()\n            outputs.append(F.normalize(self.image_bind({type : input})[type], dim=-1))\n            outputs_weights.append(input_weight)\n        outputs_weights = [x/(sum(outputs_weights)+1e-6) for x in outputs_weights]\n\n        visual_feats = sum([output*output_weight for output, output_weight in zip(outputs, outputs_weights)])\n        device = visual_feats.device\n\n        if self.knn:\n            visual_feats_ori = visual_feats\n            sims, indices = self.index.search(visual_feats.cpu(), int(cache_size))\n            B = sims.shape[0]\n            prototypes = [self.index.reconstruct(x) for x in indices.reshape(-1, ).tolist()]\n            prototypes = np.vstack(prototypes).reshape(B, int(cache_size), -1)  # [N, top_k, 1024]\n            sims = torch.tensor(sims, device=device)\n            prototypes = torch.tensor(prototypes, device=device)\n\n            sims = (sims * cache_t).softmax(dim=-1)\n            visual_feats = sims @ prototypes\n            visual_feats = visual_feats / visual_feats.norm(dim=-1, keepdim=True)\n\n            visual_feats = (1-cache_weight) * visual_feats_ori + cache_weight * visual_feats\n            visual_feats = visual_feats / visual_feats.norm(dim=-1, keepdim=True)\n\n        visual_feats = visual_feats.unsqueeze(1) # B, 1, D\n        visual_feats = self.image_bind_proj(visual_feats)\n        visual_feats_norm = self.image_bind_norm_1(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_1(F.silu(self.image_bind_f1_1(visual_feats_norm)) * self.image_bind_f3_1(visual_feats_norm))\n\n        visual_feats_norm = self.image_bind_norm_2(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_2(F.silu(self.image_bind_f1_2(visual_feats_norm)) * self.image_bind_f3_2(visual_feats_norm))\n\n        visual_feats_norm = self.image_bind_norm_3(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_3(F.silu(self.image_bind_f1_3(visual_feats_norm)) * self.image_bind_f3_3(visual_feats_norm))\n        return visual_feats\n\n    @torch.inference_mode()\n    def forward_inference(self, visual_feats, tokens, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[start_pos:start_pos + seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        prefix_query = self.prefix_query.weight.reshape(\n            self.query_layer, 1, 4096).unsqueeze(1)\n        prefix_index = 0\n        visual_proj = visual_feats # B, 1, D\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            h = layer(h, start_pos, freqs_cis, mask, visual_proj + prefix_query[prefix_index].repeat(_bsz, 1, 1))\n            prefix_index = prefix_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h[:, -1, :])\n\n        return output.float()\n\n    def forward(self, tokens, labels, imgs):\n        visual_feats = self.forward_visual({'Image': [imgs, 1]})\n\n        _bsz, seqlen = tokens.shape\n\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, 0, freqs_cis, mask)\n        prefix_query = self.prefix_query.weight.reshape(\n            self.query_layer, 1, 4096).unsqueeze(1)\n        prefix_index = 0\n        visual_proj = visual_feats\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            h = layer(h, 0, freqs_cis, mask, visual_proj + prefix_query[prefix_index])\n            prefix_index = prefix_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n            c_loss = output.mean() * 0\n        else:\n            assert self.llama.vocab_size == 32000\n            c_loss = self.criterion(output.reshape(-1, self.llama.vocab_size), labels.flatten())\n\n        return c_loss, c_loss\n\n    @torch.inference_mode()\n    def generate(\n            self,\n            inputs,\n            prompts,\n            max_gen_len: int = 256,\n            temperature: float = 0.1,\n            top_p: float = 0.75,\n            cache_size=10,\n            cache_t=20,\n            cache_weight=0.5\n    ):\n        bsz = len(prompts)\n        params = self.llama.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        with torch.cuda.amp.autocast():\n            visual_query = self.forward_visual(inputs, cache_size, cache_t, cache_weight)\n\n        if isinstance(prompts[0], str):\n            prompts = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompts])\n        max_prompt_size = max([len(t) for t in prompts])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n\n        for k, t in enumerate(prompts):\n            tokens[k, : len(t)] = torch.tensor(t).cuda().long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            with torch.cuda.amp.autocast():\n                logits = self.forward_inference(visual_query, tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            # trick: early stop if bsz==1\n            if bsz == 1 and next_token[0] == self.tokenizer.eos_id:\n                break\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n\n            # cut to max gen len\n            t = t[len(prompts[i]): len(prompts[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n\n        return decoded\n\n\n_MODELS = {\n    \"7B\": \"https://huggingface.co/Cxxs/ImageBind-LLM/resolve/main/7B.pth\",\n}\n\ndef available_models():\n    return list(_MODELS.keys())\n\ndef load(name, llama_dir, device=\"cuda\" if torch.cuda.is_available() else \"cpu\", download_root='ckpts',\n         knn=False, llama_type=\"7B\", phase=\"finetune\"):\n    if name in _MODELS:\n        model_path = download(_MODELS[name], download_root)\n    elif os.path.isfile(name):\n        model_path = name\n    else:\n        return RuntimeError(f\"Model {name} not found; available models = {available_models()}\")\n\n    llama_ckpt_dir = os.path.join(llama_dir, llama_type)\n    llama_tokenzier_path = os.path.join(llama_dir, 'tokenizer.model')\n\n    # load llama_adapter weights and model_cfg\n    print(f'Loading LLaMA-Adapter from {model_path}')\n    adapter_ckpt = torch.load(model_path, map_location='cpu')\n    model_cfg = adapter_ckpt.get('config', {})\n\n    model = LLaMA_adapter(\n        llama_ckpt_dir, llama_tokenzier_path, knn=knn, phase=phase)\n\n    load_result = model.load_state_dict(adapter_ckpt['model'], strict=False)\n    assert len(load_result.unexpected_keys) == 0, f\"Unexpected keys: {load_result.unexpected_keys}\"\n    return model.to(device)","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter.LLaMA_adapter","uri":"program://LLaMA-Adapter/class/imagebind_LLM.llama.llama_adapter.LLaMA_adapter#L18-L287","kind":"class","name":"LLaMA_adapter","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":18,"end_line":287,"context_start_line":1,"context_end_line":307,"code":"import json\nimport os\nfrom pathlib import Path\nimport numpy as np\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\n\nfrom .llama import Transformer, ModelArgs, RMSNorm\nfrom .tokenizer import Tokenizer\nfrom util.misc import download\nfrom .utils import sample_top_p\n\nfrom ImageBind.models import imagebind_model\n\n\nclass LLaMA_adapter(nn.Module):\n    \"\"\" Masked Autoencoder with VisionTransformer backbone\n    \"\"\"\n    def __init__(self, llama_ckpt_dir, llama_tokenizer, knn=False, phase=\"finetune\", legacy_bridge=False):\n        super().__init__()\n\n        # 1. imagebind and imagebind projector\n        self.image_bind = imagebind_model.imagebind_huge(pretrained=True)\n\n        self.image_bind_proj = nn.Linear(1024, 4096)\n\n        if legacy_bridge:\n            bridge_norm_layer = nn.LayerNorm\n            bridge_bias = True\n        else:\n            bridge_norm_layer = RMSNorm\n            bridge_bias = False\n\n\n        self.image_bind_norm_1 = bridge_norm_layer(4096)\n        self.image_bind_f1_1 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n        self.image_bind_f2_1 = nn.Linear(4096 * 4, 4096, bias=bridge_bias)\n        self.image_bind_f3_1 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n\n        self.image_bind_norm_2 = bridge_norm_layer(4096)\n        self.image_bind_f1_2 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n        self.image_bind_f2_2 = nn.Linear(4096 * 4, 4096, bias=bridge_bias)\n        self.image_bind_f3_2 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n\n        self.image_bind_norm_3 = bridge_norm_layer(4096)\n        self.image_bind_f1_3 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n        self.image_bind_f2_3 = nn.Linear(4096 * 4, 4096, bias=bridge_bias)\n        self.image_bind_f3_3 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n\n        # 2. tokenizer\n        self.tokenizer = Tokenizer(model_path=llama_tokenizer)\n\n        # 3. llama\n        with open(os.path.join(llama_ckpt_dir, \"params.json\"), \"r\") as f:\n            params = json.loads(f.read())\n        bias_lora = phase == \"finetune\"\n        model_args: ModelArgs = ModelArgs(\n            max_seq_len=512, max_batch_size=1, w_bias=bias_lora, w_lora=bias_lora,  **params\n        ) # max_batch_size only affects inference\n        print(f\"model args: {model_args}\")\n        model_args.vocab_size = self.tokenizer.n_words\n        torch.set_default_tensor_type(torch.cuda.HalfTensor)\n        self.llama = Transformer(model_args)\n        torch.set_default_tensor_type(torch.FloatTensor)\n\n        ckpts = sorted(Path(llama_ckpt_dir).glob(\"*.pth\"))\n        for ckpt in ckpts:\n            ckpt = torch.load(ckpt, map_location='cpu')\n            self.llama.load_state_dict(ckpt, strict=False)\n\n        # 4. prefix\n        self.query_layer = 32\n        self.query_len = 1\n        self.prefix_query = nn.Embedding(self.query_layer * self.query_len, model_args.dim)\n\n        # 5. knn\n        self.knn = knn\n        if knn:\n            import faiss\n            self.index = faiss.read_index(download(\"https://huggingface.co/csuhan/knn/resolve/main/knn.index\", \"ckpts\"))\n\n        # 6. training criterion\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n\n        self.phase = phase\n        self.set_default_trainability(self.phase)\n\n    def get_trainable_params(self, phase='finetune'):\n        trainable = {}\n        if phase == 'finetune':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'norm' in name or 'bias' in name or 'lora' in name:\n                        trainable[name] = para\n        elif phase == 'pretrain':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'gate' in name:\n                        trainable[name] = para\n                elif name.startswith(\"image_bind_\"):  # not 'image_bind.' so image_bind won't be trained.\n                    trainable[name] = para\n                elif name.startswith(\"prefix_query.\"):\n                    trainable[name] = para\n        else:\n            raise ValueError(f\"Unknown model phase: {phase}\")\n        return trainable\n\n    def set_default_trainability(self, phase='finetune'):\n        for key, value in self.named_parameters():\n            value.requires_grad = False\n        for key, value in self.get_trainable_params(phase).items():\n            value.data = value.data.float()\n            value.requires_grad = True\n\n    def forward_visual(self, inputs, cache_size=10, cache_t=20, cache_weight=0.5):\n        outputs = []\n        outputs_weights = []\n        for input_type, (input, input_weight) in inputs.items():\n            if input_type in ['Image', 'Video']:\n                type = 'vision'\n            else:\n                type = input_type.lower()\n            outputs.append(F.normalize(self.image_bind({type : input})[type], dim=-1))\n            outputs_weights.append(input_weight)\n        outputs_weights = [x/(sum(outputs_weights)+1e-6) for x in outputs_weights]\n\n        visual_feats = sum([output*output_weight for output, output_weight in zip(outputs, outputs_weights)])\n        device = visual_feats.device\n\n        if self.knn:\n            visual_feats_ori = visual_feats\n            sims, indices = self.index.search(visual_feats.cpu(), int(cache_size))\n            B = sims.shape[0]\n            prototypes = [self.index.reconstruct(x) for x in indices.reshape(-1, ).tolist()]\n            prototypes = np.vstack(prototypes).reshape(B, int(cache_size), -1)  # [N, top_k, 1024]\n            sims = torch.tensor(sims, device=device)\n            prototypes = torch.tensor(prototypes, device=device)\n\n            sims = (sims * cache_t).softmax(dim=-1)\n            visual_feats = sims @ prototypes\n            visual_feats = visual_feats / visual_feats.norm(dim=-1, keepdim=True)\n\n            visual_feats = (1-cache_weight) * visual_feats_ori + cache_weight * visual_feats\n            visual_feats = visual_feats / visual_feats.norm(dim=-1, keepdim=True)\n\n        visual_feats = visual_feats.unsqueeze(1) # B, 1, D\n        visual_feats = self.image_bind_proj(visual_feats)\n        visual_feats_norm = self.image_bind_norm_1(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_1(F.silu(self.image_bind_f1_1(visual_feats_norm)) * self.image_bind_f3_1(visual_feats_norm))\n\n        visual_feats_norm = self.image_bind_norm_2(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_2(F.silu(self.image_bind_f1_2(visual_feats_norm)) * self.image_bind_f3_2(visual_feats_norm))\n\n        visual_feats_norm = self.image_bind_norm_3(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_3(F.silu(self.image_bind_f1_3(visual_feats_norm)) * self.image_bind_f3_3(visual_feats_norm))\n        return visual_feats\n\n    @torch.inference_mode()\n    def forward_inference(self, visual_feats, tokens, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[start_pos:start_pos + seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        prefix_query = self.prefix_query.weight.reshape(\n            self.query_layer, 1, 4096).unsqueeze(1)\n        prefix_index = 0\n        visual_proj = visual_feats # B, 1, D\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            h = layer(h, start_pos, freqs_cis, mask, visual_proj + prefix_query[prefix_index].repeat(_bsz, 1, 1))\n            prefix_index = prefix_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h[:, -1, :])\n\n        return output.float()\n\n    def forward(self, tokens, labels, imgs):\n        visual_feats = self.forward_visual({'Image': [imgs, 1]})\n\n        _bsz, seqlen = tokens.shape\n\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, 0, freqs_cis, mask)\n        prefix_query = self.prefix_query.weight.reshape(\n            self.query_layer, 1, 4096).unsqueeze(1)\n        prefix_index = 0\n        visual_proj = visual_feats\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            h = layer(h, 0, freqs_cis, mask, visual_proj + prefix_query[prefix_index])\n            prefix_index = prefix_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n            c_loss = output.mean() * 0\n        else:\n            assert self.llama.vocab_size == 32000\n            c_loss = self.criterion(output.reshape(-1, self.llama.vocab_size), labels.flatten())\n\n        return c_loss, c_loss\n\n    @torch.inference_mode()\n    def generate(\n            self,\n            inputs,\n            prompts,\n            max_gen_len: int = 256,\n            temperature: float = 0.1,\n            top_p: float = 0.75,\n            cache_size=10,\n            cache_t=20,\n            cache_weight=0.5\n    ):\n        bsz = len(prompts)\n        params = self.llama.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        with torch.cuda.amp.autocast():\n            visual_query = self.forward_visual(inputs, cache_size, cache_t, cache_weight)\n\n        if isinstance(prompts[0], str):\n            prompts = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompts])\n        max_prompt_size = max([len(t) for t in prompts])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n\n        for k, t in enumerate(prompts):\n            tokens[k, : len(t)] = torch.tensor(t).cuda().long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            with torch.cuda.amp.autocast():\n                logits = self.forward_inference(visual_query, tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            # trick: early stop if bsz==1\n            if bsz == 1 and next_token[0] == self.tokenizer.eos_id:\n                break\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n\n            # cut to max gen len\n            t = t[len(prompts[i]): len(prompts[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n\n        return decoded\n\n\n_MODELS = {\n    \"7B\": \"https://huggingface.co/Cxxs/ImageBind-LLM/resolve/main/7B.pth\",\n}\n\ndef available_models():\n    return list(_MODELS.keys())\n\ndef load(name, llama_dir, device=\"cuda\" if torch.cuda.is_available() else \"cpu\", download_root='ckpts',\n         knn=False, llama_type=\"7B\", phase=\"finetune\"):\n    if name in _MODELS:\n        model_path = download(_MODELS[name], download_root)\n    elif os.path.isfile(name):\n        model_path = name\n    else:\n        return RuntimeError(f\"Model {name} not found; available models = {available_models()}\")\n\n    llama_ckpt_dir = os.path.join(llama_dir, llama_type)\n    llama_tokenzier_path = os.path.join(llama_dir, 'tokenizer.model')","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter.available_models","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama_adapter.available_models#L294-L295","kind":"function","name":"available_models","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":294,"end_line":295,"context_start_line":274,"context_end_line":315,"code":"\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n\n            # cut to max gen len\n            t = t[len(prompts[i]): len(prompts[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n\n        return decoded\n\n\n_MODELS = {\n    \"7B\": \"https://huggingface.co/Cxxs/ImageBind-LLM/resolve/main/7B.pth\",\n}\n\ndef available_models():\n    return list(_MODELS.keys())\n\ndef load(name, llama_dir, device=\"cuda\" if torch.cuda.is_available() else \"cpu\", download_root='ckpts',\n         knn=False, llama_type=\"7B\", phase=\"finetune\"):\n    if name in _MODELS:\n        model_path = download(_MODELS[name], download_root)\n    elif os.path.isfile(name):\n        model_path = name\n    else:\n        return RuntimeError(f\"Model {name} not found; available models = {available_models()}\")\n\n    llama_ckpt_dir = os.path.join(llama_dir, llama_type)\n    llama_tokenzier_path = os.path.join(llama_dir, 'tokenizer.model')\n\n    # load llama_adapter weights and model_cfg\n    print(f'Loading LLaMA-Adapter from {model_path}')\n    adapter_ckpt = torch.load(model_path, map_location='cpu')\n    model_cfg = adapter_ckpt.get('config', {})\n\n    model = LLaMA_adapter(\n        llama_ckpt_dir, llama_tokenzier_path, knn=knn, phase=phase)","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter.load","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama_adapter.load#L297-L319","kind":"function","name":"load","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":297,"end_line":319,"context_start_line":277,"context_end_line":319,"code":"\n            # cut to max gen len\n            t = t[len(prompts[i]): len(prompts[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n\n        return decoded\n\n\n_MODELS = {\n    \"7B\": \"https://huggingface.co/Cxxs/ImageBind-LLM/resolve/main/7B.pth\",\n}\n\ndef available_models():\n    return list(_MODELS.keys())\n\ndef load(name, llama_dir, device=\"cuda\" if torch.cuda.is_available() else \"cpu\", download_root='ckpts',\n         knn=False, llama_type=\"7B\", phase=\"finetune\"):\n    if name in _MODELS:\n        model_path = download(_MODELS[name], download_root)\n    elif os.path.isfile(name):\n        model_path = name\n    else:\n        return RuntimeError(f\"Model {name} not found; available models = {available_models()}\")\n\n    llama_ckpt_dir = os.path.join(llama_dir, llama_type)\n    llama_tokenzier_path = os.path.join(llama_dir, 'tokenizer.model')\n\n    # load llama_adapter weights and model_cfg\n    print(f'Loading LLaMA-Adapter from {model_path}')\n    adapter_ckpt = torch.load(model_path, map_location='cpu')\n    model_cfg = adapter_ckpt.get('config', {})\n\n    model = LLaMA_adapter(\n        llama_ckpt_dir, llama_tokenzier_path, knn=knn, phase=phase)\n\n    load_result = model.load_state_dict(adapter_ckpt['model'], strict=False)\n    assert len(load_result.unexpected_keys) == 0, f\"Unexpected keys: {load_result.unexpected_keys}\"\n    return model.to(device)","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama_adapter.__init__#L21-L89","kind":"function","name":"__init__","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":21,"end_line":89,"context_start_line":1,"context_end_line":109,"code":"import json\nimport os\nfrom pathlib import Path\nimport numpy as np\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\n\nfrom .llama import Transformer, ModelArgs, RMSNorm\nfrom .tokenizer import Tokenizer\nfrom util.misc import download\nfrom .utils import sample_top_p\n\nfrom ImageBind.models import imagebind_model\n\n\nclass LLaMA_adapter(nn.Module):\n    \"\"\" Masked Autoencoder with VisionTransformer backbone\n    \"\"\"\n    def __init__(self, llama_ckpt_dir, llama_tokenizer, knn=False, phase=\"finetune\", legacy_bridge=False):\n        super().__init__()\n\n        # 1. imagebind and imagebind projector\n        self.image_bind = imagebind_model.imagebind_huge(pretrained=True)\n\n        self.image_bind_proj = nn.Linear(1024, 4096)\n\n        if legacy_bridge:\n            bridge_norm_layer = nn.LayerNorm\n            bridge_bias = True\n        else:\n            bridge_norm_layer = RMSNorm\n            bridge_bias = False\n\n\n        self.image_bind_norm_1 = bridge_norm_layer(4096)\n        self.image_bind_f1_1 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n        self.image_bind_f2_1 = nn.Linear(4096 * 4, 4096, bias=bridge_bias)\n        self.image_bind_f3_1 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n\n        self.image_bind_norm_2 = bridge_norm_layer(4096)\n        self.image_bind_f1_2 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n        self.image_bind_f2_2 = nn.Linear(4096 * 4, 4096, bias=bridge_bias)\n        self.image_bind_f3_2 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n\n        self.image_bind_norm_3 = bridge_norm_layer(4096)\n        self.image_bind_f1_3 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n        self.image_bind_f2_3 = nn.Linear(4096 * 4, 4096, bias=bridge_bias)\n        self.image_bind_f3_3 = nn.Linear(4096, 4096 * 4, bias=bridge_bias)\n\n        # 2. tokenizer\n        self.tokenizer = Tokenizer(model_path=llama_tokenizer)\n\n        # 3. llama\n        with open(os.path.join(llama_ckpt_dir, \"params.json\"), \"r\") as f:\n            params = json.loads(f.read())\n        bias_lora = phase == \"finetune\"\n        model_args: ModelArgs = ModelArgs(\n            max_seq_len=512, max_batch_size=1, w_bias=bias_lora, w_lora=bias_lora,  **params\n        ) # max_batch_size only affects inference\n        print(f\"model args: {model_args}\")\n        model_args.vocab_size = self.tokenizer.n_words\n        torch.set_default_tensor_type(torch.cuda.HalfTensor)\n        self.llama = Transformer(model_args)\n        torch.set_default_tensor_type(torch.FloatTensor)\n\n        ckpts = sorted(Path(llama_ckpt_dir).glob(\"*.pth\"))\n        for ckpt in ckpts:\n            ckpt = torch.load(ckpt, map_location='cpu')\n            self.llama.load_state_dict(ckpt, strict=False)\n\n        # 4. prefix\n        self.query_layer = 32\n        self.query_len = 1\n        self.prefix_query = nn.Embedding(self.query_layer * self.query_len, model_args.dim)\n\n        # 5. knn\n        self.knn = knn\n        if knn:\n            import faiss\n            self.index = faiss.read_index(download(\"https://huggingface.co/csuhan/knn/resolve/main/knn.index\", \"ckpts\"))\n\n        # 6. training criterion\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n\n        self.phase = phase\n        self.set_default_trainability(self.phase)\n\n    def get_trainable_params(self, phase='finetune'):\n        trainable = {}\n        if phase == 'finetune':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'norm' in name or 'bias' in name or 'lora' in name:\n                        trainable[name] = para\n        elif phase == 'pretrain':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'gate' in name:\n                        trainable[name] = para\n                elif name.startswith(\"image_bind_\"):  # not 'image_bind.' so image_bind won't be trained.\n                    trainable[name] = para\n                elif name.startswith(\"prefix_query.\"):\n                    trainable[name] = para\n        else:\n            raise ValueError(f\"Unknown model phase: {phase}\")\n        return trainable","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter.get_trainable_params","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama_adapter.get_trainable_params#L91-L109","kind":"function","name":"get_trainable_params","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":91,"end_line":109,"context_start_line":71,"context_end_line":129,"code":"            self.llama.load_state_dict(ckpt, strict=False)\n\n        # 4. prefix\n        self.query_layer = 32\n        self.query_len = 1\n        self.prefix_query = nn.Embedding(self.query_layer * self.query_len, model_args.dim)\n\n        # 5. knn\n        self.knn = knn\n        if knn:\n            import faiss\n            self.index = faiss.read_index(download(\"https://huggingface.co/csuhan/knn/resolve/main/knn.index\", \"ckpts\"))\n\n        # 6. training criterion\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n\n        self.phase = phase\n        self.set_default_trainability(self.phase)\n\n    def get_trainable_params(self, phase='finetune'):\n        trainable = {}\n        if phase == 'finetune':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'norm' in name or 'bias' in name or 'lora' in name:\n                        trainable[name] = para\n        elif phase == 'pretrain':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'gate' in name:\n                        trainable[name] = para\n                elif name.startswith(\"image_bind_\"):  # not 'image_bind.' so image_bind won't be trained.\n                    trainable[name] = para\n                elif name.startswith(\"prefix_query.\"):\n                    trainable[name] = para\n        else:\n            raise ValueError(f\"Unknown model phase: {phase}\")\n        return trainable\n\n    def set_default_trainability(self, phase='finetune'):\n        for key, value in self.named_parameters():\n            value.requires_grad = False\n        for key, value in self.get_trainable_params(phase).items():\n            value.data = value.data.float()\n            value.requires_grad = True\n\n    def forward_visual(self, inputs, cache_size=10, cache_t=20, cache_weight=0.5):\n        outputs = []\n        outputs_weights = []\n        for input_type, (input, input_weight) in inputs.items():\n            if input_type in ['Image', 'Video']:\n                type = 'vision'\n            else:\n                type = input_type.lower()\n            outputs.append(F.normalize(self.image_bind({type : input})[type], dim=-1))\n            outputs_weights.append(input_weight)\n        outputs_weights = [x/(sum(outputs_weights)+1e-6) for x in outputs_weights]\n","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter.set_default_trainability","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama_adapter.set_default_trainability#L111-L116","kind":"function","name":"set_default_trainability","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":111,"end_line":116,"context_start_line":91,"context_end_line":136,"code":"    def get_trainable_params(self, phase='finetune'):\n        trainable = {}\n        if phase == 'finetune':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'norm' in name or 'bias' in name or 'lora' in name:\n                        trainable[name] = para\n        elif phase == 'pretrain':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'gate' in name:\n                        trainable[name] = para\n                elif name.startswith(\"image_bind_\"):  # not 'image_bind.' so image_bind won't be trained.\n                    trainable[name] = para\n                elif name.startswith(\"prefix_query.\"):\n                    trainable[name] = para\n        else:\n            raise ValueError(f\"Unknown model phase: {phase}\")\n        return trainable\n\n    def set_default_trainability(self, phase='finetune'):\n        for key, value in self.named_parameters():\n            value.requires_grad = False\n        for key, value in self.get_trainable_params(phase).items():\n            value.data = value.data.float()\n            value.requires_grad = True\n\n    def forward_visual(self, inputs, cache_size=10, cache_t=20, cache_weight=0.5):\n        outputs = []\n        outputs_weights = []\n        for input_type, (input, input_weight) in inputs.items():\n            if input_type in ['Image', 'Video']:\n                type = 'vision'\n            else:\n                type = input_type.lower()\n            outputs.append(F.normalize(self.image_bind({type : input})[type], dim=-1))\n            outputs_weights.append(input_weight)\n        outputs_weights = [x/(sum(outputs_weights)+1e-6) for x in outputs_weights]\n\n        visual_feats = sum([output*output_weight for output, output_weight in zip(outputs, outputs_weights)])\n        device = visual_feats.device\n\n        if self.knn:\n            visual_feats_ori = visual_feats\n            sims, indices = self.index.search(visual_feats.cpu(), int(cache_size))\n            B = sims.shape[0]","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter.forward_visual","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama_adapter.forward_visual#L118-L159","kind":"function","name":"forward_visual","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":118,"end_line":159,"context_start_line":98,"context_end_line":179,"code":"        elif phase == 'pretrain':\n            for name, para in self.named_parameters():\n                if name.startswith(\"llama.\"):\n                    if 'gate' in name:\n                        trainable[name] = para\n                elif name.startswith(\"image_bind_\"):  # not 'image_bind.' so image_bind won't be trained.\n                    trainable[name] = para\n                elif name.startswith(\"prefix_query.\"):\n                    trainable[name] = para\n        else:\n            raise ValueError(f\"Unknown model phase: {phase}\")\n        return trainable\n\n    def set_default_trainability(self, phase='finetune'):\n        for key, value in self.named_parameters():\n            value.requires_grad = False\n        for key, value in self.get_trainable_params(phase).items():\n            value.data = value.data.float()\n            value.requires_grad = True\n\n    def forward_visual(self, inputs, cache_size=10, cache_t=20, cache_weight=0.5):\n        outputs = []\n        outputs_weights = []\n        for input_type, (input, input_weight) in inputs.items():\n            if input_type in ['Image', 'Video']:\n                type = 'vision'\n            else:\n                type = input_type.lower()\n            outputs.append(F.normalize(self.image_bind({type : input})[type], dim=-1))\n            outputs_weights.append(input_weight)\n        outputs_weights = [x/(sum(outputs_weights)+1e-6) for x in outputs_weights]\n\n        visual_feats = sum([output*output_weight for output, output_weight in zip(outputs, outputs_weights)])\n        device = visual_feats.device\n\n        if self.knn:\n            visual_feats_ori = visual_feats\n            sims, indices = self.index.search(visual_feats.cpu(), int(cache_size))\n            B = sims.shape[0]\n            prototypes = [self.index.reconstruct(x) for x in indices.reshape(-1, ).tolist()]\n            prototypes = np.vstack(prototypes).reshape(B, int(cache_size), -1)  # [N, top_k, 1024]\n            sims = torch.tensor(sims, device=device)\n            prototypes = torch.tensor(prototypes, device=device)\n\n            sims = (sims * cache_t).softmax(dim=-1)\n            visual_feats = sims @ prototypes\n            visual_feats = visual_feats / visual_feats.norm(dim=-1, keepdim=True)\n\n            visual_feats = (1-cache_weight) * visual_feats_ori + cache_weight * visual_feats\n            visual_feats = visual_feats / visual_feats.norm(dim=-1, keepdim=True)\n\n        visual_feats = visual_feats.unsqueeze(1) # B, 1, D\n        visual_feats = self.image_bind_proj(visual_feats)\n        visual_feats_norm = self.image_bind_norm_1(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_1(F.silu(self.image_bind_f1_1(visual_feats_norm)) * self.image_bind_f3_1(visual_feats_norm))\n\n        visual_feats_norm = self.image_bind_norm_2(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_2(F.silu(self.image_bind_f1_2(visual_feats_norm)) * self.image_bind_f3_2(visual_feats_norm))\n\n        visual_feats_norm = self.image_bind_norm_3(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_3(F.silu(self.image_bind_f1_3(visual_feats_norm)) * self.image_bind_f3_3(visual_feats_norm))\n        return visual_feats\n\n    @torch.inference_mode()\n    def forward_inference(self, visual_feats, tokens, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[start_pos:start_pos + seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        prefix_query = self.prefix_query.weight.reshape(\n            self.query_layer, 1, 4096).unsqueeze(1)\n        prefix_index = 0\n        visual_proj = visual_feats # B, 1, D\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            h = layer(h, start_pos, freqs_cis, mask, visual_proj + prefix_query[prefix_index].repeat(_bsz, 1, 1))","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter.forward_inference","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama_adapter.forward_inference#L162-L185","kind":"function","name":"forward_inference","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":162,"end_line":185,"context_start_line":142,"context_end_line":205,"code":"            sims = (sims * cache_t).softmax(dim=-1)\n            visual_feats = sims @ prototypes\n            visual_feats = visual_feats / visual_feats.norm(dim=-1, keepdim=True)\n\n            visual_feats = (1-cache_weight) * visual_feats_ori + cache_weight * visual_feats\n            visual_feats = visual_feats / visual_feats.norm(dim=-1, keepdim=True)\n\n        visual_feats = visual_feats.unsqueeze(1) # B, 1, D\n        visual_feats = self.image_bind_proj(visual_feats)\n        visual_feats_norm = self.image_bind_norm_1(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_1(F.silu(self.image_bind_f1_1(visual_feats_norm)) * self.image_bind_f3_1(visual_feats_norm))\n\n        visual_feats_norm = self.image_bind_norm_2(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_2(F.silu(self.image_bind_f1_2(visual_feats_norm)) * self.image_bind_f3_2(visual_feats_norm))\n\n        visual_feats_norm = self.image_bind_norm_3(visual_feats)\n        visual_feats = visual_feats + self.image_bind_f2_3(F.silu(self.image_bind_f1_3(visual_feats_norm)) * self.image_bind_f3_3(visual_feats_norm))\n        return visual_feats\n\n    @torch.inference_mode()\n    def forward_inference(self, visual_feats, tokens, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[start_pos:start_pos + seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        prefix_query = self.prefix_query.weight.reshape(\n            self.query_layer, 1, 4096).unsqueeze(1)\n        prefix_index = 0\n        visual_proj = visual_feats # B, 1, D\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            h = layer(h, start_pos, freqs_cis, mask, visual_proj + prefix_query[prefix_index].repeat(_bsz, 1, 1))\n            prefix_index = prefix_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h[:, -1, :])\n\n        return output.float()\n\n    def forward(self, tokens, labels, imgs):\n        visual_feats = self.forward_visual({'Image': [imgs, 1]})\n\n        _bsz, seqlen = tokens.shape\n\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, 0, freqs_cis, mask)\n        prefix_query = self.prefix_query.weight.reshape(\n            self.query_layer, 1, 4096).unsqueeze(1)\n        prefix_index = 0\n        visual_proj = visual_feats\n        for layer in self.llama.layers[-1 * self.query_layer:]:","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter.forward","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama_adapter.forward#L187-L220","kind":"function","name":"forward","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":187,"end_line":220,"context_start_line":167,"context_end_line":240,"code":"        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, start_pos, freqs_cis, mask)\n        prefix_query = self.prefix_query.weight.reshape(\n            self.query_layer, 1, 4096).unsqueeze(1)\n        prefix_index = 0\n        visual_proj = visual_feats # B, 1, D\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            h = layer(h, start_pos, freqs_cis, mask, visual_proj + prefix_query[prefix_index].repeat(_bsz, 1, 1))\n            prefix_index = prefix_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h[:, -1, :])\n\n        return output.float()\n\n    def forward(self, tokens, labels, imgs):\n        visual_feats = self.forward_visual({'Image': [imgs, 1]})\n\n        _bsz, seqlen = tokens.shape\n\n        h = self.llama.tok_embeddings(tokens)\n        freqs_cis = self.llama.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n\n        for layer in self.llama.layers[:-1 * self.query_layer]:\n            h = layer(h, 0, freqs_cis, mask)\n        prefix_query = self.prefix_query.weight.reshape(\n            self.query_layer, 1, 4096).unsqueeze(1)\n        prefix_index = 0\n        visual_proj = visual_feats\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            h = layer(h, 0, freqs_cis, mask, visual_proj + prefix_query[prefix_index])\n            prefix_index = prefix_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n            c_loss = output.mean() * 0\n        else:\n            assert self.llama.vocab_size == 32000\n            c_loss = self.criterion(output.reshape(-1, self.llama.vocab_size), labels.flatten())\n\n        return c_loss, c_loss\n\n    @torch.inference_mode()\n    def generate(\n            self,\n            inputs,\n            prompts,\n            max_gen_len: int = 256,\n            temperature: float = 0.1,\n            top_p: float = 0.75,\n            cache_size=10,\n            cache_t=20,\n            cache_weight=0.5\n    ):\n        bsz = len(prompts)\n        params = self.llama.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        with torch.cuda.amp.autocast():\n            visual_query = self.forward_visual(inputs, cache_size, cache_t, cache_weight)\n","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama_adapter.generate","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama_adapter.generate#L223-L287","kind":"function","name":"generate","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":223,"end_line":287,"context_start_line":203,"context_end_line":307,"code":"        prefix_index = 0\n        visual_proj = visual_feats\n        for layer in self.llama.layers[-1 * self.query_layer:]:\n            h = layer(h, 0, freqs_cis, mask, visual_proj + prefix_query[prefix_index])\n            prefix_index = prefix_index + 1\n\n        h = self.llama.norm(h)\n        output = self.llama.output(h)\n        output = output[:, :-1, :]\n        labels = labels[:, 1:]\n\n        if labels.sum() == 0:\n            c_loss = output.mean() * 0\n        else:\n            assert self.llama.vocab_size == 32000\n            c_loss = self.criterion(output.reshape(-1, self.llama.vocab_size), labels.flatten())\n\n        return c_loss, c_loss\n\n    @torch.inference_mode()\n    def generate(\n            self,\n            inputs,\n            prompts,\n            max_gen_len: int = 256,\n            temperature: float = 0.1,\n            top_p: float = 0.75,\n            cache_size=10,\n            cache_t=20,\n            cache_weight=0.5\n    ):\n        bsz = len(prompts)\n        params = self.llama.params\n        assert bsz <= params.max_batch_size, (bsz, params.max_batch_size)\n\n        with torch.cuda.amp.autocast():\n            visual_query = self.forward_visual(inputs, cache_size, cache_t, cache_weight)\n\n        if isinstance(prompts[0], str):\n            prompts = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n\n        min_prompt_size = min([len(t) for t in prompts])\n        max_prompt_size = max([len(t) for t in prompts])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n\n        for k, t in enumerate(prompts):\n            tokens[k, : len(t)] = torch.tensor(t).cuda().long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            with torch.cuda.amp.autocast():\n                logits = self.forward_inference(visual_query, tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n\n            next_token = torch.where(\n                input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token\n            )\n            tokens[:, cur_pos] = next_token\n            # trick: early stop if bsz==1\n            if bsz == 1 and next_token[0] == self.tokenizer.eos_id:\n                break\n            prev_pos = cur_pos\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n\n            # cut to max gen len\n            t = t[len(prompts[i]): len(prompts[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n\n        return decoded\n\n\n_MODELS = {\n    \"7B\": \"https://huggingface.co/Cxxs/ImageBind-LLM/resolve/main/7B.pth\",\n}\n\ndef available_models():\n    return list(_MODELS.keys())\n\ndef load(name, llama_dir, device=\"cuda\" if torch.cuda.is_available() else \"cpu\", download_root='ckpts',\n         knn=False, llama_type=\"7B\", phase=\"finetune\"):\n    if name in _MODELS:\n        model_path = download(_MODELS[name], download_root)\n    elif os.path.isfile(name):\n        model_path = name\n    else:\n        return RuntimeError(f\"Model {name} not found; available models = {available_models()}\")\n\n    llama_ckpt_dir = os.path.join(llama_dir, llama_type)\n    llama_tokenzier_path = os.path.join(llama_dir, 'tokenizer.model')","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama","uri":"program://LLaMA-Adapter/module/imagebind_LLM.llama.llama#L1-L317","kind":"module","name":"imagebind_LLM.llama.llama","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":1,"end_line":317,"context_start_line":1,"context_end_line":317,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\n\nimport torch\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\nimport torch.nn.functional as F\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    w_bias: bool = True # use bias tuning\n    w_lora: bool = True # use lora tuning\n    lora_rank: int = 16\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n        self.args = args\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=args.w_bias\n        )\n        self.wk = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wv = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wo = Linear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=args.w_bias\n        )\n        if args.w_bias:\n            nn.init.constant_(self.wq.bias.data, 0)\n            nn.init.constant_(self.wo.bias.data, 0)\n\n        self.w_lora = args.w_lora\n        if args.w_lora:\n           self.lora_wq_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wq_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wk_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wk_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wv_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wv_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wo_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wo_l2 = Linear(args.lora_rank, args.dim, bias=False)\n           nn.init.constant_(self.lora_wq_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wk_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wv_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wo_l2.weight.data, 0)\n\n        self.cache_k = None\n        self.cache_v = None\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n\n    def train(self, mode: bool = True):\n        if mode:\n            self.cache_k = None\n            self.cache_v = None\n        else:\n            self.cache_k = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n            self.cache_v = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n        return super().train(mode)\n\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n        if self.w_lora:\n           xq = xq + self.lora_wq_l2(self.lora_wq_l1(x))\n           xk = xk + self.lora_wk_l2(self.lora_wk_l1(x))\n           xv = xv + self.lora_wv_l2(self.lora_wv_l1(x))\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if not self.training:\n            self.cache_k = self.cache_k.to(xq)\n            self.cache_v = self.cache_v.to(xq)\n\n            self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk\n            self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n            keys = self.cache_k[:bsz, : start_pos + seqlen]\n            values = self.cache_v[:bsz, : start_pos + seqlen]\n        else:\n            assert start_pos==0\n            keys = xk\n            values = xv\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_v = self.wv(adapter).view(bsz, adapter_len, self.n_local_heads, self.head_dim)\n            adapter_v = adapter_v.transpose(1, 2)\n\n            if adapter_len > 1:\n                adapter_k = self.wk(adapter).view(bsz, adapter_len, self.n_local_heads, self.head_dim)\n                adapter_k = adapter_k.transpose(1, 2)\n\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n\n        if adapter is not None:\n            if adapter_len > 1:\n                adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n                adapter_scores = self.gate.tanh() * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n                output = output + torch.matmul(adapter_scores, adapter_v)\n            else:\n                output = output + self.gate.tanh() * adapter_v\n\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        if self.w_lora:\n           return self.wo(output) + self.lora_wo_l2(self.lora_wo_l1(output))\n        else:\n           return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n        args: ModelArgs\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=args.w_bias\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=args.w_bias\n        )\n        self.w3 = Linear(\n            dim, hidden_dim, bias=args.w_bias\n        )\n        if args.w_bias:\n            nn.init.constant_(self.w1.bias.data, 0)\n            nn.init.constant_(self.w2.bias.data, 0)\n            nn.init.constant_(self.w3.bias.data, 0)\n\n        self.w_lora = args.w_lora\n        if args.w_lora:\n           self.lora_w1_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w1_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           self.lora_w2_l1 = Linear(hidden_dim, args.lora_rank, bias=False)\n           self.lora_w2_l2 = Linear(args.lora_rank, dim, bias=False)\n           self.lora_w3_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w3_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           nn.init.constant_(self.lora_w1_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w2_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w3_l2.weight.data, 0)\n\n    def forward(self, x):\n        if self.w_lora:\n           out = F.silu(self.w1(x) + self.lora_w1_l2(self.lora_w1_l1(x))) * (self.w3(x) + self.lora_w3_l2(self.lora_w3_l1(x)))\n           return self.w2(out) + self.lora_w2_l2(self.lora_w2_l1(out))\n        else:\n           return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, args=args\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(\n            params.dim, params.vocab_size, bias=False\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.ModelArgs","uri":"program://LLaMA-Adapter/class/imagebind_LLM.llama.llama.ModelArgs#L15-L28","kind":"class","name":"ModelArgs","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":15,"end_line":28,"context_start_line":1,"context_end_line":48,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\n\nimport torch\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\nimport torch.nn.functional as F\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    w_bias: bool = True # use bias tuning\n    w_lora: bool = True # use lora tuning\n    lora_rank: int = 16\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.RMSNorm","uri":"program://LLaMA-Adapter/class/imagebind_LLM.llama.llama.RMSNorm#L31-L42","kind":"class","name":"RMSNorm","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":31,"end_line":42,"context_start_line":11,"context_end_line":62,"code":"import torch.nn.functional as F\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    w_bias: bool = True # use bias tuning\n    w_lora: bool = True # use lora tuning\n    lora_rank: int = 16\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.precompute_freqs_cis","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama.precompute_freqs_cis#L45-L50","kind":"function","name":"precompute_freqs_cis","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":45,"end_line":50,"context_start_line":25,"context_end_line":70,"code":"\n    w_bias: bool = True # use bias tuning\n    w_lora: bool = True # use lora tuning\n    lora_rank: int = 16\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.reshape_for_broadcast","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama.reshape_for_broadcast#L53-L58","kind":"function","name":"reshape_for_broadcast","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":53,"end_line":58,"context_start_line":33,"context_end_line":78,"code":"        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n        self.args = args\n","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.apply_rotary_emb","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama.apply_rotary_emb#L61-L71","kind":"function","name":"apply_rotary_emb","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":61,"end_line":71,"context_start_line":41,"context_end_line":91,"code":"        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n        self.args = args\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=args.w_bias\n        )\n        self.wk = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.Attention","uri":"program://LLaMA-Adapter/class/imagebind_LLM.llama.llama.Attention#L74-L208","kind":"class","name":"Attention","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":74,"end_line":208,"context_start_line":54,"context_end_line":228,"code":"    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n        self.args = args\n\n        self.n_local_heads = args.n_heads\n        self.head_dim = args.dim // args.n_heads\n\n        self.wq = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=args.w_bias\n        )\n        self.wk = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wv = Linear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False\n        )\n        self.wo = Linear(\n            args.n_heads * self.head_dim,\n            args.dim,\n            bias=args.w_bias\n        )\n        if args.w_bias:\n            nn.init.constant_(self.wq.bias.data, 0)\n            nn.init.constant_(self.wo.bias.data, 0)\n\n        self.w_lora = args.w_lora\n        if args.w_lora:\n           self.lora_wq_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wq_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wk_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wk_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wv_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wv_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wo_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wo_l2 = Linear(args.lora_rank, args.dim, bias=False)\n           nn.init.constant_(self.lora_wq_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wk_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wv_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wo_l2.weight.data, 0)\n\n        self.cache_k = None\n        self.cache_v = None\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n\n    def train(self, mode: bool = True):\n        if mode:\n            self.cache_k = None\n            self.cache_v = None\n        else:\n            self.cache_k = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n            self.cache_v = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n        return super().train(mode)\n\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n        if self.w_lora:\n           xq = xq + self.lora_wq_l2(self.lora_wq_l1(x))\n           xk = xk + self.lora_wk_l2(self.lora_wk_l1(x))\n           xv = xv + self.lora_wv_l2(self.lora_wv_l1(x))\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if not self.training:\n            self.cache_k = self.cache_k.to(xq)\n            self.cache_v = self.cache_v.to(xq)\n\n            self.cache_k[:bsz, start_pos : start_pos + seqlen] = xk\n            self.cache_v[:bsz, start_pos : start_pos + seqlen] = xv\n\n            keys = self.cache_k[:bsz, : start_pos + seqlen]\n            values = self.cache_v[:bsz, : start_pos + seqlen]\n        else:\n            assert start_pos==0\n            keys = xk\n            values = xv\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_v = self.wv(adapter).view(bsz, adapter_len, self.n_local_heads, self.head_dim)\n            adapter_v = adapter_v.transpose(1, 2)\n\n            if adapter_len > 1:\n                adapter_k = self.wk(adapter).view(bsz, adapter_len, self.n_local_heads, self.head_dim)\n                adapter_k = adapter_k.transpose(1, 2)\n\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n        scores = torch.matmul(xq, keys.transpose(2, 3)) / math.sqrt(self.head_dim)\n\n        if mask is not None:\n            scores = scores + mask  # (bs, n_local_heads, slen, cache_len + slen)\n\n        scores = F.softmax(scores.float(), dim=-1).type_as(xq)\n        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n\n        if adapter is not None:\n            if adapter_len > 1:\n                adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n                adapter_scores = self.gate.tanh() * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n                output = output + torch.matmul(adapter_scores, adapter_v)\n            else:\n                output = output + self.gate.tanh() * adapter_v\n\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        if self.w_lora:\n           return self.wo(output) + self.lora_wo_l2(self.lora_wo_l1(output))\n        else:\n           return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n        args: ModelArgs\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=args.w_bias\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=args.w_bias\n        )","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.FeedForward","uri":"program://LLaMA-Adapter/class/imagebind_LLM.llama.llama.FeedForward#L211-L254","kind":"class","name":"FeedForward","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":211,"end_line":254,"context_start_line":191,"context_end_line":274,"code":"        output = torch.matmul(scores, values)  # (bs, n_local_heads, slen, head_dim)\n\n        if adapter is not None:\n            if adapter_len > 1:\n                adapter_scores = torch.matmul(xq, adapter_k.transpose(2, 3)) / math.sqrt(self.head_dim)\n                adapter_scores = self.gate.tanh() * F.softmax(adapter_scores.float(), dim=-1).type_as(xq)\n                output = output + torch.matmul(adapter_scores, adapter_v)\n            else:\n                output = output + self.gate.tanh() * adapter_v\n\n        output = output.transpose(\n            1, 2\n        ).contiguous().view(bsz, seqlen, -1)\n\n        if self.w_lora:\n           return self.wo(output) + self.lora_wo_l2(self.lora_wo_l1(output))\n        else:\n           return self.wo(output)\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n        args: ModelArgs\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.w1 = Linear(\n            dim, hidden_dim, bias=args.w_bias\n        )\n        self.w2 = Linear(\n            hidden_dim, dim, bias=args.w_bias\n        )\n        self.w3 = Linear(\n            dim, hidden_dim, bias=args.w_bias\n        )\n        if args.w_bias:\n            nn.init.constant_(self.w1.bias.data, 0)\n            nn.init.constant_(self.w2.bias.data, 0)\n            nn.init.constant_(self.w3.bias.data, 0)\n\n        self.w_lora = args.w_lora\n        if args.w_lora:\n           self.lora_w1_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w1_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           self.lora_w2_l1 = Linear(hidden_dim, args.lora_rank, bias=False)\n           self.lora_w2_l2 = Linear(args.lora_rank, dim, bias=False)\n           self.lora_w3_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w3_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           nn.init.constant_(self.lora_w1_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w2_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w3_l2.weight.data, 0)\n\n    def forward(self, x):\n        if self.w_lora:\n           out = F.silu(self.w1(x) + self.lora_w1_l2(self.lora_w1_l1(x))) * (self.w3(x) + self.lora_w3_l2(self.lora_w3_l1(x)))\n           return self.w2(out) + self.lora_w2_l2(self.lora_w2_l1(out))\n        else:\n           return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, args=args\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.TransformerBlock","uri":"program://LLaMA-Adapter/class/imagebind_LLM.llama.llama.TransformerBlock#L257-L275","kind":"class","name":"TransformerBlock","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":257,"end_line":275,"context_start_line":237,"context_end_line":295,"code":"        self.w_lora = args.w_lora\n        if args.w_lora:\n           self.lora_w1_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w1_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           self.lora_w2_l1 = Linear(hidden_dim, args.lora_rank, bias=False)\n           self.lora_w2_l2 = Linear(args.lora_rank, dim, bias=False)\n           self.lora_w3_l1 = Linear(dim, args.lora_rank, bias=False)\n           self.lora_w3_l2 = Linear(args.lora_rank, hidden_dim, bias=False)\n           nn.init.constant_(self.lora_w1_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w2_l2.weight.data, 0)\n           nn.init.constant_(self.lora_w3_l2.weight.data, 0)\n\n    def forward(self, x):\n        if self.w_lora:\n           out = F.silu(self.w1(x) + self.lora_w1_l2(self.lora_w1_l1(x))) * (self.w3(x) + self.lora_w3_l2(self.lora_w3_l1(x)))\n           return self.w2(out) + self.lora_w2_l2(self.lora_w2_l1(out))\n        else:\n           return self.w2(F.silu(self.w1(x)) * self.w3(x))\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, args=args\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(\n            params.dim, params.vocab_size, bias=False\n        )","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.Transformer","uri":"program://LLaMA-Adapter/class/imagebind_LLM.llama.llama.Transformer#L278-L317","kind":"class","name":"Transformer","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":278,"end_line":317,"context_start_line":258,"context_end_line":317,"code":"    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, args=args\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(\n            params.dim, params.vocab_size, bias=False\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama.__init__#L279-L299","kind":"function","name":"__init__","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":279,"end_line":299,"context_start_line":259,"context_end_line":317,"code":"        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim, hidden_dim=4 * args.dim, multiple_of=args.multiple_of, args=args\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], prompt=None):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, prompt)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(\n            params.dim, params.vocab_size, bias=False\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama._norm","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama._norm#L37-L38","kind":"function","name":"_norm","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":37,"end_line":38,"context_start_line":17,"context_end_line":58,"code":"    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_batch_size: int = 32\n    max_seq_len: int = 2048\n\n    w_bias: bool = True # use bias tuning\n    w_lora: bool = True # use lora tuning\n    lora_rank: int = 16\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return output * self.weight\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.forward","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama.forward#L302-L317","kind":"function","name":"forward","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":302,"end_line":317,"context_start_line":282,"context_end_line":317,"code":"        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = Embedding(\n            params.vocab_size, params.dim\n        )\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = Linear(\n            params.dim, params.vocab_size, bias=False\n        )\n\n        self.freqs_cis = precompute_freqs_cis(\n            self.params.dim // self.params.n_heads, self.params.max_seq_len * 2\n        )\n\n    @torch.inference_mode()\n    def forward(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n\n        mask = None\n        if seqlen > 1:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=start_pos + 1).type_as(h)\n\n        for layer in self.layers:\n            h = layer(h, start_pos, freqs_cis, mask)\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])  # only compute last logits\n        return output.float()","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.llama.train","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.llama.train#L130-L141","kind":"function","name":"train","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":130,"end_line":141,"context_start_line":110,"context_end_line":161,"code":"\n           self.lora_wk_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wk_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wv_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wv_l2 = Linear(args.lora_rank, args.dim, bias=False)\n\n           self.lora_wo_l1 = Linear(args.dim, args.lora_rank, bias=False)\n           self.lora_wo_l2 = Linear(args.lora_rank, args.dim, bias=False)\n           nn.init.constant_(self.lora_wq_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wk_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wv_l2.weight.data, 0)\n           nn.init.constant_(self.lora_wo_l2.weight.data, 0)\n\n        self.cache_k = None\n        self.cache_v = None\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, self.n_local_heads, 1, 1))\n\n\n    def train(self, mode: bool = True):\n        if mode:\n            self.cache_k = None\n            self.cache_v = None\n        else:\n            self.cache_k = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n            self.cache_v = torch.zeros(\n                (self.args.max_batch_size, self.args.max_seq_len, self.n_local_heads, self.head_dim)\n            ).cuda()\n        return super().train(mode)\n\n\n    def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None):\n        bsz, seqlen, _ = x.shape\n        xq, xk, xv = self.wq(x), self.wk(x), self.wv(x)\n        if self.w_lora:\n           xq = xq + self.lora_wq_l2(self.lora_wq_l1(x))\n           xk = xk + self.lora_wk_l2(self.lora_wk_l1(x))\n           xv = xv + self.lora_wv_l2(self.lora_wv_l1(x))\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if not self.training:\n            self.cache_k = self.cache_k.to(xq)\n            self.cache_v = self.cache_v.to(xq)\n","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.utils","uri":"program://LLaMA-Adapter/module/imagebind_LLM.llama.utils#L1-L33","kind":"module","name":"imagebind_LLM.llama.utils","path":"imagebind_LLM/llama/utils.py","language":"python","start_line":1,"end_line":33,"context_start_line":1,"context_end_line":33,"code":"import torch\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token\n\n\ndef format_prompt(instruction, input=None):\n\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n","source_hash":"e4a4149bd7b1c9a320c9eec061ba9666d770f9cf901685563ae4f0486aa245ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.utils.sample_top_p","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.utils.sample_top_p#L4-L12","kind":"function","name":"sample_top_p","path":"imagebind_LLM/llama/utils.py","language":"python","start_line":4,"end_line":12,"context_start_line":1,"context_end_line":32,"code":"import torch\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token\n\n\ndef format_prompt(instruction, input=None):\n\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})","source_hash":"e4a4149bd7b1c9a320c9eec061ba9666d770f9cf901685563ae4f0486aa245ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.utils.format_prompt","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.utils.format_prompt#L15-L32","kind":"function","name":"format_prompt","path":"imagebind_LLM/llama/utils.py","language":"python","start_line":15,"end_line":32,"context_start_line":1,"context_end_line":33,"code":"import torch\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token\n\n\ndef format_prompt(instruction, input=None):\n\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n        ),\n        \"prompt_no_input\": (\n            \"Below is an instruction that describes a task. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n        ),\n    }\n    if input is None or input=='':\n        return PROMPT_DICT['prompt_no_input'].format_map({'instruction': instruction})\n    else:\n        return PROMPT_DICT[\"prompt_input\"].format_map({'instruction': instruction, 'input': input})\n","source_hash":"e4a4149bd7b1c9a320c9eec061ba9666d770f9cf901685563ae4f0486aa245ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.tokenizer","uri":"program://LLaMA-Adapter/module/imagebind_LLM.llama.tokenizer#L1-L40","kind":"module","name":"imagebind_LLM.llama.tokenizer","path":"imagebind_LLM/llama/tokenizer.py","language":"python","start_line":1,"end_line":40,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.tokenizer.Tokenizer","uri":"program://LLaMA-Adapter/class/imagebind_LLM.llama.tokenizer.Tokenizer#L13-L40","kind":"class","name":"Tokenizer","path":"imagebind_LLM/llama/tokenizer.py","language":"python","start_line":13,"end_line":40,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.tokenizer.__init__","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.tokenizer.__init__#L14-L28","kind":"function","name":"__init__","path":"imagebind_LLM/llama/tokenizer.py","language":"python","start_line":14,"end_line":28,"context_start_line":1,"context_end_line":40,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.tokenizer.encode","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.tokenizer.encode#L30-L37","kind":"function","name":"encode","path":"imagebind_LLM/llama/tokenizer.py","language":"python","start_line":30,"end_line":37,"context_start_line":10,"context_end_line":40,"code":"logger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.llama.tokenizer.decode","uri":"program://LLaMA-Adapter/function/imagebind_LLM.llama.tokenizer.decode#L39-L40","kind":"function","name":"decode","path":"imagebind_LLM/llama/tokenizer.py","language":"python","start_line":39,"end_line":40,"context_start_line":19,"context_end_line":40,"code":"\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(\n            f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\"\n        )\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:imagebind_LLM.tools.get_chinese_llama","uri":"program://LLaMA-Adapter/module/imagebind_LLM.tools.get_chinese_llama#L1-L40","kind":"module","name":"imagebind_LLM.tools.get_chinese_llama","path":"imagebind_LLM/tools/get_chinese_llama.py","language":"python","start_line":1,"end_line":40,"context_start_line":1,"context_end_line":40,"code":"# Script for obtaining Chinese LLaMA weights from the OpenChineseLLaMA project (https://github.com/OpenLMLab/OpenChineseLLaMA)\n# Due to the License of LLaMA, we only provide a delta-version patch\n# Adding the patch to the original LLaMA weights makes the Chinese LLaMA weights\nimport os\nimport sys\nsys.path.append(os.path.abspath(__file__).rsplit('/', 2)[0])\nimport shutil\nimport torch\nimport argparse\nfrom util.misc import download\n\nparser = argparse.ArgumentParser()\nparser.add_argument(\n    \"--llama_dir\", default=\"/path/to/llama\", type=str,\n    help=\"Path to official LLaMA weights\",\n)\nargs = parser.parse_args()\n\nori_path = os.path.join(args.llama_dir, '7B')\ndelta_path = os.path.join(args.llama_dir, '7B_chinese_delta')\nnew_path = os.path.join(args.llama_dir, '7B_chinese')\n\n\ndownload('https://huggingface.co/Cxxs/Open-Chinese-LLaMA/resolve/main/7B_chinese_delta/consolidated.00.pth',\n         delta_path)\ndownload('https://huggingface.co/Cxxs/Open-Chinese-LLaMA/resolve/main/7B_chinese_delta/params.json',\n         delta_path)\n\nos.makedirs(new_path, exist_ok=True)\nshutil.copyfile(os.path.join(delta_path, 'params.json'), os.path.join(new_path, 'params.json'))\n\nori_dict = torch.load(os.path.join(ori_path, 'consolidated.00.pth'), map_location=\"cpu\")\ndelta_dict = torch.load(os.path.join(delta_path, 'consolidated.00.pth'), map_location=\"cpu\")\nnew_dict = {}\nfor key in ori_dict:\n    if \"rope\" in key: continue\n    new_value = (ori_dict[key].float() + delta_dict[key].float()).half()\n    new_dict[key] = new_value\n\ntorch.save(new_dict, os.path.join(new_path, 'consolidated.00.pth'))","source_hash":"50b5ab89be0b84d50efc629d1b5d2adb5f3cb5dd09f8ec97bd9ea6f3eab5b10f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.conversation","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_chat65b.conversation#L1-L153","kind":"module","name":"llama_adapter_v2_chat65b.conversation","path":"llama_adapter_v2_chat65b/conversation.py","language":"python","start_line":1,"end_line":153,"context_start_line":1,"context_end_line":153,"code":"import dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Tuple\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Different separator style.\"\"\"\n    SINGLE = auto()\n    TWO = auto()\n\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n    system: str\n    roles: List[str]\n    messages: List[List[str]]\n    offset: int\n    sep_style: SeparatorStyle = SeparatorStyle.SINGLE\n    sep: str = \"###\"\n    sep2: str = None\n\n    skip_next: bool = False\n\n    def get_prompt(self):\n        if self.sep_style == SeparatorStyle.SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n\n    def dict(self):\n        return {\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n            \"sep\": self.sep,\n            \"sep2\": self.sep2,\n        }\n\n\nconv_v1 = Conversation(\n    system=\"A chat between a curious human and an artificial intelligence assistant. \"\n           \"The assistant gives helpful, detailed, and polite answers to the human's questions.\",\n    roles=(\"Human\", \"Assistant\"),\n    messages=(\n        (\"Human\", \"Give three tips for staying healthy.\"),\n        (\"Assistant\",\n            \"Sure, here are three tips for staying healthy:\\n\"\n            \"1. Exercise regularly: Regular physical activity can help improve your overall health and wellbeing. \"\n            \"It can also help reduce your risk of chronic conditions such as obesity, diabetes, heart disease, \"\n            \"and certain cancers. Aim for at least 150 minutes of moderate-intensity aerobic exercise or \"\n            \"75 minutes of vigorous-intensity aerobic exercise per week, along with muscle-strengthening \"\n            \"activities at least two days per week.\\n\"\n            \"2. Eat a balanced diet: Eating a balanced diet that is rich in fruits, \"\n            \"vegetables, whole grains, lean proteins, and healthy fats can help support \"\n            \"your overall health. Try to limit your intake of processed and high-sugar foods, \"\n            \"and aim to drink plenty of water throughout the day.\\n\"\n            \"3. Get enough sleep: Getting enough quality sleep is essential for your physical \"\n            \"and mental health. Adults should aim for seven to nine hours of sleep per night. \"\n            \"Establish a regular sleep schedule and try to create a relaxing bedtime routine to \"\n            \"help improve the quality of your sleep.\")\n    ),\n    offset=2,\n    sep_style=SeparatorStyle.SINGLE,\n    sep=\"###\",\n)\n\nconv_v1_2 = Conversation(\n    system=\"A chat between a curious human and an artificial intelligence assistant. \"\n           \"The assistant gives helpful, detailed, and polite answers to the human's questions.\",\n    roles=(\"Human\", \"Assistant\"),\n    messages=(\n        (\"Human\", \"What are the key differences between renewable and non-renewable energy sources?\"),\n        (\"Assistant\",\n            \"Renewable energy sources are those that can be replenished naturally in a relatively \"\n            \"short amount of time, such as solar, wind, hydro, geothermal, and biomass. \"\n            \"Non-renewable energy sources, on the other hand, are finite and will eventually be \"\n            \"depleted, such as coal, oil, and natural gas. Here are some key differences between \"\n            \"renewable and non-renewable energy sources:\\n\"\n            \"1. Availability: Renewable energy sources are virtually inexhaustible, while non-renewable \"\n            \"energy sources are finite and will eventually run out.\\n\"\n            \"2. Environmental impact: Renewable energy sources have a much lower environmental impact \"\n            \"than non-renewable sources, which can lead to air and water pollution, greenhouse gas emissions, \"\n            \"and other negative effects.\\n\"\n            \"3. Cost: Renewable energy sources can be more expensive to initially set up, but they typically \"\n            \"have lower operational costs than non-renewable sources.\\n\"\n            \"4. Reliability: Renewable energy sources are often more reliable and can be used in more remote \"\n            \"locations than non-renewable sources.\\n\"\n            \"5. Flexibility: Renewable energy sources are often more flexible and can be adapted to different \"\n            \"situations and needs, while non-renewable sources are more rigid and inflexible.\\n\"\n            \"6. Sustainability: Renewable energy sources are more sustainable over the long term, while \"\n            \"non-renewable sources are not, and their depletion can lead to economic and social instability.\\n\")\n    ),\n    offset=2,\n    sep_style=SeparatorStyle.SINGLE,\n    sep=\"###\",\n)\n\nconv_bair_v1 = Conversation(\n    system=\"BEGINNING OF CONVERSATION:\",\n    roles=(\"USER\", \"GPT\"),\n    messages=(),\n    offset=0,\n    sep_style=SeparatorStyle.TWO,\n    sep=\" \",\n    sep2=\"</s>\",\n)\n\n\ndefault_conversation = conv_v1_2\nconv_templates = {\n    \"v1\": conv_v1_2,\n    \"bair_v1\": conv_bair_v1,\n}\n","source_hash":"b49b07dda1d3c118b5df73393388b8ee70c6639a9b2a3cf7e11ad5abda3bb0ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.conversation.SeparatorStyle","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.conversation.SeparatorStyle#L6-L9","kind":"class","name":"SeparatorStyle","path":"llama_adapter_v2_chat65b/conversation.py","language":"python","start_line":6,"end_line":9,"context_start_line":1,"context_end_line":29,"code":"import dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Tuple\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Different separator style.\"\"\"\n    SINGLE = auto()\n    TWO = auto()\n\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n    system: str\n    roles: List[str]\n    messages: List[List[str]]\n    offset: int\n    sep_style: SeparatorStyle = SeparatorStyle.SINGLE\n    sep: str = \"###\"\n    sep2: str = None\n\n    skip_next: bool = False\n\n    def get_prompt(self):\n        if self.sep_style == SeparatorStyle.SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:","source_hash":"b49b07dda1d3c118b5df73393388b8ee70c6639a9b2a3cf7e11ad5abda3bb0ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.conversation.Conversation","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.conversation.Conversation#L13-L76","kind":"class","name":"Conversation","path":"llama_adapter_v2_chat65b/conversation.py","language":"python","start_line":13,"end_line":76,"context_start_line":1,"context_end_line":96,"code":"import dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Tuple\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Different separator style.\"\"\"\n    SINGLE = auto()\n    TWO = auto()\n\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n    system: str\n    roles: List[str]\n    messages: List[List[str]]\n    offset: int\n    sep_style: SeparatorStyle = SeparatorStyle.SINGLE\n    sep: str = \"###\"\n    sep2: str = None\n\n    skip_next: bool = False\n\n    def get_prompt(self):\n        if self.sep_style == SeparatorStyle.SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n\n    def dict(self):\n        return {\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n            \"sep\": self.sep,\n            \"sep2\": self.sep2,\n        }\n\n\nconv_v1 = Conversation(\n    system=\"A chat between a curious human and an artificial intelligence assistant. \"\n           \"The assistant gives helpful, detailed, and polite answers to the human's questions.\",\n    roles=(\"Human\", \"Assistant\"),\n    messages=(\n        (\"Human\", \"Give three tips for staying healthy.\"),\n        (\"Assistant\",\n            \"Sure, here are three tips for staying healthy:\\n\"\n            \"1. Exercise regularly: Regular physical activity can help improve your overall health and wellbeing. \"\n            \"It can also help reduce your risk of chronic conditions such as obesity, diabetes, heart disease, \"\n            \"and certain cancers. Aim for at least 150 minutes of moderate-intensity aerobic exercise or \"\n            \"75 minutes of vigorous-intensity aerobic exercise per week, along with muscle-strengthening \"\n            \"activities at least two days per week.\\n\"\n            \"2. Eat a balanced diet: Eating a balanced diet that is rich in fruits, \"\n            \"vegetables, whole grains, lean proteins, and healthy fats can help support \"\n            \"your overall health. Try to limit your intake of processed and high-sugar foods, \"\n            \"and aim to drink plenty of water throughout the day.\\n\"\n            \"3. Get enough sleep: Getting enough quality sleep is essential for your physical \"","source_hash":"b49b07dda1d3c118b5df73393388b8ee70c6639a9b2a3cf7e11ad5abda3bb0ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.conversation.get_prompt","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.conversation.get_prompt#L25-L44","kind":"function","name":"get_prompt","path":"llama_adapter_v2_chat65b/conversation.py","language":"python","start_line":25,"end_line":44,"context_start_line":5,"context_end_line":64,"code":"\nclass SeparatorStyle(Enum):\n    \"\"\"Different separator style.\"\"\"\n    SINGLE = auto()\n    TWO = auto()\n\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n    system: str\n    roles: List[str]\n    messages: List[List[str]]\n    offset: int\n    sep_style: SeparatorStyle = SeparatorStyle.SINGLE\n    sep: str = \"###\"\n    sep2: str = None\n\n    skip_next: bool = False\n\n    def get_prompt(self):\n        if self.sep_style == SeparatorStyle.SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,","source_hash":"b49b07dda1d3c118b5df73393388b8ee70c6639a9b2a3cf7e11ad5abda3bb0ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.conversation.append_message","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.conversation.append_message#L46-L47","kind":"function","name":"append_message","path":"llama_adapter_v2_chat65b/conversation.py","language":"python","start_line":46,"end_line":47,"context_start_line":26,"context_end_line":67,"code":"        if self.sep_style == SeparatorStyle.SINGLE:\n            ret = self.system + self.sep\n            for role, message in self.messages:\n                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n","source_hash":"b49b07dda1d3c118b5df73393388b8ee70c6639a9b2a3cf7e11ad5abda3bb0ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.conversation.to_gradio_chatbot","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.conversation.to_gradio_chatbot#L49-L56","kind":"function","name":"to_gradio_chatbot","path":"llama_adapter_v2_chat65b/conversation.py","language":"python","start_line":49,"end_line":56,"context_start_line":29,"context_end_line":76,"code":"                if message:\n                    ret += role + \": \" + message + self.sep\n                else:\n                    ret += role + \":\"\n            return ret\n        elif self.sep_style == SeparatorStyle.TWO:\n            seps = [self.sep, self.sep2]\n            ret = self.system + seps[0]\n            for i, (role, message) in enumerate(self.messages):\n                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n\n    def dict(self):\n        return {\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n            \"sep\": self.sep,\n            \"sep2\": self.sep2,\n        }","source_hash":"b49b07dda1d3c118b5df73393388b8ee70c6639a9b2a3cf7e11ad5abda3bb0ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.conversation.copy","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.conversation.copy#L58-L66","kind":"function","name":"copy","path":"llama_adapter_v2_chat65b/conversation.py","language":"python","start_line":58,"end_line":66,"context_start_line":38,"context_end_line":86,"code":"                if message:\n                    ret += role + \": \" + message + seps[i % 2]\n                else:\n                    ret += role + \":\"\n            return ret\n        else:\n            raise ValueError(f\"Invalid style: {self.sep_style}\")\n\n    def append_message(self, role, message):\n        self.messages.append([role, message])\n\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n\n    def dict(self):\n        return {\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n            \"sep\": self.sep,\n            \"sep2\": self.sep2,\n        }\n\n\nconv_v1 = Conversation(\n    system=\"A chat between a curious human and an artificial intelligence assistant. \"\n           \"The assistant gives helpful, detailed, and polite answers to the human's questions.\",\n    roles=(\"Human\", \"Assistant\"),\n    messages=(\n        (\"Human\", \"Give three tips for staying healthy.\"),\n        (\"Assistant\",\n            \"Sure, here are three tips for staying healthy:\\n\"","source_hash":"b49b07dda1d3c118b5df73393388b8ee70c6639a9b2a3cf7e11ad5abda3bb0ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.conversation.dict","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.conversation.dict#L68-L76","kind":"function","name":"dict","path":"llama_adapter_v2_chat65b/conversation.py","language":"python","start_line":68,"end_line":76,"context_start_line":48,"context_end_line":96,"code":"\n    def to_gradio_chatbot(self):\n        ret = []\n        for i, (role, msg) in enumerate(self.messages[self.offset:]):\n            if i % 2 == 0:\n                ret.append([msg, None])\n            else:\n                ret[-1][-1] = msg\n        return ret\n\n    def copy(self):\n        return Conversation(\n            system=self.system,\n            roles=self.roles,\n            messages=[[x, y] for x, y in self.messages],\n            offset=self.offset,\n            sep_style=self.sep_style,\n            sep=self.sep,\n            sep2=self.sep2)\n\n    def dict(self):\n        return {\n            \"system\": self.system,\n            \"roles\": self.roles,\n            \"messages\": self.messages,\n            \"offset\": self.offset,\n            \"sep\": self.sep,\n            \"sep2\": self.sep2,\n        }\n\n\nconv_v1 = Conversation(\n    system=\"A chat between a curious human and an artificial intelligence assistant. \"\n           \"The assistant gives helpful, detailed, and polite answers to the human's questions.\",\n    roles=(\"Human\", \"Assistant\"),\n    messages=(\n        (\"Human\", \"Give three tips for staying healthy.\"),\n        (\"Assistant\",\n            \"Sure, here are three tips for staying healthy:\\n\"\n            \"1. Exercise regularly: Regular physical activity can help improve your overall health and wellbeing. \"\n            \"It can also help reduce your risk of chronic conditions such as obesity, diabetes, heart disease, \"\n            \"and certain cancers. Aim for at least 150 minutes of moderate-intensity aerobic exercise or \"\n            \"75 minutes of vigorous-intensity aerobic exercise per week, along with muscle-strengthening \"\n            \"activities at least two days per week.\\n\"\n            \"2. Eat a balanced diet: Eating a balanced diet that is rich in fruits, \"\n            \"vegetables, whole grains, lean proteins, and healthy fats can help support \"\n            \"your overall health. Try to limit your intake of processed and high-sugar foods, \"\n            \"and aim to drink plenty of water throughout the day.\\n\"\n            \"3. Get enough sleep: Getting enough quality sleep is essential for your physical \"","source_hash":"b49b07dda1d3c118b5df73393388b8ee70c6639a9b2a3cf7e11ad5abda3bb0ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.chat_demo","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_chat65b.chat_demo#L1-L141","kind":"module","name":"llama_adapter_v2_chat65b.chat_demo","path":"llama_adapter_v2_chat65b/chat_demo.py","language":"python","start_line":1,"end_line":141,"context_start_line":1,"context_end_line":141,"code":"import argparse\nimport os\nimport sys\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport models_llama_adapter\nimport torch\nimport torch.distributed as dist\nfrom conversation import conv_templates, SeparatorStyle\nfrom util import misc\n\nfrom llama import LLaMA, Tokenizer\n\n\ndef load_model(args, load_8bit=False):\n    model = models_llama_adapter.__dict__[args.model_name](args)\n    model.eval()\n    if args.model_path is None:\n        print(\"Warning: not loading instruct tuned weights.\")\n    else:\n        print(\"Using instruct tuned weights from:\", args.model_path)\n        checkpoint = torch.load(args.model_path, map_location=\"cpu\")\n        for k, v in checkpoint[\"model\"].items():\n            if (\n                k.endswith(\".wq_bias\")\n                or k.endswith(\".wk_bias\")\n                or k.endswith(\".wv_bias\")\n                or k.endswith(\".wo_scale\")\n                or k.endswith(\".w1_bias\")\n                or k.endswith(\".w3_bias\")\n                or k.endswith(\".w2_scale\")\n            ):\n                assert v.ndim == 1\n                mp_size = fs_init.get_model_parallel_world_size()\n                mp_rank = fs_init.get_model_parallel_rank()\n                shard_size = v.size(0) // mp_size\n                checkpoint[\"model\"][k] = v[shard_size * mp_rank : shard_size * (mp_rank + 1)]\n        missing_keys, unexpected_keys = model.load_state_dict(checkpoint[\"model\"], strict=False)\n\n    generator = LLaMA(\n        model,\n        Tokenizer(model_path=os.path.join(args.llama_model_path, \"tokenizer.model\")),\n    )\n\n    return generator\n\n\n@torch.inference_mode()\ndef generate_stream(model, params):\n    \"\"\"Adapted from fastchat/serve/model_worker.py::generate_stream\"\"\"\n\n    prompt = params[\"prompt\"]\n    len(prompt)\n    temperature = float(params.get(\"temperature\", 1.0))\n    top_p = float(params.get(\"top_p\", 0.95))\n    max_new_tokens = int(params.get(\"max_new_tokens\", 256))\n    stop_str = params.get(\"stop\", None)\n    with torch.cuda.amp.autocast():\n        decoded = model.generate(\n            [prompt],\n            max_gen_len=max_new_tokens,\n            top_p=top_p,\n            temperature=temperature,\n        )\n    decoded = decoded[0]\n\n    pos = decoded.find(stop_str)\n    if pos != -1:\n        decoded = decoded[:pos]\n\n    return decoded\n\n\ndef main(args):\n    misc.init_distributed_mode()\n    fs_init.initialize_model_parallel(dist.get_world_size())\n    torch.manual_seed(1)\n\n    # Model\n    model = load_model(args)\n\n    # Chat\n    conv = conv_templates[args.conv_template].copy()\n    while True:\n        if dist.get_rank() == 0:\n            try:\n                sys.stdout.write(f\"\\n{conv.roles[0]}: \")\n                sys.stdout.flush()\n                inp = input()\n            except EOFError:\n                inp = \"\"\n            dist.broadcast_object_list([inp], src=0)\n        else:\n            recv_obj = [None]\n            dist.broadcast_object_list(recv_obj, src=0)\n            inp = recv_obj[0]\n\n        if not inp:\n            print(\"exit...\")\n            break\n\n        conv.append_message(conv.roles[0], inp)\n        conv.append_message(conv.roles[1], None)\n        prompt = conv.get_prompt()\n\n        params = {\n            \"model\": \"LLaMA-Adapter\",\n            \"prompt\": prompt,\n            \"temperature\": args.temperature,\n            \"top_p\": args.top_p,\n            \"max_new_tokens\": args.max_new_tokens,\n            \"stop\": conv.sep if conv.sep_style == SeparatorStyle.SINGLE else conv.sep2,\n        }\n\n        if dist.get_rank() == 0:\n            sys.stdout.write(f\"{conv.roles[1]}: \")\n            sys.stdout.flush()\n        outputs = generate_stream(model, params)\n        outputs = outputs.strip()\n        if dist.get_rank() == 0:\n            sys.stdout.write(outputs + \"\\n\")\n            sys.stdout.flush()\n\n        conv.messages[-1][-1] = outputs\n\n        if args.debug:\n            print(\"\\n\", {\"prompt\": prompt, \"outputs\": outputs}, \"\\n\")\n\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--model_name\", type=str, required=True)\n    parser.add_argument(\"--model_path\", type=str)\n    parser.add_argument(\"--llama_model_path\", type=str, required=True)\n    parser.add_argument(\"--conv_template\", type=str, default=\"v1\")\n    parser.add_argument(\"--temperature\", type=float, default=0.7)\n    parser.add_argument(\"--max_new_tokens\", type=int, default=512)\n    parser.add_argument(\"--top_p\", type=float, default=0.95)\n    parser.add_argument(\"--debug\", action=\"store_true\")\n    args = parser.parse_args()\n    main(args)","source_hash":"6220deba345664385d054b284402ca14848a15aa842d58c5ac568bf678052946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.chat_demo.load_model","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.chat_demo.load_model#L15-L45","kind":"function","name":"load_model","path":"llama_adapter_v2_chat65b/chat_demo.py","language":"python","start_line":15,"end_line":45,"context_start_line":1,"context_end_line":65,"code":"import argparse\nimport os\nimport sys\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport models_llama_adapter\nimport torch\nimport torch.distributed as dist\nfrom conversation import conv_templates, SeparatorStyle\nfrom util import misc\n\nfrom llama import LLaMA, Tokenizer\n\n\ndef load_model(args, load_8bit=False):\n    model = models_llama_adapter.__dict__[args.model_name](args)\n    model.eval()\n    if args.model_path is None:\n        print(\"Warning: not loading instruct tuned weights.\")\n    else:\n        print(\"Using instruct tuned weights from:\", args.model_path)\n        checkpoint = torch.load(args.model_path, map_location=\"cpu\")\n        for k, v in checkpoint[\"model\"].items():\n            if (\n                k.endswith(\".wq_bias\")\n                or k.endswith(\".wk_bias\")\n                or k.endswith(\".wv_bias\")\n                or k.endswith(\".wo_scale\")\n                or k.endswith(\".w1_bias\")\n                or k.endswith(\".w3_bias\")\n                or k.endswith(\".w2_scale\")\n            ):\n                assert v.ndim == 1\n                mp_size = fs_init.get_model_parallel_world_size()\n                mp_rank = fs_init.get_model_parallel_rank()\n                shard_size = v.size(0) // mp_size\n                checkpoint[\"model\"][k] = v[shard_size * mp_rank : shard_size * (mp_rank + 1)]\n        missing_keys, unexpected_keys = model.load_state_dict(checkpoint[\"model\"], strict=False)\n\n    generator = LLaMA(\n        model,\n        Tokenizer(model_path=os.path.join(args.llama_model_path, \"tokenizer.model\")),\n    )\n\n    return generator\n\n\n@torch.inference_mode()\ndef generate_stream(model, params):\n    \"\"\"Adapted from fastchat/serve/model_worker.py::generate_stream\"\"\"\n\n    prompt = params[\"prompt\"]\n    len(prompt)\n    temperature = float(params.get(\"temperature\", 1.0))\n    top_p = float(params.get(\"top_p\", 0.95))\n    max_new_tokens = int(params.get(\"max_new_tokens\", 256))\n    stop_str = params.get(\"stop\", None)\n    with torch.cuda.amp.autocast():\n        decoded = model.generate(\n            [prompt],\n            max_gen_len=max_new_tokens,\n            top_p=top_p,\n            temperature=temperature,\n        )\n    decoded = decoded[0]","source_hash":"6220deba345664385d054b284402ca14848a15aa842d58c5ac568bf678052946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.chat_demo.generate_stream","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.chat_demo.generate_stream#L49-L71","kind":"function","name":"generate_stream","path":"llama_adapter_v2_chat65b/chat_demo.py","language":"python","start_line":49,"end_line":71,"context_start_line":29,"context_end_line":91,"code":"                or k.endswith(\".w1_bias\")\n                or k.endswith(\".w3_bias\")\n                or k.endswith(\".w2_scale\")\n            ):\n                assert v.ndim == 1\n                mp_size = fs_init.get_model_parallel_world_size()\n                mp_rank = fs_init.get_model_parallel_rank()\n                shard_size = v.size(0) // mp_size\n                checkpoint[\"model\"][k] = v[shard_size * mp_rank : shard_size * (mp_rank + 1)]\n        missing_keys, unexpected_keys = model.load_state_dict(checkpoint[\"model\"], strict=False)\n\n    generator = LLaMA(\n        model,\n        Tokenizer(model_path=os.path.join(args.llama_model_path, \"tokenizer.model\")),\n    )\n\n    return generator\n\n\n@torch.inference_mode()\ndef generate_stream(model, params):\n    \"\"\"Adapted from fastchat/serve/model_worker.py::generate_stream\"\"\"\n\n    prompt = params[\"prompt\"]\n    len(prompt)\n    temperature = float(params.get(\"temperature\", 1.0))\n    top_p = float(params.get(\"top_p\", 0.95))\n    max_new_tokens = int(params.get(\"max_new_tokens\", 256))\n    stop_str = params.get(\"stop\", None)\n    with torch.cuda.amp.autocast():\n        decoded = model.generate(\n            [prompt],\n            max_gen_len=max_new_tokens,\n            top_p=top_p,\n            temperature=temperature,\n        )\n    decoded = decoded[0]\n\n    pos = decoded.find(stop_str)\n    if pos != -1:\n        decoded = decoded[:pos]\n\n    return decoded\n\n\ndef main(args):\n    misc.init_distributed_mode()\n    fs_init.initialize_model_parallel(dist.get_world_size())\n    torch.manual_seed(1)\n\n    # Model\n    model = load_model(args)\n\n    # Chat\n    conv = conv_templates[args.conv_template].copy()\n    while True:\n        if dist.get_rank() == 0:\n            try:\n                sys.stdout.write(f\"\\n{conv.roles[0]}: \")\n                sys.stdout.flush()\n                inp = input()\n            except EOFError:\n                inp = \"\"","source_hash":"6220deba345664385d054b284402ca14848a15aa842d58c5ac568bf678052946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.chat_demo.main","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.chat_demo.main#L74-L127","kind":"function","name":"main","path":"llama_adapter_v2_chat65b/chat_demo.py","language":"python","start_line":74,"end_line":127,"context_start_line":54,"context_end_line":141,"code":"    temperature = float(params.get(\"temperature\", 1.0))\n    top_p = float(params.get(\"top_p\", 0.95))\n    max_new_tokens = int(params.get(\"max_new_tokens\", 256))\n    stop_str = params.get(\"stop\", None)\n    with torch.cuda.amp.autocast():\n        decoded = model.generate(\n            [prompt],\n            max_gen_len=max_new_tokens,\n            top_p=top_p,\n            temperature=temperature,\n        )\n    decoded = decoded[0]\n\n    pos = decoded.find(stop_str)\n    if pos != -1:\n        decoded = decoded[:pos]\n\n    return decoded\n\n\ndef main(args):\n    misc.init_distributed_mode()\n    fs_init.initialize_model_parallel(dist.get_world_size())\n    torch.manual_seed(1)\n\n    # Model\n    model = load_model(args)\n\n    # Chat\n    conv = conv_templates[args.conv_template].copy()\n    while True:\n        if dist.get_rank() == 0:\n            try:\n                sys.stdout.write(f\"\\n{conv.roles[0]}: \")\n                sys.stdout.flush()\n                inp = input()\n            except EOFError:\n                inp = \"\"\n            dist.broadcast_object_list([inp], src=0)\n        else:\n            recv_obj = [None]\n            dist.broadcast_object_list(recv_obj, src=0)\n            inp = recv_obj[0]\n\n        if not inp:\n            print(\"exit...\")\n            break\n\n        conv.append_message(conv.roles[0], inp)\n        conv.append_message(conv.roles[1], None)\n        prompt = conv.get_prompt()\n\n        params = {\n            \"model\": \"LLaMA-Adapter\",\n            \"prompt\": prompt,\n            \"temperature\": args.temperature,\n            \"top_p\": args.top_p,\n            \"max_new_tokens\": args.max_new_tokens,\n            \"stop\": conv.sep if conv.sep_style == SeparatorStyle.SINGLE else conv.sep2,\n        }\n\n        if dist.get_rank() == 0:\n            sys.stdout.write(f\"{conv.roles[1]}: \")\n            sys.stdout.flush()\n        outputs = generate_stream(model, params)\n        outputs = outputs.strip()\n        if dist.get_rank() == 0:\n            sys.stdout.write(outputs + \"\\n\")\n            sys.stdout.flush()\n\n        conv.messages[-1][-1] = outputs\n\n        if args.debug:\n            print(\"\\n\", {\"prompt\": prompt, \"outputs\": outputs}, \"\\n\")\n\n\nif __name__ == \"__main__\":\n    parser = argparse.ArgumentParser()\n    parser.add_argument(\"--model_name\", type=str, required=True)\n    parser.add_argument(\"--model_path\", type=str)\n    parser.add_argument(\"--llama_model_path\", type=str, required=True)\n    parser.add_argument(\"--conv_template\", type=str, default=\"v1\")\n    parser.add_argument(\"--temperature\", type=float, default=0.7)\n    parser.add_argument(\"--max_new_tokens\", type=int, default=512)\n    parser.add_argument(\"--top_p\", type=float, default=0.95)\n    parser.add_argument(\"--debug\", action=\"store_true\")\n    args = parser.parse_args()\n    main(args)","source_hash":"6220deba345664385d054b284402ca14848a15aa842d58c5ac568bf678052946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.models_llama_adapter","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_chat65b.models_llama_adapter#L1-L163","kind":"module","name":"llama_adapter_v2_chat65b.models_llama_adapter","path":"llama_adapter_v2_chat65b/models_llama_adapter.py","language":"python","start_line":1,"end_line":163,"context_start_line":1,"context_end_line":163,"code":"import functools\nimport json\nfrom pathlib import Path\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.distributed as dist\n\nfrom llama import ModelArgs, Tokenizer, Transformer\n\n\ndef _load_and_redistribute_checkpoint(llama_model_path, model_name):\n    with open(Path(llama_model_path) / model_name / \"params.json\") as f:\n        params = json.load(f)\n    tokenizer = Tokenizer(model_path=str(Path(llama_model_path) / \"tokenizer.model\"))\n    print(\"Using model path: %s, model_name: %s\" % (llama_model_path, model_name))\n\n    checkpoints = (Path(llama_model_path) / model_name).glob(\"*.pth\")\n    checkpoints = sorted(checkpoints)\n\n    mp_world_size = fs_init.get_model_parallel_world_size()\n    mp_rank = fs_init.get_model_parallel_rank()\n    if mp_world_size == len(checkpoints):\n        print(\"same number of shards of checkpoints and training, loading directly...\")\n        dist.barrier()\n        print(\"[rank=%d, mp_rank=%d] loading from %s\" % (dist.get_rank(), mp_rank, checkpoints[mp_rank]), force=True)\n        checkpoint = torch.load(checkpoints[mp_rank], map_location=\"cpu\")\n    else:\n        print(\"different number of shards of checkpoints and training, redistributing...\")\n        if dist.get_rank() == 0:\n            loaded = []\n            for x in checkpoints:\n                print(\"loading from\", x)\n                loaded.append(torch.load(x, map_location=\"cpu\"))\n\n            full_state_dict = {}\n            split_dims = {}\n\n            def add_weight_with_split_dim(name, dim):\n                if dim < 0:  # bcast without split\n                    full_state_dict[name] = loaded[0][name].clone()\n                else:\n                    full_state_dict[name] = torch.cat([x[name] for x in loaded], dim=dim)\n                for x in loaded:\n                    del x[name]\n                split_dims[name] = dim\n\n            add_weight_with_split_dim(\"tok_embeddings.weight\", 1)\n            add_weight_with_split_dim(\"norm.weight\", -1)\n            add_weight_with_split_dim(\"output.weight\", 0)\n            for i in range(params[\"n_layers\"]):\n                print(\"gathering layer %d of %d\" % (i, params[\"n_layers\"]))\n                layer_prefix = f\"layers.{i}.\"\n                bcast_names = [\n                    \"attention_norm.weight\",\n                    \"ffn_norm.weight\",\n                ]\n                column_parallel_names = [\n                    \"attention.wq.weight\",\n                    \"attention.wk.weight\",\n                    \"attention.wv.weight\",\n                    \"feed_forward.w1.weight\",\n                    \"feed_forward.w3.weight\",\n                ]\n                row_parallel_names = [\n                    \"attention.wo.weight\",\n                    \"feed_forward.w2.weight\",\n                ]\n                for key in bcast_names:\n                    add_weight_with_split_dim(layer_prefix + key, -1)\n                for key in column_parallel_names:\n                    add_weight_with_split_dim(layer_prefix + key, 0)\n                for key in row_parallel_names:\n                    add_weight_with_split_dim(layer_prefix + key, 1)\n\n            full_state_dict_meta = dict((k, v.shape) for k, v in full_state_dict.items())\n            dist.broadcast_object_list([full_state_dict_meta, split_dims], src=0)\n\n        else:  # dist.get_rank() != 0\n            recv_objs = [None, None]\n            dist.broadcast_object_list(recv_objs, src=0)\n            full_state_dict_meta, split_dims = recv_objs\n\n        local_state_dict = {}\n        for k in sorted(full_state_dict_meta.keys()):\n            print(\"redistributing weights: %s\" % k)\n            if dist.get_rank() == 0:\n                value = full_state_dict[k].cuda().half()\n                del full_state_dict[k]\n            else:\n                value = torch.empty(full_state_dict_meta[k], device=\"cuda\", dtype=torch.half)\n            dist.broadcast(value, src=0)\n            value = value.cpu()\n            if split_dims[k] < 0:\n                local_state_dict[k] = value\n            else:\n                dim = split_dims[k]\n                assert dim >= 0 and dim < value.ndim and value.size(dim) % mp_world_size == 0\n                shard_size = value.size(dim) // mp_world_size\n                shard_st, shard_ed = shard_size * mp_rank, shard_size * (mp_rank + 1)\n                # TODO: make more general\n                if dim == 0:\n                    value = value[shard_st:shard_ed]\n                elif dim == 1:\n                    value = value[:, shard_st:shard_ed]\n                else:\n                    raise NotImplementedError()\n                local_state_dict[k] = value.clone()\n\n        checkpoint = local_state_dict\n\n    return checkpoint, tokenizer, params\n\n\ndef Llama_adapter(\n    args, model_name, adapter_len=0, adapter_layer=0, add_bias=False, add_scale=False, train_norm=False, **kwargs\n):\n    checkpoint, tokenizer, params = _load_and_redistribute_checkpoint(args.llama_model_path, model_name)\n\n    model_args: ModelArgs = ModelArgs(\n        # caching configuration\n        max_seq_len=args.max_seq_len if hasattr(args, \"max_seq_len\") else 2048,\n        adapter_len=adapter_len,\n        adapter_layer=adapter_layer,\n        add_bias=add_bias,\n        add_scale=add_scale,\n        train_norm=train_norm,\n        # other args\n        **params,\n    )\n\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model_llama_adapter = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    missing_keys, unexpected_keys = model_llama_adapter.load_state_dict(checkpoint, strict=False)\n\n    for i in range(model_args.n_layers):\n        if i < model_args.n_layers - adapter_layer:\n            del model_llama_adapter.layers[i].attention.gate\n\n    for name, param in model_llama_adapter.named_parameters():\n        requires_grad = (\n            name.endswith(\".gate\")\n            or name == \"adapter_query\"\n            or (train_norm and \"_norm.\" in name)\n            or name.endswith(\".added_bias\")\n            or name.endswith(\".added_scale\")\n        )\n\n        if requires_grad:\n            param.data = param.data.float()\n            param.requires_grad_(True)\n        else:\n            param.requires_grad_(False)\n\n    return model_llama_adapter\n\n\n# set recommended archs\nLlama65B_bias_scale_norm_tuning = functools.partial(\n    Llama_adapter, model_name=\"65B\", add_bias=True, add_scale=True, train_norm=True\n)","source_hash":"febc64422e500736302f0044a0038e15ccd6a033ca0713a0b0b4e848a839caf8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.models_llama_adapter._load_and_redistribute_checkpoint","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.models_llama_adapter._load_and_redistribute_checkpoint#L12-L112","kind":"function","name":"_load_and_redistribute_checkpoint","path":"llama_adapter_v2_chat65b/models_llama_adapter.py","language":"python","start_line":12,"end_line":112,"context_start_line":1,"context_end_line":132,"code":"import functools\nimport json\nfrom pathlib import Path\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.distributed as dist\n\nfrom llama import ModelArgs, Tokenizer, Transformer\n\n\ndef _load_and_redistribute_checkpoint(llama_model_path, model_name):\n    with open(Path(llama_model_path) / model_name / \"params.json\") as f:\n        params = json.load(f)\n    tokenizer = Tokenizer(model_path=str(Path(llama_model_path) / \"tokenizer.model\"))\n    print(\"Using model path: %s, model_name: %s\" % (llama_model_path, model_name))\n\n    checkpoints = (Path(llama_model_path) / model_name).glob(\"*.pth\")\n    checkpoints = sorted(checkpoints)\n\n    mp_world_size = fs_init.get_model_parallel_world_size()\n    mp_rank = fs_init.get_model_parallel_rank()\n    if mp_world_size == len(checkpoints):\n        print(\"same number of shards of checkpoints and training, loading directly...\")\n        dist.barrier()\n        print(\"[rank=%d, mp_rank=%d] loading from %s\" % (dist.get_rank(), mp_rank, checkpoints[mp_rank]), force=True)\n        checkpoint = torch.load(checkpoints[mp_rank], map_location=\"cpu\")\n    else:\n        print(\"different number of shards of checkpoints and training, redistributing...\")\n        if dist.get_rank() == 0:\n            loaded = []\n            for x in checkpoints:\n                print(\"loading from\", x)\n                loaded.append(torch.load(x, map_location=\"cpu\"))\n\n            full_state_dict = {}\n            split_dims = {}\n\n            def add_weight_with_split_dim(name, dim):\n                if dim < 0:  # bcast without split\n                    full_state_dict[name] = loaded[0][name].clone()\n                else:\n                    full_state_dict[name] = torch.cat([x[name] for x in loaded], dim=dim)\n                for x in loaded:\n                    del x[name]\n                split_dims[name] = dim\n\n            add_weight_with_split_dim(\"tok_embeddings.weight\", 1)\n            add_weight_with_split_dim(\"norm.weight\", -1)\n            add_weight_with_split_dim(\"output.weight\", 0)\n            for i in range(params[\"n_layers\"]):\n                print(\"gathering layer %d of %d\" % (i, params[\"n_layers\"]))\n                layer_prefix = f\"layers.{i}.\"\n                bcast_names = [\n                    \"attention_norm.weight\",\n                    \"ffn_norm.weight\",\n                ]\n                column_parallel_names = [\n                    \"attention.wq.weight\",\n                    \"attention.wk.weight\",\n                    \"attention.wv.weight\",\n                    \"feed_forward.w1.weight\",\n                    \"feed_forward.w3.weight\",\n                ]\n                row_parallel_names = [\n                    \"attention.wo.weight\",\n                    \"feed_forward.w2.weight\",\n                ]\n                for key in bcast_names:\n                    add_weight_with_split_dim(layer_prefix + key, -1)\n                for key in column_parallel_names:\n                    add_weight_with_split_dim(layer_prefix + key, 0)\n                for key in row_parallel_names:\n                    add_weight_with_split_dim(layer_prefix + key, 1)\n\n            full_state_dict_meta = dict((k, v.shape) for k, v in full_state_dict.items())\n            dist.broadcast_object_list([full_state_dict_meta, split_dims], src=0)\n\n        else:  # dist.get_rank() != 0\n            recv_objs = [None, None]\n            dist.broadcast_object_list(recv_objs, src=0)\n            full_state_dict_meta, split_dims = recv_objs\n\n        local_state_dict = {}\n        for k in sorted(full_state_dict_meta.keys()):\n            print(\"redistributing weights: %s\" % k)\n            if dist.get_rank() == 0:\n                value = full_state_dict[k].cuda().half()\n                del full_state_dict[k]\n            else:\n                value = torch.empty(full_state_dict_meta[k], device=\"cuda\", dtype=torch.half)\n            dist.broadcast(value, src=0)\n            value = value.cpu()\n            if split_dims[k] < 0:\n                local_state_dict[k] = value\n            else:\n                dim = split_dims[k]\n                assert dim >= 0 and dim < value.ndim and value.size(dim) % mp_world_size == 0\n                shard_size = value.size(dim) // mp_world_size\n                shard_st, shard_ed = shard_size * mp_rank, shard_size * (mp_rank + 1)\n                # TODO: make more general\n                if dim == 0:\n                    value = value[shard_st:shard_ed]\n                elif dim == 1:\n                    value = value[:, shard_st:shard_ed]\n                else:\n                    raise NotImplementedError()\n                local_state_dict[k] = value.clone()\n\n        checkpoint = local_state_dict\n\n    return checkpoint, tokenizer, params\n\n\ndef Llama_adapter(\n    args, model_name, adapter_len=0, adapter_layer=0, add_bias=False, add_scale=False, train_norm=False, **kwargs\n):\n    checkpoint, tokenizer, params = _load_and_redistribute_checkpoint(args.llama_model_path, model_name)\n\n    model_args: ModelArgs = ModelArgs(\n        # caching configuration\n        max_seq_len=args.max_seq_len if hasattr(args, \"max_seq_len\") else 2048,\n        adapter_len=adapter_len,\n        adapter_layer=adapter_layer,\n        add_bias=add_bias,\n        add_scale=add_scale,\n        train_norm=train_norm,\n        # other args\n        **params,\n    )\n\n    model_args.vocab_size = tokenizer.n_words","source_hash":"febc64422e500736302f0044a0038e15ccd6a033ca0713a0b0b4e848a839caf8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.models_llama_adapter.Llama_adapter","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.models_llama_adapter.Llama_adapter#L115-L157","kind":"function","name":"Llama_adapter","path":"llama_adapter_v2_chat65b/models_llama_adapter.py","language":"python","start_line":115,"end_line":157,"context_start_line":95,"context_end_line":163,"code":"                local_state_dict[k] = value\n            else:\n                dim = split_dims[k]\n                assert dim >= 0 and dim < value.ndim and value.size(dim) % mp_world_size == 0\n                shard_size = value.size(dim) // mp_world_size\n                shard_st, shard_ed = shard_size * mp_rank, shard_size * (mp_rank + 1)\n                # TODO: make more general\n                if dim == 0:\n                    value = value[shard_st:shard_ed]\n                elif dim == 1:\n                    value = value[:, shard_st:shard_ed]\n                else:\n                    raise NotImplementedError()\n                local_state_dict[k] = value.clone()\n\n        checkpoint = local_state_dict\n\n    return checkpoint, tokenizer, params\n\n\ndef Llama_adapter(\n    args, model_name, adapter_len=0, adapter_layer=0, add_bias=False, add_scale=False, train_norm=False, **kwargs\n):\n    checkpoint, tokenizer, params = _load_and_redistribute_checkpoint(args.llama_model_path, model_name)\n\n    model_args: ModelArgs = ModelArgs(\n        # caching configuration\n        max_seq_len=args.max_seq_len if hasattr(args, \"max_seq_len\") else 2048,\n        adapter_len=adapter_len,\n        adapter_layer=adapter_layer,\n        add_bias=add_bias,\n        add_scale=add_scale,\n        train_norm=train_norm,\n        # other args\n        **params,\n    )\n\n    model_args.vocab_size = tokenizer.n_words\n    torch.set_default_tensor_type(torch.cuda.HalfTensor)\n    model_llama_adapter = Transformer(model_args)\n    torch.set_default_tensor_type(torch.FloatTensor)\n    missing_keys, unexpected_keys = model_llama_adapter.load_state_dict(checkpoint, strict=False)\n\n    for i in range(model_args.n_layers):\n        if i < model_args.n_layers - adapter_layer:\n            del model_llama_adapter.layers[i].attention.gate\n\n    for name, param in model_llama_adapter.named_parameters():\n        requires_grad = (\n            name.endswith(\".gate\")\n            or name == \"adapter_query\"\n            or (train_norm and \"_norm.\" in name)\n            or name.endswith(\".added_bias\")\n            or name.endswith(\".added_scale\")\n        )\n\n        if requires_grad:\n            param.data = param.data.float()\n            param.requires_grad_(True)\n        else:\n            param.requires_grad_(False)\n\n    return model_llama_adapter\n\n\n# set recommended archs\nLlama65B_bias_scale_norm_tuning = functools.partial(\n    Llama_adapter, model_name=\"65B\", add_bias=True, add_scale=True, train_norm=True\n)","source_hash":"febc64422e500736302f0044a0038e15ccd6a033ca0713a0b0b4e848a839caf8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.models_llama_adapter.add_weight_with_split_dim","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.models_llama_adapter.add_weight_with_split_dim#L39-L46","kind":"function","name":"add_weight_with_split_dim","path":"llama_adapter_v2_chat65b/models_llama_adapter.py","language":"python","start_line":39,"end_line":46,"context_start_line":19,"context_end_line":66,"code":"    checkpoints = sorted(checkpoints)\n\n    mp_world_size = fs_init.get_model_parallel_world_size()\n    mp_rank = fs_init.get_model_parallel_rank()\n    if mp_world_size == len(checkpoints):\n        print(\"same number of shards of checkpoints and training, loading directly...\")\n        dist.barrier()\n        print(\"[rank=%d, mp_rank=%d] loading from %s\" % (dist.get_rank(), mp_rank, checkpoints[mp_rank]), force=True)\n        checkpoint = torch.load(checkpoints[mp_rank], map_location=\"cpu\")\n    else:\n        print(\"different number of shards of checkpoints and training, redistributing...\")\n        if dist.get_rank() == 0:\n            loaded = []\n            for x in checkpoints:\n                print(\"loading from\", x)\n                loaded.append(torch.load(x, map_location=\"cpu\"))\n\n            full_state_dict = {}\n            split_dims = {}\n\n            def add_weight_with_split_dim(name, dim):\n                if dim < 0:  # bcast without split\n                    full_state_dict[name] = loaded[0][name].clone()\n                else:\n                    full_state_dict[name] = torch.cat([x[name] for x in loaded], dim=dim)\n                for x in loaded:\n                    del x[name]\n                split_dims[name] = dim\n\n            add_weight_with_split_dim(\"tok_embeddings.weight\", 1)\n            add_weight_with_split_dim(\"norm.weight\", -1)\n            add_weight_with_split_dim(\"output.weight\", 0)\n            for i in range(params[\"n_layers\"]):\n                print(\"gathering layer %d of %d\" % (i, params[\"n_layers\"]))\n                layer_prefix = f\"layers.{i}.\"\n                bcast_names = [\n                    \"attention_norm.weight\",\n                    \"ffn_norm.weight\",\n                ]\n                column_parallel_names = [\n                    \"attention.wq.weight\",\n                    \"attention.wk.weight\",\n                    \"attention.wv.weight\",\n                    \"feed_forward.w1.weight\",\n                    \"feed_forward.w3.weight\",\n                ]\n                row_parallel_names = [\n                    \"attention.wo.weight\",","source_hash":"febc64422e500736302f0044a0038e15ccd6a033ca0713a0b0b4e848a839caf8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_chat65b.util.misc#L1-L357","kind":"module","name":"llama_adapter_v2_chat65b.util.misc","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":1,"end_line":357,"context_start_line":1,"context_end_line":357,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport subprocess\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\n\nimport torch\nimport torch.distributed as dist\nfrom torch import inf\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device=\"cuda\")\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value\n        )\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\"{}: {}\".format(name, str(meter)))\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = \"\"\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        data_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        space_fmt = \":\" + str(len(str(len(iterable)))) + \"d\"\n        log_msg = [header, \"[{0\" + space_fmt + \"}/{1}]\", \"eta: {eta}\", \"{meters}\", \"time: {time}\", \"data: {data}\"]\n        if torch.cuda.is_available():\n            log_msg.append(\"max mem: {memory:.0f}\")\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(\n                        log_msg.format(\n                            i,\n                            len(iterable),\n                            eta=eta_string,\n                            meters=str(self),\n                            time=str(iter_time),\n                            data=str(data_time),\n                            memory=torch.cuda.max_memory_allocated() / MB,\n                        )\n                    )\n                else:\n                    print(\n                        log_msg.format(\n                            i, len(iterable), eta=eta_string, meters=str(self), time=str(iter_time), data=str(data_time)\n                        )\n                    )\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print(\"{} Total time: {} ({:.4f} s / it)\".format(header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop(\"force\", False)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print(\"[{}] \".format(now), end=\"\")  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode():\n    if \"RANK\" in os.environ and \"WORLD_SIZE\" in os.environ:\n        # torchrun\n        rank = int(os.environ[\"RANK\"])\n        world_size = int(os.environ[\"WORLD_SIZE\"])\n        gpu = int(os.environ[\"LOCAL_RANK\"])\n    elif \"SLURM_PROCID\" in os.environ:\n        # slurm\n        rank = int(os.environ[\"SLURM_PROCID\"])\n        world_size = int(os.environ[\"SLURM_NPROCS\"])\n        if \"MASTER_PORT\" not in os.environ:\n            os.environ[\"MASTER_PORT\"] = \"8964\"\n        os.environ[\"MASTER_ADDR\"] = (\n            subprocess.check_output('scontrol show hostnames \"$SLURM_JOB_NODELIST\"', shell=True)\n            .decode(\"utf-8\")\n            .splitlines()[0]\n        )\n        gpu = rank % torch.cuda.device_count()\n    else:\n        print(\n            \"Only distributed mode is supported but distributed env cannot be initialized. \"\n            \"Use torchrun --nproc_per_node 1 ... to run on a single gpu.\"\n        )\n\n    torch.cuda.set_device(gpu)\n    print(\"| distributed init (rank {}), gpu {}\".format(rank, gpu), flush=True)\n    torch.distributed.init_process_group(\"nccl\", world_size=world_size, rank=rank)\n    torch.distributed.barrier()\n    setup_for_distributed(rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.0)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(\n            torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type\n        )\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / (\"checkpoint-%s.pth\" % epoch_name)]\n        model_state_dict = model_without_ddp.state_dict()\n        for n, p in model_without_ddp.named_parameters():\n            if not p.requires_grad:\n                del model_state_dict[n]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                \"model\": model_state_dict,\n                \"optimizer\": optimizer.state_dict(),\n                \"epoch\": epoch,\n                \"scaler\": loss_scaler.state_dict(),\n                \"args\": args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {\"epoch\": epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.auto_resume and not args.resume:\n        print(\"Trying to auto resume...\")\n        output_dir = Path(args.output_dir)\n        ckpts = output_dir.glob(\"*.pth\")\n        ckpt_iters = [int(x.stem[len(\"checkpoint-\") :]) for x in ckpts]\n        ckpt_iters.sort()\n        print(\"Candidate checkpoints:\", [str(Path(args.output_dir) / (\"checkpoint-%d.pth\" % x)) for x in ckpt_iters])\n        for x in ckpt_iters[::-1]:\n            ckpt_path = str(Path(args.output_dir) / (\"checkpoint-%d.pth\" % x))\n            try:\n                torch.load(ckpt_path, map_location=\"cpu\")\n            except Exception as e:\n                print(\"Failed to load %s with error:\" % ckpt_path, e)\n                continue\n            print(\"Found a valid checkpoint:\", ckpt_path)\n            args.resume = ckpt_path\n            break\n\n    if args.resume:\n        if args.resume.startswith(\"https\"):\n            checkpoint = torch.hub.load_state_dict_from_url(args.resume, map_location=\"cpu\", check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location=\"cpu\")\n        missing_keys, unexpected_keys = model_without_ddp.load_state_dict(checkpoint[\"model\"], strict=False)\n        assert len(unexpected_keys) == 0\n        missing_keys = set(missing_keys)\n        for n, p in model_without_ddp.named_parameters():\n            if p.requires_grad:\n                assert n not in missing_keys\n        print(\"Resume checkpoint %s\" % args.resume)\n        if \"optimizer\" in checkpoint and \"epoch\" in checkpoint and not (hasattr(args, \"eval\") and args.eval):\n            optimizer.load_state_dict(checkpoint[\"optimizer\"])\n            args.start_epoch = checkpoint[\"epoch\"] + 1\n            if \"scaler\" in checkpoint:\n                loss_scaler.load_state_dict(checkpoint[\"scaler\"])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.SmoothedValue","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.util.misc.SmoothedValue#L25-L81","kind":"class","name":"SmoothedValue","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":25,"end_line":81,"context_start_line":5,"context_end_line":101,"code":"# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport subprocess\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\n\nimport torch\nimport torch.distributed as dist\nfrom torch import inf\n\n\nclass SmoothedValue(object):\n    \"\"\"Track a series of values and provide access to smoothed values over a\n    window or the global series average.\n    \"\"\"\n\n    def __init__(self, window_size=20, fmt=None):\n        if fmt is None:\n            fmt = \"{median:.4f} ({global_avg:.4f})\"\n        self.deque = deque(maxlen=window_size)\n        self.total = 0.0\n        self.count = 0\n        self.fmt = fmt\n\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device=\"cuda\")\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value\n        )\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.MetricLogger","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.util.misc.MetricLogger#L84-L161","kind":"class","name":"MetricLogger","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":84,"end_line":161,"context_start_line":64,"context_end_line":181,"code":"        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value\n        )\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\"{}: {}\".format(name, str(meter)))\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = \"\"\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        data_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        space_fmt = \":\" + str(len(str(len(iterable)))) + \"d\"\n        log_msg = [header, \"[{0\" + space_fmt + \"}/{1}]\", \"eta: {eta}\", \"{meters}\", \"time: {time}\", \"data: {data}\"]\n        if torch.cuda.is_available():\n            log_msg.append(\"max mem: {memory:.0f}\")\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(\n                        log_msg.format(\n                            i,\n                            len(iterable),\n                            eta=eta_string,\n                            meters=str(self),\n                            time=str(iter_time),\n                            data=str(data_time),\n                            memory=torch.cuda.max_memory_allocated() / MB,\n                        )\n                    )\n                else:\n                    print(\n                        log_msg.format(\n                            i, len(iterable), eta=eta_string, meters=str(self), time=str(iter_time), data=str(data_time)\n                        )\n                    )\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print(\"{} Total time: {} ({:.4f} s / it)\".format(header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop(\"force\", False)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print(\"[{}] \".format(now), end=\"\")  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.setup_for_distributed","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.setup_for_distributed#L164-L177","kind":"function","name":"setup_for_distributed","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":164,"end_line":177,"context_start_line":144,"context_end_line":197,"code":"                            eta=eta_string,\n                            meters=str(self),\n                            time=str(iter_time),\n                            data=str(data_time),\n                            memory=torch.cuda.max_memory_allocated() / MB,\n                        )\n                    )\n                else:\n                    print(\n                        log_msg.format(\n                            i, len(iterable), eta=eta_string, meters=str(self), time=str(iter_time), data=str(data_time)\n                        )\n                    )\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print(\"{} Total time: {} ({:.4f} s / it)\".format(header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop(\"force\", False)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print(\"[{}] \".format(now), end=\"\")  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.is_dist_avail_and_initialized","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.is_dist_avail_and_initialized#L180-L185","kind":"function","name":"is_dist_avail_and_initialized","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":180,"end_line":185,"context_start_line":160,"context_end_line":205,"code":"        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print(\"{} Total time: {} ({:.4f} s / it)\".format(header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop(\"force\", False)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print(\"[{}] \".format(now), end=\"\")  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.get_world_size","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.get_world_size#L188-L191","kind":"function","name":"get_world_size","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":188,"end_line":191,"context_start_line":168,"context_end_line":211,"code":"    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop(\"force\", False)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print(\"[{}] \".format(now), end=\"\")  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode():\n    if \"RANK\" in os.environ and \"WORLD_SIZE\" in os.environ:\n        # torchrun","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.get_rank","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.get_rank#L194-L197","kind":"function","name":"get_rank","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":194,"end_line":197,"context_start_line":174,"context_end_line":217,"code":"            builtin_print(\"[{}] \".format(now), end=\"\")  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode():\n    if \"RANK\" in os.environ and \"WORLD_SIZE\" in os.environ:\n        # torchrun\n        rank = int(os.environ[\"RANK\"])\n        world_size = int(os.environ[\"WORLD_SIZE\"])\n        gpu = int(os.environ[\"LOCAL_RANK\"])\n    elif \"SLURM_PROCID\" in os.environ:\n        # slurm\n        rank = int(os.environ[\"SLURM_PROCID\"])","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.is_main_process","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.is_main_process#L200-L201","kind":"function","name":"is_main_process","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":200,"end_line":201,"context_start_line":180,"context_end_line":221,"code":"def is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode():\n    if \"RANK\" in os.environ and \"WORLD_SIZE\" in os.environ:\n        # torchrun\n        rank = int(os.environ[\"RANK\"])\n        world_size = int(os.environ[\"WORLD_SIZE\"])\n        gpu = int(os.environ[\"LOCAL_RANK\"])\n    elif \"SLURM_PROCID\" in os.environ:\n        # slurm\n        rank = int(os.environ[\"SLURM_PROCID\"])\n        world_size = int(os.environ[\"SLURM_NPROCS\"])\n        if \"MASTER_PORT\" not in os.environ:\n            os.environ[\"MASTER_PORT\"] = \"8964\"\n        os.environ[\"MASTER_ADDR\"] = (","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.save_on_master","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.save_on_master#L204-L206","kind":"function","name":"save_on_master","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":204,"end_line":206,"context_start_line":184,"context_end_line":226,"code":"        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode():\n    if \"RANK\" in os.environ and \"WORLD_SIZE\" in os.environ:\n        # torchrun\n        rank = int(os.environ[\"RANK\"])\n        world_size = int(os.environ[\"WORLD_SIZE\"])\n        gpu = int(os.environ[\"LOCAL_RANK\"])\n    elif \"SLURM_PROCID\" in os.environ:\n        # slurm\n        rank = int(os.environ[\"SLURM_PROCID\"])\n        world_size = int(os.environ[\"SLURM_NPROCS\"])\n        if \"MASTER_PORT\" not in os.environ:\n            os.environ[\"MASTER_PORT\"] = \"8964\"\n        os.environ[\"MASTER_ADDR\"] = (\n            subprocess.check_output('scontrol show hostnames \"$SLURM_JOB_NODELIST\"', shell=True)\n            .decode(\"utf-8\")\n            .splitlines()[0]\n        )\n        gpu = rank % torch.cuda.device_count()","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.init_distributed_mode","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.init_distributed_mode#L209-L237","kind":"function","name":"init_distributed_mode","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":209,"end_line":237,"context_start_line":189,"context_end_line":257,"code":"    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():\n        return 0\n    return dist.get_rank()\n\n\ndef is_main_process():\n    return get_rank() == 0\n\n\ndef save_on_master(*args, **kwargs):\n    if is_main_process():\n        torch.save(*args, **kwargs)\n\n\ndef init_distributed_mode():\n    if \"RANK\" in os.environ and \"WORLD_SIZE\" in os.environ:\n        # torchrun\n        rank = int(os.environ[\"RANK\"])\n        world_size = int(os.environ[\"WORLD_SIZE\"])\n        gpu = int(os.environ[\"LOCAL_RANK\"])\n    elif \"SLURM_PROCID\" in os.environ:\n        # slurm\n        rank = int(os.environ[\"SLURM_PROCID\"])\n        world_size = int(os.environ[\"SLURM_NPROCS\"])\n        if \"MASTER_PORT\" not in os.environ:\n            os.environ[\"MASTER_PORT\"] = \"8964\"\n        os.environ[\"MASTER_ADDR\"] = (\n            subprocess.check_output('scontrol show hostnames \"$SLURM_JOB_NODELIST\"', shell=True)\n            .decode(\"utf-8\")\n            .splitlines()[0]\n        )\n        gpu = rank % torch.cuda.device_count()\n    else:\n        print(\n            \"Only distributed mode is supported but distributed env cannot be initialized. \"\n            \"Use torchrun --nproc_per_node 1 ... to run on a single gpu.\"\n        )\n\n    torch.cuda.set_device(gpu)\n    print(\"| distributed init (rank {}), gpu {}\".format(rank, gpu), flush=True)\n    torch.distributed.init_process_group(\"nccl\", world_size=world_size, rank=rank)\n    torch.distributed.barrier()\n    setup_for_distributed(rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.NativeScalerWithGradNormCount","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.util.misc.NativeScalerWithGradNormCount#L240-L266","kind":"class","name":"NativeScalerWithGradNormCount","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":240,"end_line":266,"context_start_line":220,"context_end_line":286,"code":"            os.environ[\"MASTER_PORT\"] = \"8964\"\n        os.environ[\"MASTER_ADDR\"] = (\n            subprocess.check_output('scontrol show hostnames \"$SLURM_JOB_NODELIST\"', shell=True)\n            .decode(\"utf-8\")\n            .splitlines()[0]\n        )\n        gpu = rank % torch.cuda.device_count()\n    else:\n        print(\n            \"Only distributed mode is supported but distributed env cannot be initialized. \"\n            \"Use torchrun --nproc_per_node 1 ... to run on a single gpu.\"\n        )\n\n    torch.cuda.set_device(gpu)\n    print(\"| distributed init (rank {}), gpu {}\".format(rank, gpu), flush=True)\n    torch.distributed.init_process_group(\"nccl\", world_size=world_size, rank=rank)\n    torch.distributed.barrier()\n    setup_for_distributed(rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.0)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(\n            torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type\n        )\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.get_grad_norm_","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.get_grad_norm_#L269-L283","kind":"function","name":"get_grad_norm_","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":269,"end_line":283,"context_start_line":249,"context_end_line":303,"code":"            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.0)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(\n            torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type\n        )\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / (\"checkpoint-%s.pth\" % epoch_name)]\n        model_state_dict = model_without_ddp.state_dict()\n        for n, p in model_without_ddp.named_parameters():\n            if not p.requires_grad:\n                del model_state_dict[n]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                \"model\": model_state_dict,\n                \"optimizer\": optimizer.state_dict(),\n                \"epoch\": epoch,\n                \"scaler\": loss_scaler.state_dict(),\n                \"args\": args,\n            }\n","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.save_model","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.save_model#L286-L307","kind":"function","name":"save_model","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":286,"end_line":307,"context_start_line":266,"context_end_line":327,"code":"        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.0)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(\n            torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type\n        )\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):\n    output_dir = Path(args.output_dir)\n    epoch_name = str(epoch)\n    if loss_scaler is not None:\n        checkpoint_paths = [output_dir / (\"checkpoint-%s.pth\" % epoch_name)]\n        model_state_dict = model_without_ddp.state_dict()\n        for n, p in model_without_ddp.named_parameters():\n            if not p.requires_grad:\n                del model_state_dict[n]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                \"model\": model_state_dict,\n                \"optimizer\": optimizer.state_dict(),\n                \"epoch\": epoch,\n                \"scaler\": loss_scaler.state_dict(),\n                \"args\": args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {\"epoch\": epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.auto_resume and not args.resume:\n        print(\"Trying to auto resume...\")\n        output_dir = Path(args.output_dir)\n        ckpts = output_dir.glob(\"*.pth\")\n        ckpt_iters = [int(x.stem[len(\"checkpoint-\") :]) for x in ckpts]\n        ckpt_iters.sort()\n        print(\"Candidate checkpoints:\", [str(Path(args.output_dir) / (\"checkpoint-%d.pth\" % x)) for x in ckpt_iters])\n        for x in ckpt_iters[::-1]:\n            ckpt_path = str(Path(args.output_dir) / (\"checkpoint-%d.pth\" % x))\n            try:\n                torch.load(ckpt_path, map_location=\"cpu\")\n            except Exception as e:\n                print(\"Failed to load %s with error:\" % ckpt_path, e)\n                continue\n            print(\"Found a valid checkpoint:\", ckpt_path)\n            args.resume = ckpt_path\n            break","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.load_model","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.load_model#L310-L346","kind":"function","name":"load_model","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":310,"end_line":346,"context_start_line":290,"context_end_line":357,"code":"        checkpoint_paths = [output_dir / (\"checkpoint-%s.pth\" % epoch_name)]\n        model_state_dict = model_without_ddp.state_dict()\n        for n, p in model_without_ddp.named_parameters():\n            if not p.requires_grad:\n                del model_state_dict[n]\n        for checkpoint_path in checkpoint_paths:\n            to_save = {\n                \"model\": model_state_dict,\n                \"optimizer\": optimizer.state_dict(),\n                \"epoch\": epoch,\n                \"scaler\": loss_scaler.state_dict(),\n                \"args\": args,\n            }\n\n            save_on_master(to_save, checkpoint_path)\n    else:\n        client_state = {\"epoch\": epoch}\n        model.save_checkpoint(save_dir=args.output_dir, tag=\"checkpoint-%s\" % epoch_name, client_state=client_state)\n\n\ndef load_model(args, model_without_ddp, optimizer, loss_scaler):\n    if args.auto_resume and not args.resume:\n        print(\"Trying to auto resume...\")\n        output_dir = Path(args.output_dir)\n        ckpts = output_dir.glob(\"*.pth\")\n        ckpt_iters = [int(x.stem[len(\"checkpoint-\") :]) for x in ckpts]\n        ckpt_iters.sort()\n        print(\"Candidate checkpoints:\", [str(Path(args.output_dir) / (\"checkpoint-%d.pth\" % x)) for x in ckpt_iters])\n        for x in ckpt_iters[::-1]:\n            ckpt_path = str(Path(args.output_dir) / (\"checkpoint-%d.pth\" % x))\n            try:\n                torch.load(ckpt_path, map_location=\"cpu\")\n            except Exception as e:\n                print(\"Failed to load %s with error:\" % ckpt_path, e)\n                continue\n            print(\"Found a valid checkpoint:\", ckpt_path)\n            args.resume = ckpt_path\n            break\n\n    if args.resume:\n        if args.resume.startswith(\"https\"):\n            checkpoint = torch.hub.load_state_dict_from_url(args.resume, map_location=\"cpu\", check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location=\"cpu\")\n        missing_keys, unexpected_keys = model_without_ddp.load_state_dict(checkpoint[\"model\"], strict=False)\n        assert len(unexpected_keys) == 0\n        missing_keys = set(missing_keys)\n        for n, p in model_without_ddp.named_parameters():\n            if p.requires_grad:\n                assert n not in missing_keys\n        print(\"Resume checkpoint %s\" % args.resume)\n        if \"optimizer\" in checkpoint and \"epoch\" in checkpoint and not (hasattr(args, \"eval\") and args.eval):\n            optimizer.load_state_dict(checkpoint[\"optimizer\"])\n            args.start_epoch = checkpoint[\"epoch\"] + 1\n            if \"scaler\" in checkpoint:\n                loss_scaler.load_state_dict(checkpoint[\"scaler\"])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.all_reduce_mean","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.all_reduce_mean#L349-L357","kind":"function","name":"all_reduce_mean","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":349,"end_line":357,"context_start_line":329,"context_end_line":357,"code":"    if args.resume:\n        if args.resume.startswith(\"https\"):\n            checkpoint = torch.hub.load_state_dict_from_url(args.resume, map_location=\"cpu\", check_hash=True)\n        else:\n            checkpoint = torch.load(args.resume, map_location=\"cpu\")\n        missing_keys, unexpected_keys = model_without_ddp.load_state_dict(checkpoint[\"model\"], strict=False)\n        assert len(unexpected_keys) == 0\n        missing_keys = set(missing_keys)\n        for n, p in model_without_ddp.named_parameters():\n            if p.requires_grad:\n                assert n not in missing_keys\n        print(\"Resume checkpoint %s\" % args.resume)\n        if \"optimizer\" in checkpoint and \"epoch\" in checkpoint and not (hasattr(args, \"eval\") and args.eval):\n            optimizer.load_state_dict(checkpoint[\"optimizer\"])\n            args.start_epoch = checkpoint[\"epoch\"] + 1\n            if \"scaler\" in checkpoint:\n                loss_scaler.load_state_dict(checkpoint[\"scaler\"])\n            print(\"With optim & sched!\")\n\n\ndef all_reduce_mean(x):\n    world_size = get_world_size()\n    if world_size > 1:\n        x_reduce = torch.tensor(x).cuda()\n        dist.all_reduce(x_reduce)\n        x_reduce /= world_size\n        return x_reduce.item()\n    else:\n        return x","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.__init__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.__init__#L243-L244","kind":"function","name":"__init__","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":243,"end_line":244,"context_start_line":223,"context_end_line":264,"code":"            .decode(\"utf-8\")\n            .splitlines()[0]\n        )\n        gpu = rank % torch.cuda.device_count()\n    else:\n        print(\n            \"Only distributed mode is supported but distributed env cannot be initialized. \"\n            \"Use torchrun --nproc_per_node 1 ... to run on a single gpu.\"\n        )\n\n    torch.cuda.set_device(gpu)\n    print(\"| distributed init (rank {}), gpu {}\".format(rank, gpu), flush=True)\n    torch.distributed.init_process_group(\"nccl\", world_size=world_size, rank=rank)\n    torch.distributed.barrier()\n    setup_for_distributed(rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.update","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.update#L89-L96","kind":"function","name":"update","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":89,"end_line":96,"context_start_line":69,"context_end_line":116,"code":"\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value\n        )\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\"{}: {}\".format(name, str(meter)))\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.synchronize_between_processes","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.synchronize_between_processes#L111-L113","kind":"function","name":"synchronize_between_processes","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":111,"end_line":113,"context_start_line":91,"context_end_line":133,"code":"            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\"{}: {}\".format(name, str(meter)))\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = \"\"\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        data_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        space_fmt = \":\" + str(len(str(len(iterable)))) + \"d\"\n        log_msg = [header, \"[{0\" + space_fmt + \"}/{1}]\", \"eta: {eta}\", \"{meters}\", \"time: {time}\", \"data: {data}\"]\n        if torch.cuda.is_available():\n            log_msg.append(\"max mem: {memory:.0f}\")\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.median","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.median#L57-L59","kind":"function","name":"median","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":57,"end_line":59,"context_start_line":37,"context_end_line":79,"code":"\n    def update(self, value, n=1):\n        self.deque.append(value)\n        self.count += n\n        self.total += value * n\n\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device=\"cuda\")\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.avg","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.avg#L62-L64","kind":"function","name":"avg","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":62,"end_line":64,"context_start_line":42,"context_end_line":84,"code":"\n    def synchronize_between_processes(self):\n        \"\"\"\n        Warning: does not synchronize the deque!\n        \"\"\"\n        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device=\"cuda\")\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value\n        )\n\n\nclass MetricLogger(object):","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.global_avg","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.global_avg#L67-L68","kind":"function","name":"global_avg","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":67,"end_line":68,"context_start_line":47,"context_end_line":88,"code":"        if not is_dist_avail_and_initialized():\n            return\n        t = torch.tensor([self.count, self.total], dtype=torch.float64, device=\"cuda\")\n        dist.barrier()\n        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value\n        )\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.max","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.max#L71-L72","kind":"function","name":"max","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":71,"end_line":72,"context_start_line":51,"context_end_line":92,"code":"        dist.all_reduce(t)\n        t = t.tolist()\n        self.count = int(t[0])\n        self.total = t[1]\n\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value\n        )\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.value","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.value#L75-L76","kind":"function","name":"value","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":75,"end_line":76,"context_start_line":55,"context_end_line":96,"code":"\n    @property\n    def median(self):\n        d = torch.tensor(list(self.deque))\n        return d.median().item()\n\n    @property\n    def avg(self):\n        d = torch.tensor(list(self.deque), dtype=torch.float32)\n        return d.mean().item()\n\n    @property\n    def global_avg(self):\n        return self.total / self.count\n\n    @property\n    def max(self):\n        return max(self.deque)\n\n    @property\n    def value(self):\n        return self.deque[-1]\n\n    def __str__(self):\n        return self.fmt.format(\n            median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value\n        )\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.__str__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.__str__#L105-L109","kind":"function","name":"__str__","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":105,"end_line":109,"context_start_line":85,"context_end_line":129,"code":"    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\"{}: {}\".format(name, str(meter)))\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = \"\"\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        data_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        space_fmt = \":\" + str(len(str(len(iterable)))) + \"d\"\n        log_msg = [header, \"[{0\" + space_fmt + \"}/{1}]\", \"eta: {eta}\", \"{meters}\", \"time: {time}\", \"data: {data}\"]\n        if torch.cuda.is_available():\n            log_msg.append(\"max mem: {memory:.0f}\")","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.__getattr__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.__getattr__#L98-L103","kind":"function","name":"__getattr__","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":98,"end_line":103,"context_start_line":78,"context_end_line":123,"code":"    def __str__(self):\n        return self.fmt.format(\n            median=self.median, avg=self.avg, global_avg=self.global_avg, max=self.max, value=self.value\n        )\n\n\nclass MetricLogger(object):\n    def __init__(self, delimiter=\"\\t\"):\n        self.meters = defaultdict(SmoothedValue)\n        self.delimiter = delimiter\n\n    def update(self, **kwargs):\n        for k, v in kwargs.items():\n            if v is None:\n                continue\n            if isinstance(v, torch.Tensor):\n                v = v.item()\n            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\"{}: {}\".format(name, str(meter)))\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = \"\"\n        start_time = time.time()\n        end = time.time()","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.add_meter","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.add_meter#L115-L116","kind":"function","name":"add_meter","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":115,"end_line":116,"context_start_line":95,"context_end_line":136,"code":"            assert isinstance(v, (float, int))\n            self.meters[k].update(v)\n\n    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\"{}: {}\".format(name, str(meter)))\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = \"\"\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        data_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        space_fmt = \":\" + str(len(str(len(iterable)))) + \"d\"\n        log_msg = [header, \"[{0\" + space_fmt + \"}/{1}]\", \"eta: {eta}\", \"{meters}\", \"time: {time}\", \"data: {data}\"]\n        if torch.cuda.is_available():\n            log_msg.append(\"max mem: {memory:.0f}\")\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.log_every","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.log_every#L118-L161","kind":"function","name":"log_every","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":118,"end_line":161,"context_start_line":98,"context_end_line":181,"code":"    def __getattr__(self, attr):\n        if attr in self.meters:\n            return self.meters[attr]\n        if attr in self.__dict__:\n            return self.__dict__[attr]\n        raise AttributeError(\"'{}' object has no attribute '{}'\".format(type(self).__name__, attr))\n\n    def __str__(self):\n        loss_str = []\n        for name, meter in self.meters.items():\n            loss_str.append(\"{}: {}\".format(name, str(meter)))\n        return self.delimiter.join(loss_str)\n\n    def synchronize_between_processes(self):\n        for meter in self.meters.values():\n            meter.synchronize_between_processes()\n\n    def add_meter(self, name, meter):\n        self.meters[name] = meter\n\n    def log_every(self, iterable, print_freq, header=None):\n        i = 0\n        if not header:\n            header = \"\"\n        start_time = time.time()\n        end = time.time()\n        iter_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        data_time = SmoothedValue(fmt=\"{avg:.4f}\")\n        space_fmt = \":\" + str(len(str(len(iterable)))) + \"d\"\n        log_msg = [header, \"[{0\" + space_fmt + \"}/{1}]\", \"eta: {eta}\", \"{meters}\", \"time: {time}\", \"data: {data}\"]\n        if torch.cuda.is_available():\n            log_msg.append(\"max mem: {memory:.0f}\")\n        log_msg = self.delimiter.join(log_msg)\n        MB = 1024.0 * 1024.0\n        for obj in iterable:\n            data_time.update(time.time() - end)\n            yield obj\n            iter_time.update(time.time() - end)\n            if i % print_freq == 0 or i == len(iterable) - 1:\n                eta_seconds = iter_time.global_avg * (len(iterable) - i)\n                eta_string = str(datetime.timedelta(seconds=int(eta_seconds)))\n                if torch.cuda.is_available():\n                    print(\n                        log_msg.format(\n                            i,\n                            len(iterable),\n                            eta=eta_string,\n                            meters=str(self),\n                            time=str(iter_time),\n                            data=str(data_time),\n                            memory=torch.cuda.max_memory_allocated() / MB,\n                        )\n                    )\n                else:\n                    print(\n                        log_msg.format(\n                            i, len(iterable), eta=eta_string, meters=str(self), time=str(iter_time), data=str(data_time)\n                        )\n                    )\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print(\"{} Total time: {} ({:.4f} s / it)\".format(header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop(\"force\", False)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print(\"[{}] \".format(now), end=\"\")  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.print","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.print#L170-L175","kind":"function","name":"print","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":170,"end_line":175,"context_start_line":150,"context_end_line":195,"code":"                    )\n                else:\n                    print(\n                        log_msg.format(\n                            i, len(iterable), eta=eta_string, meters=str(self), time=str(iter_time), data=str(data_time)\n                        )\n                    )\n            i += 1\n            end = time.time()\n        total_time = time.time() - start_time\n        total_time_str = str(datetime.timedelta(seconds=int(total_time)))\n        print(\"{} Total time: {} ({:.4f} s / it)\".format(header, total_time_str, total_time / len(iterable)))\n\n\ndef setup_for_distributed(is_master):\n    \"\"\"\n    This function disables printing when not in master process\n    \"\"\"\n    builtin_print = builtins.print\n\n    def print(*args, **kwargs):\n        force = kwargs.pop(\"force\", False)\n        if is_master or force:\n            now = datetime.datetime.now().time()\n            builtin_print(\"[{}] \".format(now), end=\"\")  # print with time stamp\n            builtin_print(*args, **kwargs)\n\n    builtins.print = print\n\n\ndef is_dist_avail_and_initialized():\n    if not dist.is_available():\n        return False\n    if not dist.is_initialized():\n        return False\n    return True\n\n\ndef get_world_size():\n    if not is_dist_avail_and_initialized():\n        return 1\n    return dist.get_world_size()\n\n\ndef get_rank():\n    if not is_dist_avail_and_initialized():","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.__call__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.__call__#L246-L260","kind":"function","name":"__call__","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":246,"end_line":260,"context_start_line":226,"context_end_line":280,"code":"        gpu = rank % torch.cuda.device_count()\n    else:\n        print(\n            \"Only distributed mode is supported but distributed env cannot be initialized. \"\n            \"Use torchrun --nproc_per_node 1 ... to run on a single gpu.\"\n        )\n\n    torch.cuda.set_device(gpu)\n    print(\"| distributed init (rank {}), gpu {}\".format(rank, gpu), flush=True)\n    torch.distributed.init_process_group(\"nccl\", world_size=world_size, rank=rank)\n    torch.distributed.barrier()\n    setup_for_distributed(rank == 0)\n\n\nclass NativeScalerWithGradNormCount:\n    state_dict_key = \"amp_scaler\"\n\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.0)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.state_dict","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.state_dict#L262-L263","kind":"function","name":"state_dict","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":262,"end_line":263,"context_start_line":242,"context_end_line":283,"code":"\n    def __init__(self):\n        self._scaler = torch.cuda.amp.GradScaler()\n\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.0)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(\n            torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type\n        )\n    return total_norm","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.util.misc.load_state_dict","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.util.misc.load_state_dict#L265-L266","kind":"function","name":"load_state_dict","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":265,"end_line":266,"context_start_line":245,"context_end_line":286,"code":"\n    def __call__(self, loss, optimizer, clip_grad=None, parameters=None, create_graph=False, update_grad=True):\n        self._scaler.scale(loss).backward(create_graph=create_graph)\n        if update_grad:\n            if clip_grad is not None:\n                assert parameters is not None\n                self._scaler.unscale_(optimizer)  # unscale the gradients of optimizer's assigned params in-place\n                norm = torch.nn.utils.clip_grad_norm_(parameters, clip_grad)\n            else:\n                self._scaler.unscale_(optimizer)\n                norm = get_grad_norm_(parameters)\n            self._scaler.step(optimizer)\n            self._scaler.update()\n        else:\n            norm = None\n        return norm\n\n    def state_dict(self):\n        return self._scaler.state_dict()\n\n    def load_state_dict(self, state_dict):\n        self._scaler.load_state_dict(state_dict)\n\n\ndef get_grad_norm_(parameters, norm_type: float = 2.0) -> torch.Tensor:\n    if isinstance(parameters, torch.Tensor):\n        parameters = [parameters]\n    parameters = [p for p in parameters if p.grad is not None]\n    norm_type = float(norm_type)\n    if len(parameters) == 0:\n        return torch.tensor(0.0)\n    device = parameters[0].grad.device\n    if norm_type == inf:\n        total_norm = max(p.grad.detach().abs().max().to(device) for p in parameters)\n    else:\n        total_norm = torch.norm(\n            torch.stack([torch.norm(p.grad.detach(), norm_type).to(device) for p in parameters]), norm_type\n        )\n    return total_norm\n\n\ndef save_model(args, epoch, model, model_without_ddp, optimizer, loss_scaler):","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.generation","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_chat65b.llama.generation#L1-L78","kind":"module","name":"llama_adapter_v2_chat65b.llama.generation","path":"llama_adapter_v2_chat65b/llama/generation.py","language":"python","start_line":1,"end_line":78,"context_start_line":1,"context_end_line":78,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer) -> None:\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        self.model.enable_cache()\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n        prompt_tokens = [x[-(2048 - max_gen_len) :] for x in prompt_tokens]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward_inference(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        self.model.disable_cache()\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[len(prompt_tokens[i]) : len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"8cbad0b94018145da71f990d707bdf3f9a90a5c9dead3b7406b19e940356a09a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.generation.LLaMA","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.llama.generation.LLaMA#L12-L67","kind":"class","name":"LLaMA","path":"llama_adapter_v2_chat65b/llama/generation.py","language":"python","start_line":12,"end_line":67,"context_start_line":1,"context_end_line":78,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer) -> None:\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        self.model.enable_cache()\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n        prompt_tokens = [x[-(2048 - max_gen_len) :] for x in prompt_tokens]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward_inference(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        self.model.disable_cache()\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[len(prompt_tokens[i]) : len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"8cbad0b94018145da71f990d707bdf3f9a90a5c9dead3b7406b19e940356a09a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.generation.sample_top_p","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.generation.sample_top_p#L70-L78","kind":"function","name":"sample_top_p","path":"llama_adapter_v2_chat65b/llama/generation.py","language":"python","start_line":70,"end_line":78,"context_start_line":50,"context_end_line":78,"code":"            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        self.model.disable_cache()\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[len(prompt_tokens[i]) : len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"8cbad0b94018145da71f990d707bdf3f9a90a5c9dead3b7406b19e940356a09a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.generation.__init__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.generation.__init__#L13-L15","kind":"function","name":"__init__","path":"llama_adapter_v2_chat65b/llama/generation.py","language":"python","start_line":13,"end_line":15,"context_start_line":1,"context_end_line":35,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer) -> None:\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        self.model.enable_cache()\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n        prompt_tokens = [x[-(2048 - max_gen_len) :] for x in prompt_tokens]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n","source_hash":"8cbad0b94018145da71f990d707bdf3f9a90a5c9dead3b7406b19e940356a09a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.generation.generate","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.generation.generate#L17-L67","kind":"function","name":"generate","path":"llama_adapter_v2_chat65b/llama/generation.py","language":"python","start_line":17,"end_line":67,"context_start_line":1,"context_end_line":78,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer) -> None:\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,\n        top_p: float = 0.95,\n    ) -> List[str]:\n        bsz = len(prompts)\n        params = self.model.params\n        self.model.enable_cache()\n\n        prompt_tokens = [self.tokenizer.encode(x, bos=True, eos=False) for x in prompts]\n        prompt_tokens = [x[-(2048 - max_gen_len) :] for x in prompt_tokens]\n\n        min_prompt_size = min([len(t) for t in prompt_tokens])\n        max_prompt_size = max([len(t) for t in prompt_tokens])\n\n        total_len = min(params.max_seq_len, max_gen_len + max_prompt_size)\n\n        tokens = torch.full((bsz, total_len), self.tokenizer.pad_id).cuda().long()\n        for k, t in enumerate(prompt_tokens):\n            tokens[k, : len(t)] = torch.tensor(t).long()\n        input_text_mask = tokens != self.tokenizer.pad_id\n        start_pos = min_prompt_size\n        prev_pos = 0\n        for cur_pos in range(start_pos, total_len):\n            logits = self.model.forward_inference(tokens[:, prev_pos:cur_pos], prev_pos)\n            if temperature > 0:\n                probs = torch.softmax(logits / temperature, dim=-1)\n                next_token = sample_top_p(probs, top_p)\n            else:\n                next_token = torch.argmax(logits, dim=-1)\n            next_token = next_token.reshape(-1)\n            # only replace token if prompt has already been generated\n            next_token = torch.where(input_text_mask[:, cur_pos], tokens[:, cur_pos], next_token)\n            tokens[:, cur_pos] = next_token\n            prev_pos = cur_pos\n\n        self.model.disable_cache()\n\n        decoded = []\n        for i, t in enumerate(tokens.tolist()):\n            # cut to max gen len\n            t = t[len(prompt_tokens[i]) : len(prompt_tokens[i]) + max_gen_len]\n            # cut to eos tok if any\n            try:\n                t = t[: t.index(self.tokenizer.eos_id)]\n            except ValueError:\n                pass\n            decoded.append(self.tokenizer.decode(t))\n        return decoded\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token","source_hash":"8cbad0b94018145da71f990d707bdf3f9a90a5c9dead3b7406b19e940356a09a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_chat65b.llama.model#L1-L365","kind":"module","name":"llama_adapter_v2_chat65b.llama.model","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":1,"end_line":365,"context_start_line":1,"context_end_line":365,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.nn.functional as F\nfrom fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_seq_len: int = 2048\n\n    adapter_len: int = 0\n    adapter_layer: int = 0\n\n    add_bias: bool = False\n    add_scale: bool = False\n    train_norm: bool = False\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return (output * self.weight).type_as(x)\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\ndef forward_linear_with_scale_and_bias(x, module, scale=None, bias=None):\n    if scale is not None:\n        x = x * scale\n    x = module(x)\n    if bias is not None:\n        x = x + bias\n    return x\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n        self.dim = args.dim\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wv = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wo = RowParallelLinear(\n            args.n_heads * self.head_dim, args.dim, bias=False, input_is_parallel=True, init_method=lambda x: x\n        )\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, args.n_heads, 1, 1))\n        self.head_start = self.n_local_heads * fs_init.get_model_parallel_rank()\n        self.head_end = self.n_local_heads * (fs_init.get_model_parallel_rank() + 1)\n\n        self.cache_enabled = False\n        self.cache_k, self.cache_v = None, None\n\n        if args.add_bias:\n            self.wq_bias, self.wk_bias, self.wv_bias = [\n                nn.Parameter(torch.zeros([self.n_local_heads * self.head_dim])) for _ in range(3)\n            ]\n            self.wo_bias = nn.Parameter(torch.zeros([args.dim]))\n        else:\n            self.wq_bias = self.wk_bias = self.wv_bias = self.wo_bias = None\n\n        if args.add_scale:\n            self.wq_scale, self.wk_scale, self.wv_scale = [nn.Parameter(torch.ones([args.dim])) for _ in range(3)]\n            self.wo_scale = nn.Parameter(torch.ones([self.n_local_heads * self.head_dim]))\n        else:\n            self.wq_scale = self.wk_scale = self.wv_scale = self.wo_scale = None\n\n    def enable_cache(self):\n        self.cache_enabled = True\n\n    def disable_cache(self):\n        self.cache_enabled = False\n        self.cache_k, self.cache_v = None, None\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        bsz, seqlen, _ = x.shape\n        xq = forward_linear_with_scale_and_bias(x, self.wq, self.wq_scale, self.wq_bias)\n        xk = forward_linear_with_scale_and_bias(x, self.wk, self.wk_scale, self.wk_bias)\n        xv = forward_linear_with_scale_and_bias(x, self.wv, self.wv_scale, self.wv_bias)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_k = forward_linear_with_scale_and_bias(adapter, self.wk, self.wk_scale, self.wk_bias)\n            adapter_k = adapter_k.view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_v = forward_linear_with_scale_and_bias(adapter, self.wv, self.wv_scale, self.wv_bias)\n            adapter_v = adapter_v.view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_k = adapter_k.transpose(1, 2)\n            adapter_v = adapter_v.transpose(1, 2)\n        keys = xk\n        values = xv\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n\n        if self.cache_enabled:\n            if self.cache_k is None:\n                assert start_pos == 0\n                self.cache_k, self.cache_v = keys, values\n            else:\n                assert self.cache_k.size(2) >= start_pos\n                self.cache_k = torch.cat([self.cache_k[:, :, :start_pos], keys], dim=2)\n                self.cache_v = torch.cat([self.cache_v[:, :, :start_pos], values], dim=2)\n                keys, values = self.cache_k, self.cache_v\n\n        output = self._forward_scaled_dot_product_attention(xq, keys, values, mask)\n        if adapter is not None:\n            output += self.gate[\n                :, self.head_start : self.head_end\n            ].tanh().half() * self._forward_scaled_dot_product_attention(xq, adapter_k, adapter_v)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return forward_linear_with_scale_and_bias(output, self.wo, self.wo_scale, self.wo_bias)\n\n    def _forward_scaled_dot_product_attention(self, q, k, v, mask=None):\n        if hasattr(F, \"scaled_dot_product_attention\"):\n            return F.scaled_dot_product_attention(q, k, v, mask >= 0 if mask is not None else None)\n        else:\n            scores = torch.matmul(q, k.transpose(2, 3)) / math.sqrt(self.head_dim)\n            if mask is not None:\n                scores = scores + mask\n            scores = F.softmax(scores.float(), dim=-1).type_as(q)\n            output = torch.matmul(scores, v)\n            return output\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n        add_bias: bool = False,\n        add_scale: bool = False,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.dim, self.hidden_dim = dim, hidden_dim\n        mp_size = fs_init.get_model_parallel_world_size()\n        mp_rank = fs_init.get_model_parallel_rank()\n        self.mp_dim_st = self.hidden_dim // mp_size * mp_rank\n        self.mp_dim_ed = self.hidden_dim // mp_size * (mp_rank + 1)\n\n        self.w1 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n        self.w2 = RowParallelLinear(hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x)\n        self.w3 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n\n        if add_bias:\n            self.w1_bias, self.w3_bias = [nn.Parameter(torch.zeros([hidden_dim // mp_size])) for _ in range(2)]\n            self.w2_bias = nn.Parameter(torch.zeros([dim]))\n        else:\n            self.w1_bias = self.w2_bias = self.w3_bias = None\n\n        if add_scale:\n            self.w1_scale, self.w3_scale = [nn.Parameter(torch.ones([dim])) for _ in range(2)]\n            self.w2_scale = nn.Parameter(torch.ones([hidden_dim // mp_size]))\n        else:\n            self.w1_scale = self.w2_scale = self.w3_scale = None\n\n    def forward(self, x):\n        return forward_linear_with_scale_and_bias(\n            F.silu(forward_linear_with_scale_and_bias(x, self.w1, self.w1_scale, self.w1_bias))\n            * forward_linear_with_scale_and_bias(x, self.w3, self.w3_scale, self.w3_bias),\n            self.w2,\n            self.w2_scale,\n            self.w2_bias,\n        )\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim,\n            hidden_dim=4 * args.dim,\n            multiple_of=args.multiple_of,\n            add_bias=args.add_bias,\n            add_scale=args.add_scale,\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        if params.adapter_len * params.adapter_layer > 0:\n            self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n        _bsz, seqlen = examples.shape\n\n        h = self.tok_embeddings(examples)\n        freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n\n        if self.adapter_len * self.adapter_layer > 0:\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n            adapter_index = 0\n            adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, self.params.dim).unsqueeze(1)\n            for layer in self.layers[-1 * self.adapter_layer :]:\n                h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n                adapter_index = adapter_index + 1\n\n        else:\n            for layer in self.layers:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        h = self.norm(h)\n        output = self.output(h)\n        output = output[:, :-1, :].reshape(-1, self.vocab_size)\n        labels = labels[:, 1:].flatten()\n\n        c_loss = self.criterion(output, labels)\n        return c_loss\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        if self.adapter_len * self.adapter_layer > 0:\n            adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, self.params.dim).unsqueeze(1)\n        if seqlen == 1:\n            mask = None\n        elif start_pos == 0:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=1).type_as(h)\n        else:\n            raise NotImplementedError()\n\n        for i, layer in enumerate(self.layers):\n            adapter_index = i - (len(self.layers) - self.adapter_layer)\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half() if adapter_index >= 0 else None)\n\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])\n        return output.float()\n\n    def enable_cache(self):\n        for layer in self.layers:\n            layer.attention.enable_cache()\n\n    def disable_cache(self):\n        for layer in self.layers:\n            layer.attention.disable_cache()","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.ModelArgs","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.llama.model.ModelArgs#L16-L31","kind":"class","name":"ModelArgs","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":16,"end_line":31,"context_start_line":1,"context_end_line":51,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.nn.functional as F\nfrom fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_seq_len: int = 2048\n\n    adapter_len: int = 0\n    adapter_layer: int = 0\n\n    add_bias: bool = False\n    add_scale: bool = False\n    train_norm: bool = False\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return (output * self.weight).type_as(x)\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.RMSNorm","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.llama.model.RMSNorm#L34-L45","kind":"class","name":"RMSNorm","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":34,"end_line":45,"context_start_line":14,"context_end_line":65,"code":"\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_seq_len: int = 2048\n\n    adapter_len: int = 0\n    adapter_layer: int = 0\n\n    add_bias: bool = False\n    add_scale: bool = False\n    train_norm: bool = False\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return (output * self.weight).type_as(x)\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.precompute_freqs_cis","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model.precompute_freqs_cis#L48-L53","kind":"function","name":"precompute_freqs_cis","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":48,"end_line":53,"context_start_line":28,"context_end_line":73,"code":"\n    add_bias: bool = False\n    add_scale: bool = False\n    train_norm: bool = False\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return (output * self.weight).type_as(x)\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.reshape_for_broadcast","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model.reshape_for_broadcast#L56-L61","kind":"function","name":"reshape_for_broadcast","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":56,"end_line":61,"context_start_line":36,"context_end_line":81,"code":"        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return (output * self.weight).type_as(x)\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\ndef forward_linear_with_scale_and_bias(x, module, scale=None, bias=None):\n    if scale is not None:\n        x = x * scale\n    x = module(x)\n    if bias is not None:","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.apply_rotary_emb","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model.apply_rotary_emb#L64-L74","kind":"function","name":"apply_rotary_emb","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":64,"end_line":74,"context_start_line":44,"context_end_line":94,"code":"        output = self._norm(x.float()).type_as(x)\n        return (output * self.weight).type_as(x)\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\ndef forward_linear_with_scale_and_bias(x, module, scale=None, bias=None):\n    if scale is not None:\n        x = x * scale\n    x = module(x)\n    if bias is not None:\n        x = x + bias\n    return x\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n        self.dim = args.dim\n\n        self.wq = ColumnParallelLinear(","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.forward_linear_with_scale_and_bias","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model.forward_linear_with_scale_and_bias#L77-L83","kind":"function","name":"forward_linear_with_scale_and_bias","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":77,"end_line":83,"context_start_line":57,"context_end_line":103,"code":"    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)\n\n\ndef apply_rotary_emb(\n    xq: torch.Tensor,\n    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\ndef forward_linear_with_scale_and_bias(x, module, scale=None, bias=None):\n    if scale is not None:\n        x = x * scale\n    x = module(x)\n    if bias is not None:\n        x = x + bias\n    return x\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n        self.dim = args.dim\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.Attention","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.llama.model.Attention#L86-L204","kind":"class","name":"Attention","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":86,"end_line":204,"context_start_line":66,"context_end_line":224,"code":"    xk: torch.Tensor,\n    freqs_cis: torch.Tensor,\n) -> Tuple[torch.Tensor, torch.Tensor]:\n    xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2))\n    xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2))\n    freqs_cis = reshape_for_broadcast(freqs_cis, xq_)\n    xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3)\n    xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3)\n    return xq_out.type_as(xq), xk_out.type_as(xk)\n\n\ndef forward_linear_with_scale_and_bias(x, module, scale=None, bias=None):\n    if scale is not None:\n        x = x * scale\n    x = module(x)\n    if bias is not None:\n        x = x + bias\n    return x\n\n\nclass Attention(nn.Module):\n    def __init__(self, args: ModelArgs):\n        super().__init__()\n\n        self.n_local_heads = args.n_heads // fs_init.get_model_parallel_world_size()\n        self.head_dim = args.dim // args.n_heads\n        self.dim = args.dim\n\n        self.wq = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wk = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wv = ColumnParallelLinear(\n            args.dim,\n            args.n_heads * self.head_dim,\n            bias=False,\n            gather_output=False,\n            init_method=lambda x: x,\n        )\n        self.wo = RowParallelLinear(\n            args.n_heads * self.head_dim, args.dim, bias=False, input_is_parallel=True, init_method=lambda x: x\n        )\n\n        self.gate = torch.nn.Parameter(torch.zeros(1, args.n_heads, 1, 1))\n        self.head_start = self.n_local_heads * fs_init.get_model_parallel_rank()\n        self.head_end = self.n_local_heads * (fs_init.get_model_parallel_rank() + 1)\n\n        self.cache_enabled = False\n        self.cache_k, self.cache_v = None, None\n\n        if args.add_bias:\n            self.wq_bias, self.wk_bias, self.wv_bias = [\n                nn.Parameter(torch.zeros([self.n_local_heads * self.head_dim])) for _ in range(3)\n            ]\n            self.wo_bias = nn.Parameter(torch.zeros([args.dim]))\n        else:\n            self.wq_bias = self.wk_bias = self.wv_bias = self.wo_bias = None\n\n        if args.add_scale:\n            self.wq_scale, self.wk_scale, self.wv_scale = [nn.Parameter(torch.ones([args.dim])) for _ in range(3)]\n            self.wo_scale = nn.Parameter(torch.ones([self.n_local_heads * self.head_dim]))\n        else:\n            self.wq_scale = self.wk_scale = self.wv_scale = self.wo_scale = None\n\n    def enable_cache(self):\n        self.cache_enabled = True\n\n    def disable_cache(self):\n        self.cache_enabled = False\n        self.cache_k, self.cache_v = None, None\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n        bsz, seqlen, _ = x.shape\n        xq = forward_linear_with_scale_and_bias(x, self.wq, self.wq_scale, self.wq_bias)\n        xk = forward_linear_with_scale_and_bias(x, self.wk, self.wk_scale, self.wk_bias)\n        xv = forward_linear_with_scale_and_bias(x, self.wv, self.wv_scale, self.wv_bias)\n\n        xq = xq.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xk = xk.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n        xv = xv.view(bsz, seqlen, self.n_local_heads, self.head_dim)\n\n        xq, xk = apply_rotary_emb(xq, xk, freqs_cis=freqs_cis)\n\n        if adapter is not None:\n            adapter_len = adapter.shape[1]\n            adapter_k = forward_linear_with_scale_and_bias(adapter, self.wk, self.wk_scale, self.wk_bias)\n            adapter_k = adapter_k.view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_v = forward_linear_with_scale_and_bias(adapter, self.wv, self.wv_scale, self.wv_bias)\n            adapter_v = adapter_v.view(1, adapter_len, self.n_local_heads, self.head_dim).repeat(bsz, 1, 1, 1)\n            adapter_k = adapter_k.transpose(1, 2)\n            adapter_v = adapter_v.transpose(1, 2)\n        keys = xk\n        values = xv\n\n        xq = xq.transpose(1, 2)\n        keys = keys.transpose(1, 2)\n        values = values.transpose(1, 2)\n\n        if self.cache_enabled:\n            if self.cache_k is None:\n                assert start_pos == 0\n                self.cache_k, self.cache_v = keys, values\n            else:\n                assert self.cache_k.size(2) >= start_pos\n                self.cache_k = torch.cat([self.cache_k[:, :, :start_pos], keys], dim=2)\n                self.cache_v = torch.cat([self.cache_v[:, :, :start_pos], values], dim=2)\n                keys, values = self.cache_k, self.cache_v\n\n        output = self._forward_scaled_dot_product_attention(xq, keys, values, mask)\n        if adapter is not None:\n            output += self.gate[\n                :, self.head_start : self.head_end\n            ].tanh().half() * self._forward_scaled_dot_product_attention(xq, adapter_k, adapter_v)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return forward_linear_with_scale_and_bias(output, self.wo, self.wo_scale, self.wo_bias)\n\n    def _forward_scaled_dot_product_attention(self, q, k, v, mask=None):\n        if hasattr(F, \"scaled_dot_product_attention\"):\n            return F.scaled_dot_product_attention(q, k, v, mask >= 0 if mask is not None else None)\n        else:\n            scores = torch.matmul(q, k.transpose(2, 3)) / math.sqrt(self.head_dim)\n            if mask is not None:\n                scores = scores + mask\n            scores = F.softmax(scores.float(), dim=-1).type_as(q)\n            output = torch.matmul(scores, v)\n            return output\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n        add_bias: bool = False,\n        add_scale: bool = False,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.dim, self.hidden_dim = dim, hidden_dim\n        mp_size = fs_init.get_model_parallel_world_size()\n        mp_rank = fs_init.get_model_parallel_rank()\n        self.mp_dim_st = self.hidden_dim // mp_size * mp_rank\n        self.mp_dim_ed = self.hidden_dim // mp_size * (mp_rank + 1)","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.FeedForward","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.llama.model.FeedForward#L207-L249","kind":"class","name":"FeedForward","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":207,"end_line":249,"context_start_line":187,"context_end_line":269,"code":"        if adapter is not None:\n            output += self.gate[\n                :, self.head_start : self.head_end\n            ].tanh().half() * self._forward_scaled_dot_product_attention(xq, adapter_k, adapter_v)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return forward_linear_with_scale_and_bias(output, self.wo, self.wo_scale, self.wo_bias)\n\n    def _forward_scaled_dot_product_attention(self, q, k, v, mask=None):\n        if hasattr(F, \"scaled_dot_product_attention\"):\n            return F.scaled_dot_product_attention(q, k, v, mask >= 0 if mask is not None else None)\n        else:\n            scores = torch.matmul(q, k.transpose(2, 3)) / math.sqrt(self.head_dim)\n            if mask is not None:\n                scores = scores + mask\n            scores = F.softmax(scores.float(), dim=-1).type_as(q)\n            output = torch.matmul(scores, v)\n            return output\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n        add_bias: bool = False,\n        add_scale: bool = False,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.dim, self.hidden_dim = dim, hidden_dim\n        mp_size = fs_init.get_model_parallel_world_size()\n        mp_rank = fs_init.get_model_parallel_rank()\n        self.mp_dim_st = self.hidden_dim // mp_size * mp_rank\n        self.mp_dim_ed = self.hidden_dim // mp_size * (mp_rank + 1)\n\n        self.w1 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n        self.w2 = RowParallelLinear(hidden_dim, dim, bias=False, input_is_parallel=True, init_method=lambda x: x)\n        self.w3 = ColumnParallelLinear(dim, hidden_dim, bias=False, gather_output=False, init_method=lambda x: x)\n\n        if add_bias:\n            self.w1_bias, self.w3_bias = [nn.Parameter(torch.zeros([hidden_dim // mp_size])) for _ in range(2)]\n            self.w2_bias = nn.Parameter(torch.zeros([dim]))\n        else:\n            self.w1_bias = self.w2_bias = self.w3_bias = None\n\n        if add_scale:\n            self.w1_scale, self.w3_scale = [nn.Parameter(torch.ones([dim])) for _ in range(2)]\n            self.w2_scale = nn.Parameter(torch.ones([hidden_dim // mp_size]))\n        else:\n            self.w1_scale = self.w2_scale = self.w3_scale = None\n\n    def forward(self, x):\n        return forward_linear_with_scale_and_bias(\n            F.silu(forward_linear_with_scale_and_bias(x, self.w1, self.w1_scale, self.w1_bias))\n            * forward_linear_with_scale_and_bias(x, self.w3, self.w3_scale, self.w3_bias),\n            self.w2,\n            self.w2_scale,\n            self.w2_bias,\n        )\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim,\n            hidden_dim=4 * args.dim,\n            multiple_of=args.multiple_of,\n            add_bias=args.add_bias,\n            add_scale=args.add_scale,\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.TransformerBlock","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.llama.model.TransformerBlock#L252-L276","kind":"class","name":"TransformerBlock","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":252,"end_line":276,"context_start_line":232,"context_end_line":296,"code":"            self.w2_bias = nn.Parameter(torch.zeros([dim]))\n        else:\n            self.w1_bias = self.w2_bias = self.w3_bias = None\n\n        if add_scale:\n            self.w1_scale, self.w3_scale = [nn.Parameter(torch.ones([dim])) for _ in range(2)]\n            self.w2_scale = nn.Parameter(torch.ones([hidden_dim // mp_size]))\n        else:\n            self.w1_scale = self.w2_scale = self.w3_scale = None\n\n    def forward(self, x):\n        return forward_linear_with_scale_and_bias(\n            F.silu(forward_linear_with_scale_and_bias(x, self.w1, self.w1_scale, self.w1_bias))\n            * forward_linear_with_scale_and_bias(x, self.w3, self.w3_scale, self.w3_bias),\n            self.w2,\n            self.w2_scale,\n            self.w2_bias,\n        )\n\n\nclass TransformerBlock(nn.Module):\n    def __init__(self, layer_id: int, args: ModelArgs):\n        super().__init__()\n        self.n_heads = args.n_heads\n        self.dim = args.dim\n        self.head_dim = args.dim // args.n_heads\n        self.attention = Attention(args)\n        self.feed_forward = FeedForward(\n            dim=args.dim,\n            hidden_dim=4 * args.dim,\n            multiple_of=args.multiple_of,\n            add_bias=args.add_bias,\n            add_scale=args.add_scale,\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        if params.adapter_len * params.adapter_layer > 0:\n            self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.Transformer","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.llama.model.Transformer#L279-L365","kind":"class","name":"Transformer","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":279,"end_line":365,"context_start_line":259,"context_end_line":365,"code":"        self.feed_forward = FeedForward(\n            dim=args.dim,\n            hidden_dim=4 * args.dim,\n            multiple_of=args.multiple_of,\n            add_bias=args.add_bias,\n            add_scale=args.add_scale,\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        if params.adapter_len * params.adapter_layer > 0:\n            self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n        _bsz, seqlen = examples.shape\n\n        h = self.tok_embeddings(examples)\n        freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n\n        if self.adapter_len * self.adapter_layer > 0:\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n            adapter_index = 0\n            adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, self.params.dim).unsqueeze(1)\n            for layer in self.layers[-1 * self.adapter_layer :]:\n                h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n                adapter_index = adapter_index + 1\n\n        else:\n            for layer in self.layers:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        h = self.norm(h)\n        output = self.output(h)\n        output = output[:, :-1, :].reshape(-1, self.vocab_size)\n        labels = labels[:, 1:].flatten()\n\n        c_loss = self.criterion(output, labels)\n        return c_loss\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        if self.adapter_len * self.adapter_layer > 0:\n            adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, self.params.dim).unsqueeze(1)\n        if seqlen == 1:\n            mask = None\n        elif start_pos == 0:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=1).type_as(h)\n        else:\n            raise NotImplementedError()\n\n        for i, layer in enumerate(self.layers):\n            adapter_index = i - (len(self.layers) - self.adapter_layer)\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half() if adapter_index >= 0 else None)\n\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])\n        return output.float()\n\n    def enable_cache(self):\n        for layer in self.layers:\n            layer.attention.enable_cache()\n\n    def disable_cache(self):\n        for layer in self.layers:\n            layer.attention.disable_cache()","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.__init__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model.__init__#L280-L301","kind":"function","name":"__init__","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":280,"end_line":301,"context_start_line":260,"context_end_line":321,"code":"            dim=args.dim,\n            hidden_dim=4 * args.dim,\n            multiple_of=args.multiple_of,\n            add_bias=args.add_bias,\n            add_scale=args.add_scale,\n        )\n        self.layer_id = layer_id\n        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)\n        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)\n\n    def forward(\n        self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor], adapter=None\n    ):\n\n        h = x + self.attention.forward(self.attention_norm(x), start_pos, freqs_cis, mask, adapter)\n        out = h + self.feed_forward.forward(self.ffn_norm(h))\n        return out\n\n\nclass Transformer(nn.Module):\n    def __init__(self, params: ModelArgs):\n        super().__init__()\n        self.params = params\n        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        if params.adapter_len * params.adapter_layer > 0:\n            self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n        _bsz, seqlen = examples.shape\n\n        h = self.tok_embeddings(examples)\n        freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n\n        if self.adapter_len * self.adapter_layer > 0:\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n            adapter_index = 0\n            adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, self.params.dim).unsqueeze(1)\n            for layer in self.layers[-1 * self.adapter_layer :]:\n                h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n                adapter_index = adapter_index + 1","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model._norm","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model._norm#L40-L41","kind":"function","name":"_norm","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":40,"end_line":41,"context_start_line":20,"context_end_line":61,"code":"    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5\n\n    max_seq_len: int = 2048\n\n    adapter_len: int = 0\n    adapter_layer: int = 0\n\n    add_bias: bool = False\n    add_scale: bool = False\n    train_norm: bool = False\n\n\nclass RMSNorm(torch.nn.Module):\n    def __init__(self, dim: int, eps: float = 1e-6):\n        super().__init__()\n        self.eps = eps\n        self.weight = nn.Parameter(torch.ones(dim))\n\n    def _norm(self, x):\n        return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)\n\n    def forward(self, x):\n        output = self._norm(x.float()).type_as(x)\n        return (output * self.weight).type_as(x)\n\n\ndef precompute_freqs_cis(dim: int, end: int, theta: float = 10000.0):\n    freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))\n    t = torch.arange(end, device=freqs.device)  # type: ignore\n    freqs = torch.outer(t, freqs).float()  # type: ignore\n    freqs_cis = torch.polar(torch.ones_like(freqs), freqs)  # complex64\n    return freqs_cis\n\n\ndef reshape_for_broadcast(freqs_cis: torch.Tensor, x: torch.Tensor):\n    ndim = x.ndim\n    assert 0 <= 1 < ndim\n    assert freqs_cis.shape == (x.shape[1], x.shape[-1])\n    shape = [d if i == 1 or i == ndim - 1 else 1 for i, d in enumerate(x.shape)]\n    return freqs_cis.view(*shape)","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.forward","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model.forward#L303-L333","kind":"function","name":"forward","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":303,"end_line":333,"context_start_line":283,"context_end_line":353,"code":"        self.vocab_size = params.vocab_size\n        self.n_layers = params.n_layers\n        self.tok_embeddings = ParallelEmbedding(params.vocab_size, params.dim, init_method=lambda x: x)\n\n        if params.adapter_len * params.adapter_layer > 0:\n            self.adapter_query = nn.Embedding(params.adapter_len * params.adapter_layer, params.dim)\n        self.adapter_len = params.adapter_len\n        self.adapter_layer = params.adapter_layer\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        self.layers = torch.nn.ModuleList()\n        for layer_id in range(params.n_layers):\n            self.layers.append(TransformerBlock(layer_id, params))\n\n        self.norm = RMSNorm(params.dim, eps=params.norm_eps)\n        self.output = ColumnParallelLinear(params.dim, params.vocab_size, bias=False, init_method=lambda x: x)\n\n        self.freqs_cis = precompute_freqs_cis(self.params.dim // self.params.n_heads, self.params.max_seq_len * 2)\n\n    def forward(self, examples, labels):\n        _bsz, seqlen = examples.shape\n\n        h = self.tok_embeddings(examples)\n        freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = freqs_cis[:seqlen]\n        mask = None\n        mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=h.device)\n        mask = torch.triu(mask, diagonal=0 + 1).type_as(h)\n        start_pos = 0\n\n        if self.adapter_len * self.adapter_layer > 0:\n            for layer in self.layers[: -1 * self.adapter_layer]:\n                h = layer(h, start_pos, freqs_cis, mask)\n            adapter_index = 0\n            adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, self.params.dim).unsqueeze(1)\n            for layer in self.layers[-1 * self.adapter_layer :]:\n                h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n                adapter_index = adapter_index + 1\n\n        else:\n            for layer in self.layers:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        h = self.norm(h)\n        output = self.output(h)\n        output = output[:, :-1, :].reshape(-1, self.vocab_size)\n        labels = labels[:, 1:].flatten()\n\n        c_loss = self.criterion(output, labels)\n        return c_loss\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        if self.adapter_len * self.adapter_layer > 0:\n            adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, self.params.dim).unsqueeze(1)\n        if seqlen == 1:\n            mask = None\n        elif start_pos == 0:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=1).type_as(h)\n        else:\n            raise NotImplementedError()\n\n        for i, layer in enumerate(self.layers):\n            adapter_index = i - (len(self.layers) - self.adapter_layer)\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half() if adapter_index >= 0 else None)","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.enable_cache","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model.enable_cache#L359-L361","kind":"function","name":"enable_cache","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":359,"end_line":361,"context_start_line":339,"context_end_line":365,"code":"        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        if self.adapter_len * self.adapter_layer > 0:\n            adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, self.params.dim).unsqueeze(1)\n        if seqlen == 1:\n            mask = None\n        elif start_pos == 0:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=1).type_as(h)\n        else:\n            raise NotImplementedError()\n\n        for i, layer in enumerate(self.layers):\n            adapter_index = i - (len(self.layers) - self.adapter_layer)\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half() if adapter_index >= 0 else None)\n\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])\n        return output.float()\n\n    def enable_cache(self):\n        for layer in self.layers:\n            layer.attention.enable_cache()\n\n    def disable_cache(self):\n        for layer in self.layers:\n            layer.attention.disable_cache()","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.disable_cache","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model.disable_cache#L363-L365","kind":"function","name":"disable_cache","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":363,"end_line":365,"context_start_line":343,"context_end_line":365,"code":"        if seqlen == 1:\n            mask = None\n        elif start_pos == 0:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=1).type_as(h)\n        else:\n            raise NotImplementedError()\n\n        for i, layer in enumerate(self.layers):\n            adapter_index = i - (len(self.layers) - self.adapter_layer)\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half() if adapter_index >= 0 else None)\n\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])\n        return output.float()\n\n    def enable_cache(self):\n        for layer in self.layers:\n            layer.attention.enable_cache()\n\n    def disable_cache(self):\n        for layer in self.layers:\n            layer.attention.disable_cache()","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model._forward_scaled_dot_product_attention","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model._forward_scaled_dot_product_attention#L195-L204","kind":"function","name":"_forward_scaled_dot_product_attention","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":195,"end_line":204,"context_start_line":175,"context_end_line":224,"code":"\n        if self.cache_enabled:\n            if self.cache_k is None:\n                assert start_pos == 0\n                self.cache_k, self.cache_v = keys, values\n            else:\n                assert self.cache_k.size(2) >= start_pos\n                self.cache_k = torch.cat([self.cache_k[:, :, :start_pos], keys], dim=2)\n                self.cache_v = torch.cat([self.cache_v[:, :, :start_pos], values], dim=2)\n                keys, values = self.cache_k, self.cache_v\n\n        output = self._forward_scaled_dot_product_attention(xq, keys, values, mask)\n        if adapter is not None:\n            output += self.gate[\n                :, self.head_start : self.head_end\n            ].tanh().half() * self._forward_scaled_dot_product_attention(xq, adapter_k, adapter_v)\n        output = output.transpose(1, 2).contiguous().view(bsz, seqlen, -1)\n\n        return forward_linear_with_scale_and_bias(output, self.wo, self.wo_scale, self.wo_bias)\n\n    def _forward_scaled_dot_product_attention(self, q, k, v, mask=None):\n        if hasattr(F, \"scaled_dot_product_attention\"):\n            return F.scaled_dot_product_attention(q, k, v, mask >= 0 if mask is not None else None)\n        else:\n            scores = torch.matmul(q, k.transpose(2, 3)) / math.sqrt(self.head_dim)\n            if mask is not None:\n                scores = scores + mask\n            scores = F.softmax(scores.float(), dim=-1).type_as(q)\n            output = torch.matmul(scores, v)\n            return output\n\n\nclass FeedForward(nn.Module):\n    def __init__(\n        self,\n        dim: int,\n        hidden_dim: int,\n        multiple_of: int,\n        add_bias: bool = False,\n        add_scale: bool = False,\n    ):\n        super().__init__()\n        hidden_dim = int(2 * hidden_dim / 3)\n        hidden_dim = multiple_of * ((hidden_dim + multiple_of - 1) // multiple_of)\n\n        self.dim, self.hidden_dim = dim, hidden_dim\n        mp_size = fs_init.get_model_parallel_world_size()\n        mp_rank = fs_init.get_model_parallel_rank()\n        self.mp_dim_st = self.hidden_dim // mp_size * mp_rank\n        self.mp_dim_ed = self.hidden_dim // mp_size * (mp_rank + 1)","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.model.forward_inference","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.model.forward_inference#L336-L357","kind":"function","name":"forward_inference","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":336,"end_line":357,"context_start_line":316,"context_end_line":365,"code":"                h = layer(h, start_pos, freqs_cis, mask)\n            adapter_index = 0\n            adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, self.params.dim).unsqueeze(1)\n            for layer in self.layers[-1 * self.adapter_layer :]:\n                h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half())\n                adapter_index = adapter_index + 1\n\n        else:\n            for layer in self.layers:\n                h = layer(h, start_pos, freqs_cis, mask)\n\n        h = self.norm(h)\n        output = self.output(h)\n        output = output[:, :-1, :].reshape(-1, self.vocab_size)\n        labels = labels[:, 1:].flatten()\n\n        c_loss = self.criterion(output, labels)\n        return c_loss\n\n    @torch.inference_mode()\n    def forward_inference(self, tokens: torch.Tensor, start_pos: int):\n        _bsz, seqlen = tokens.shape\n        h = self.tok_embeddings(tokens)\n        self.freqs_cis = self.freqs_cis.to(h.device)\n        freqs_cis = self.freqs_cis[start_pos : start_pos + seqlen]\n        if self.adapter_len * self.adapter_layer > 0:\n            adapter = self.adapter_query.weight.reshape(-1, self.adapter_len, self.params.dim).unsqueeze(1)\n        if seqlen == 1:\n            mask = None\n        elif start_pos == 0:\n            mask = torch.full((1, 1, seqlen, seqlen), float(\"-inf\"), device=tokens.device)\n            mask = torch.triu(mask, diagonal=1).type_as(h)\n        else:\n            raise NotImplementedError()\n\n        for i, layer in enumerate(self.layers):\n            adapter_index = i - (len(self.layers) - self.adapter_layer)\n            h = layer(h, start_pos, freqs_cis, mask, adapter[adapter_index].half() if adapter_index >= 0 else None)\n\n        h = self.norm(h)\n        output = self.output(h[:, -1, :])\n        return output.float()\n\n    def enable_cache(self):\n        for layer in self.layers:\n            layer.attention.enable_cache()\n\n    def disable_cache(self):\n        for layer in self.layers:\n            layer.attention.disable_cache()","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.tokenizer","uri":"program://LLaMA-Adapter/module/llama_adapter_v2_chat65b.llama.tokenizer#L1-L38","kind":"module","name":"llama_adapter_v2_chat65b.llama.tokenizer","path":"llama_adapter_v2_chat65b/llama/tokenizer.py","language":"python","start_line":1,"end_line":38,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.tokenizer.Tokenizer","uri":"program://LLaMA-Adapter/class/llama_adapter_v2_chat65b.llama.tokenizer.Tokenizer#L13-L38","kind":"class","name":"Tokenizer","path":"llama_adapter_v2_chat65b/llama/tokenizer.py","language":"python","start_line":13,"end_line":38,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.tokenizer.__init__","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.tokenizer.__init__#L14-L26","kind":"function","name":"__init__","path":"llama_adapter_v2_chat65b/llama/tokenizer.py","language":"python","start_line":14,"end_line":26,"context_start_line":1,"context_end_line":38,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.tokenizer.encode","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.tokenizer.encode#L28-L35","kind":"function","name":"encode","path":"llama_adapter_v2_chat65b/llama/tokenizer.py","language":"python","start_line":28,"end_line":35,"context_start_line":8,"context_end_line":38,"code":"from sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"py:llama_adapter_v2_chat65b.llama.tokenizer.decode","uri":"program://LLaMA-Adapter/function/llama_adapter_v2_chat65b.llama.tokenizer.decode#L37-L38","kind":"function","name":"decode","path":"llama_adapter_v2_chat65b/llama/tokenizer.py","language":"python","start_line":37,"end_line":38,"context_start_line":17,"context_end_line":38,"code":"        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()\n        self.bos_id: int = self.sp_model.bos_id()\n        self.eos_id: int = self.sp_model.eos_id()\n        self.pad_id: int = self.sp_model.pad_id()\n        logger.info(f\"#words: {self.n_words} - BOS ID: {self.bos_id} - EOS ID: {self.eos_id}\")\n        assert self.sp_model.vocab_size() == self.sp_model.get_piece_size()\n\n    def encode(self, s: str, bos: bool, eos: bool) -> List[int]:\n        assert type(s) is str\n        t = self.sp_model.encode(s)\n        if bos:\n            t = [self.bos_id] + t\n        if eos:\n            t = t + [self.eos_id]\n        return t\n\n    def decode(self, t: List[int]) -> str:\n        return self.sp_model.decode(t)","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:setup.py","uri":"program://LLaMA-Adapter/file/setup.py","kind":"file","name":"setup.py","path":"setup.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom setuptools import find_packages, setup\n\nsetup(name=\"llama\", version=\"0.0.0\", packages=find_packages())","source_hash":"2ef5fd81323c276fca0f7106c2f0d98d12499ec49e5b7ee8de222a315e9153d7","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:example.py","uri":"program://LLaMA-Adapter/file/example.py","kind":"file","name":"example.py","path":"example.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport json\nimport os\nimport sys\nimport time\nfrom pathlib import Path\nfrom typing import Tuple\n\nimport fire\nimport torch\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import LLaMA, ModelArgs, Tokenizer, Transformer\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"","source_hash":"6c23b7ee72462699c540418da3efb30bfb4b4dcc5d82e513cc14c99635883587","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:setup.cfg","uri":"program://LLaMA-Adapter/file/setup.cfg","kind":"file","name":"setup.cfg","path":"setup.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom setuptools import find_packages, setup\n\nsetup(name=\"llama\", version=\"0.0.0\", packages=find_packages())","source_hash":"2ef5fd81323c276fca0f7106c2f0d98d12499ec49e5b7ee8de222a315e9153d7","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:utils/quantization.py","uri":"program://LLaMA-Adapter/file/utils/quantization.py","kind":"file","name":"utils/quantization.py","path":"utils/quantization.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport bitsandbytes as bnb\n\n\n'''\nlit-llama\n'''\nclass Linear8bitLt(bnb.nn.Linear8bitLt):\n    \"\"\"Wraps `bnb.nn.Linear8bitLt` and enables instantiation directly on the device and\n    re-quantizaton when loading the state dict.\n\n\n    \"\"\"\n\n    def __init__(self, *args, **kwargs):\n        super().__init__(*args, **kwargs, has_fp16_weights=False, threshold=6.0)\n\n        # We quantize the initial weight here so we don't end up filling the device\n        # memory with float32 weights which could lead to OOM.\n        self._quantize_weight(self.weight.data)\n","source_hash":"8936905ab12763c28c0324a49457f6680d9ed1c568a746d284ca64bf506a2f60","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/demo.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/demo.py","kind":"file","name":"llama_adapter_v2_multimodal7b/demo.py","path":"llama_adapter_v2_multimodal7b/demo.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":20,"code":"import cv2\nimport llama\nimport torch\nfrom PIL import Image\n\ndevice = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\nllama_dir = \"/path/to/LLaMA/\"\n\n# choose from BIAS-7B, LORA-BIAS-7B, CAPTION-7B.pth\nmodel, preprocess = llama.load(\"BIAS-7B\", llama_dir, device)\nmodel.eval()\n\nprompt = llama.format_prompt('Please introduce this painting.')\nimg = Image.fromarray(cv2.imread(\"../docs/logo_v1.png\"))\nimg = preprocess(img).unsqueeze(0).to(device)\n\nresult = model.generate(img, [prompt])[0]\n\nprint(result)","source_hash":"4aa8d6e038dbd5203bfaf76e378cc880d5ffe3cd6d7e567fd92e18bebada90d7","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/main_finetune.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/main_finetune.py","kind":"file","name":"llama_adapter_v2_multimodal7b/main_finetune.py","path":"llama_adapter_v2_multimodal7b/main_finetune.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import FinetuneDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_finetune import train_one_epoch\n","source_hash":"671fe114125e0c1e86a43b56d22a85b59683c3fe053378e5743eeaa5c1506dfa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/main_pretrain.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/main_pretrain.py","kind":"file","name":"llama_adapter_v2_multimodal7b/main_pretrain.py","path":"llama_adapter_v2_multimodal7b/main_pretrain.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import PretrainDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_pretrain import train_one_epoch\n","source_hash":"25f96771b527c0cf0ac998c008639df7231da9c09660c1f448c7451fa94e1d64","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/engine_finetune.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/engine_finetune.py","kind":"file","name":"llama_adapter_v2_multimodal7b/engine_finetune.py","path":"llama_adapter_v2_multimodal7b/engine_finetune.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/gradio_app.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/gradio_app.py","kind":"file","name":"llama_adapter_v2_multimodal7b/gradio_app.py","path":"llama_adapter_v2_multimodal7b/gradio_app.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import cv2\nimport gradio as gr\nimport torch\nfrom PIL import Image\n\nimport llama\n\n\ndevice = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n\nllama_dir = \"/path/to/LLaMA/\"\n\nmodel, preprocess = llama.load(\"BIAS-7B\", llama_dir, device)\nmodel.half()\nmodel.eval()\n\ndef multi_modal_generate(\n    img_path: str,\n    prompt: str,\n    max_gen_len=256,\n    temperature: float = 0.1,","source_hash":"800829419bded0f9a137c7489c91d1883db81fc4ca25045b5176b75c169f6eb2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/engine_pretrain.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/engine_pretrain.py","kind":"file","name":"llama_adapter_v2_multimodal7b/engine_pretrain.py","path":"llama_adapter_v2_multimodal7b/engine_pretrain.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/util/extract_adapter_from_checkpoint.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/util/extract_adapter_from_checkpoint.py","kind":"file","name":"llama_adapter_v2_multimodal7b/util/extract_adapter_from_checkpoint.py","path":"llama_adapter_v2_multimodal7b/util/extract_adapter_from_checkpoint.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\n\ndef save(full_model, path, model_type = 'BIAS'):\n    if model_type == 'BIAS':\n        keys = [\n            f'visual_blocks.{i}.{key}.{suffix}'\n            for i in range(8)\n            for key in ['norm1', 'attn.qkv', 'attn.proj', 'norm2', 'mlp.fc1', 'mlp.fc2']\n            for suffix in ['weight', 'bias']\n        ] + [\n            f'llama.layers.{i}.{key}'\n            for i in range(32)\n            for key in ['attention.gate', 'attention.wq.bias', 'attention.wo.bias', 'feed_forward.w1.bias', 'feed_forward.w2.bias', 'feed_forward.w3.bias', 'attention_norm.weight', 'ffn_norm.weight']\n        ] + [\n            f'{base_key}.{suffix}'\n            for base_key in ['clip_proj_norm', 'visual_proj_norm', 'visual_proj', 'clip_proj']\n            for suffix in ['weight', 'bias']\n        ] + ['llama.norm.weight', 'visual_query.weight', 'adapter_query.weight']\n\n    \n    elif model_type == 'LORA':","source_hash":"cdf86474579900b917d88f2488ef9b151ba11a87246765b5ec3640b5c83d4838","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/util/misc.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/util/misc.py","kind":"file","name":"llama_adapter_v2_multimodal7b/util/misc.py","path":"llama_adapter_v2_multimodal7b/util/misc.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\nimport urllib\nfrom tqdm import tqdm\n\nimport torch","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/util/lr_sched.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/util/lr_sched.py","kind":"file","name":"llama_adapter_v2_multimodal7b/util/lr_sched.py","path":"llama_adapter_v2_multimodal7b/util/lr_sched.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs \n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * \\\n            (1. + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs)))\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"4ab5d5633bda0be9173ec91570bb3050326d942582ded2267702b53c3ac87c2c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/util/evaluate_mme.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/util/evaluate_mme.py","kind":"file","name":"llama_adapter_v2_multimodal7b/util/evaluate_mme.py","path":"llama_adapter_v2_multimodal7b/util/evaluate_mme.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import os\nimport glob\nimport argparse\nfrom tqdm import tqdm\nimport PIL\nfrom PIL import Image\nimport torch\nimport torch.distributed as dist\nfrom torch.utils.data import Dataset\nimport cv2\nfrom llama.llama_adapter import LLaMA_adapter\n\nDATA_DIR = \"./MME_Benchmark_release_version\"\n\ndef get_image(image):\n    if type(image) is str:\n        try:\n            return Image.open(image).convert(\"RGB\")\n        except Exception as e:\n            print(f\"Fail to read image: {image}\")\n            exit(-1)","source_hash":"d5272f273a26eb7d95fa9871bd0d6cd466b816f5c81cded28a669bda03955961","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/data/dataset.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/data/dataset.py","kind":"file","name":"llama_adapter_v2_multimodal7b/data/dataset.py","path":"llama_adapter_v2_multimodal7b/data/dataset.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport yaml\nfrom torch.utils.data import Dataset\nfrom PIL import Image\nimport json\nimport llama.utils\nfrom llama import Tokenizer\nimport copy\nimport torchvision.transforms as transforms\nimport pandas as pd\nimport random\nimport cv2\n\ntry:\n    from torchvision.transforms import InterpolationMode\n    BICUBIC = InterpolationMode.BICUBIC\nexcept ImportError:\n    BICUBIC = Image.BICUBIC\n\n\nPROMPT_DICT = {","source_hash":"55d2ae8775c52c638ce4f2a13f0f9bfd9921264010bda2eb46998367cbe47b7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/llama/llama_adapter.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/llama/llama_adapter.py","kind":"file","name":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","path":"llama_adapter_v2_multimodal7b/llama/llama_adapter.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import os\nimport json\nfrom pathlib import Path\n\nimport clip\nimport torch\nimport torch.nn as nn\nfrom timm.models.vision_transformer import Block\n\nfrom .llama import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer\nfrom .utils import sample_top_p, _download\n\n\nclass LLaMA_adapter(nn.Module):\n\n    def __init__(self, llama_ckpt_dir, llama_tokenizer,\n                 max_seq_len=512, max_batch_size=1,\n                 clip_model='ViT-L/14',\n                 v_embed_dim=768, v_depth=8,\n                 v_num_heads=16, v_mlp_ratio=4.0,","source_hash":"6f105b9367340faa7f96f79d116c5fd0a0b7a6ff77bb93ce043d37810ef0452c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/llama/llama.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/llama/llama.py","kind":"file","name":"llama_adapter_v2_multimodal7b/llama/llama.py","path":"llama_adapter_v2_multimodal7b/llama/llama.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\n\nimport torch\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\nimport torch.nn.functional as F\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5","source_hash":"94161e720fd61d7672643bcb9d3b72ce603f1ec1c01c8caf370d7e9d9a0450f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/llama/__init__.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/llama/__init__.py","kind":"file","name":"llama_adapter_v2_multimodal7b/llama/__init__.py","path":"llama_adapter_v2_multimodal7b/llama/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":4,"code":"from .llama import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer\nfrom .llama_adapter import *\nfrom .utils import format_prompt","source_hash":"1f2d46064701de2419f628100b108dfddf2dbe8f21fa4773c2027ada9b3729e4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/llama/utils.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/llama/utils.py","kind":"file","name":"llama_adapter_v2_multimodal7b/llama/utils.py","path":"llama_adapter_v2_multimodal7b/llama/utils.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import os\nimport urllib\nimport hashlib\nimport warnings\n\nfrom tqdm import tqdm\nimport torch\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token\n\n\ndef format_prompt(instruction, input=None):","source_hash":"72a4d5400ca4d22079d19105a2a641bbf30f652005b5e7c6888b7106f87aae4c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_multimodal7b/llama/tokenizer.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_multimodal7b/llama/tokenizer.py","kind":"file","name":"llama_adapter_v2_multimodal7b/llama/tokenizer.py","path":"llama_adapter_v2_multimodal7b/llama/tokenizer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/extract_adapter_from_checkpoint.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/extract_adapter_from_checkpoint.py","kind":"file","name":"alpaca_finetuning_v1/extract_adapter_from_checkpoint.py","path":"alpaca_finetuning_v1/extract_adapter_from_checkpoint.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":15,"code":"import torch\n\nmodel = torch.load(\"./checkpoint/checkpoint-4.pth\", map_location=\"cpu\")\nnew_model = dict()\nweight_list = [\"layers.\" + str(i) + \".attention.gate\" for i in range(32)]\nold_weight_list = [\"layers.\" + str(i) + \".attention.gate\" for i in range(32)]\nweight_list = weight_list + [\"adapter_query.weight\"]\n\nprint(weight_list)\nprint(model[\"model\"][\"adapter_query.weight\"].shape)\n\nfor i in range(len(weight_list)):\n    new_model[weight_list[i]] = model[\"model\"][weight_list[i]]\n\ntorch.save(new_model, \"adapter_adapter_len10_layer30_epoch5.pth\")","source_hash":"9adf0123c67c13105f937d04e39a03742783601f02d8483d772abaf3d222ffbc","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/models_llama_adapter.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/models_llama_adapter.py","kind":"file","name":"alpaca_finetuning_v1/models_llama_adapter.py","path":"alpaca_finetuning_v1/models_llama_adapter.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import json\n\nimport torch\n\nfrom llama import ModelArgs, Tokenizer, Transformer\n\n\ndef Llama7B_adapter(args, **kwargs):\n\n    llama_model_path = args.llama_model_path\n    model_name = \"7B\"\n\n    checkpoint = torch.load(llama_model_path + model_name + \"/consolidated.00.pth\", map_location=\"cpu\")\n    print(llama_model_path + model_name + \"/consolidated.00.pth\")\n\n    with open(llama_model_path + model_name + \"/params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=args.max_seq_len,\n        max_batch_size=32,","source_hash":"0890bdd4d67183d4e58df19168966750005c5cc195ba3cfb50884e31dcc24d7a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/engine_finetuning.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/engine_finetuning.py","kind":"file","name":"alpaca_finetuning_v1/engine_finetuning.py","path":"alpaca_finetuning_v1/engine_finetuning.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\nimport util.lr_sched as lr_sched\nimport util.misc as misc\n\n\ndef train_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n\n    model.train(True)","source_hash":"e90ffdf70e0a5bc1217c32533638a96c91df1ec263f2805e7eecde3789ceb5a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/finetuning.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/finetuning.py","kind":"file","name":"alpaca_finetuning_v1/finetuning.py","path":"alpaca_finetuning_v1/finetuning.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import argparse\nimport copy\nimport datetime\nimport json\nimport os\nimport time\nfrom pathlib import Path\n\nimport models_llama_adapter\nimport numpy as np\nimport timm.optim.optim_factory as optim_factory\nimport torch\nimport torch.backends.cudnn as cudnn\nimport util.misc as misc\nfrom engine_finetuning import train_one_epoch, val_one_epoch\nfrom torch.utils.data import Dataset\nfrom torch.utils.tensorboard import SummaryWriter\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\n\nfrom llama import Tokenizer\n","source_hash":"a0ec94f11bf4623533eb76da6c0ad77e1e0697d4cd35193d37bc7f4e445185d1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/finetuning.sh","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/finetuning.sh","kind":"file","name":"alpaca_finetuning_v1/finetuning.sh","path":"alpaca_finetuning_v1/finetuning.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import argparse\nimport copy\nimport datetime\nimport json\nimport os\nimport time\nfrom pathlib import Path\n\nimport models_llama_adapter\nimport numpy as np\nimport timm.optim.optim_factory as optim_factory\nimport torch\nimport torch.backends.cudnn as cudnn\nimport util.misc as misc\nfrom engine_finetuning import train_one_epoch, val_one_epoch\nfrom torch.utils.data import Dataset\nfrom torch.utils.tensorboard import SummaryWriter\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\n\nfrom llama import Tokenizer\n","source_hash":"a0ec94f11bf4623533eb76da6c0ad77e1e0697d4cd35193d37bc7f4e445185d1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/util/misc.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/util/misc.py","kind":"file","name":"alpaca_finetuning_v1/util/misc.py","path":"alpaca_finetuning_v1/util/misc.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\n\nimport torch\nimport torch.distributed as dist\nfrom torch import inf","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/util/datasets.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/util/datasets.py","kind":"file","name":"alpaca_finetuning_v1/util/datasets.py","path":"alpaca_finetuning_v1/util/datasets.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\n\nimport os\n\nimport PIL\nfrom timm.data import create_transform\nfrom timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD\nfrom torchvision import datasets, transforms\n\n\ndef build_dataset(is_train, args):\n    transform = build_transform(is_train, args)\n","source_hash":"b40d68d3bfae8ebdf0b0c35884f609d3d8a46dc73c5e6397d8f537a389ee9b1d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/util/lr_decay.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/util/lr_decay.py","kind":"file","name":"alpaca_finetuning_v1/util/lr_decay.py","path":"alpaca_finetuning_v1/util/lr_decay.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# ELECTRA https://github.com/google-research/electra\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport json\n\n\ndef param_groups_lrd(model, weight_decay=0.05, no_weight_decay_list=[], layer_decay=.75):\n    \"\"\"\n    Parameter groups for layer-wise lr decay\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L58\n    \"\"\"\n    param_group_names = {}\n    param_groups = {}","source_hash":"c21f7ac4070e7dc3ef9fe3a904b9ab59b80082d136b70d99cf7853c5ceb62dba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/util/lr_sched.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/util/lr_sched.py","kind":"file","name":"alpaca_finetuning_v1/util/lr_sched.py","path":"alpaca_finetuning_v1/util/lr_sched.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs\n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * (\n            1.0 + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs))\n        )\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:","source_hash":"136a8dba00b53af7934e7377cc6a0dce1e84fef604e55673ce8460b109aa6d0d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/util/pos_embed.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/util/pos_embed.py","kind":"file","name":"alpaca_finetuning_v1/util/pos_embed.py","path":"alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# Position embedding utils\n# --------------------------------------------------------\n\nimport numpy as np\nimport torch\n\n\n# --------------------------------------------------------\n# 2D sine-cosine position embedding\n# References:\n# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py\n# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/util/lars.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/util/lars.py","kind":"file","name":"alpaca_finetuning_v1/util/lars.py","path":"alpaca_finetuning_v1/util/lars.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# LARS optimizer, implementation from MoCo v3:\n# https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\n\nimport torch\n\n\nclass LARS(torch.optim.Optimizer):\n    \"\"\"\n    LARS optimizer, no rate scaling or weight decay for parameters <= 1D.\n    \"\"\"\n    def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, trust_coefficient=0.001):\n        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, trust_coefficient=trust_coefficient)\n        super().__init__(params, defaults)\n","source_hash":"1bb75085eecf93b8d7ed7972c507495586e71a8f483bcdcd4f37c354f88a0fc0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/llama/generation.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/llama/generation.py","kind":"file","name":"alpaca_finetuning_v1/llama/generation.py","path":"alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/llama/model.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/llama/model.py","kind":"file","name":"alpaca_finetuning_v1/llama/model.py","path":"alpaca_finetuning_v1/llama/model.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport torch\nimport torch.nn.functional as F\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/llama/__init__.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/llama/__init__.py","kind":"file","name":"alpaca_finetuning_v1/llama/__init__.py","path":"alpaca_finetuning_v1/llama/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom .generation import LLaMA\nfrom .model import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer","source_hash":"1b13d2a1a3d2443f4b3d9bf4c6fe3e08c34a0a0e03cc72ba4fbee16bea46f992","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:alpaca_finetuning_v1/llama/tokenizer.py","uri":"program://LLaMA-Adapter/file/alpaca_finetuning_v1/llama/tokenizer.py","kind":"file","name":"alpaca_finetuning_v1/llama/tokenizer.py","path":"alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama/generation.py","uri":"program://LLaMA-Adapter/file/llama/generation.py","kind":"file","name":"llama/generation.py","path":"llama/generation.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama/model.py","uri":"program://LLaMA-Adapter/file/llama/model.py","kind":"file","name":"llama/model.py","path":"llama/model.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.nn.functional as F\nfrom fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama/__init__.py","uri":"program://LLaMA-Adapter/file/llama/__init__.py","kind":"file","name":"llama/__init__.py","path":"llama/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom .generation import LLaMA\nfrom .model import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer","source_hash":"1b13d2a1a3d2443f4b3d9bf4c6fe3e08c34a0a0e03cc72ba4fbee16bea46f992","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama/tokenizer.py","uri":"program://LLaMA-Adapter/file/llama/tokenizer.py","kind":"file","name":"llama/tokenizer.py","path":"llama/tokenizer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/demo.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/demo.py","kind":"file","name":"gorilla/finetune/demo.py","path":"gorilla/finetune/demo.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import json\nimport os\nimport sys\nimport time\nfrom pathlib import Path\nfrom typing import List, Tuple\n\nimport fire\nimport torch\nfrom model.tokenizer import Tokenizer\nfrom model.meta import MetaModel\n\n\nclass LLaMA:\n    def __init__(self, model, tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],","source_hash":"96f9e8f2bc294b12c1871e05e7ad362d902383cf881a3efc503a3ea7c865f265","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/main_finetune.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/main_finetune.py","kind":"file","name":"gorilla/finetune/main_finetune.py","path":"gorilla/finetune/main_finetune.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\nimport functools\nfrom functools import partial\n\nimport torch","source_hash":"719bf3b77c2305b815d1b3a47187e66b12c8034226c996c5995e3ff628d3204d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/transformer.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/transformer.py","kind":"file","name":"gorilla/finetune/transformer.py","path":"gorilla/finetune/transformer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved\n\"\"\"\nDETR Transformer class.\nCopy-paste from torch.nn.Transformer with modifications:\n    * positional encodings are passed in MHattention\n    * extra LN at the end of encoder is removed\n    * decoder returns a stack of activations from all decoding layers\n\"\"\"\nimport copy\nfrom typing import Optional, List\n\nimport torch\nimport torch.nn.functional as F\nfrom torch import nn, Tensor\nimport pdb\n\nclass Transformer(nn.Module):\n\n    def __init__(self, d_model=512, nhead=8, num_encoder_layers=6,\n                 num_decoder_layers=6, dim_feedforward=2048, dropout=0.1,\n                 activation=\"relu\", normalize_before=False,","source_hash":"2b88e8d795af4e51c03982332247f605be025e66c286a4ae59a11051c1106829","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/conversation.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/conversation.py","kind":"file","name":"gorilla/finetune/conversation.py","path":"gorilla/finetune/conversation.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Tuple\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Different separator style.\"\"\"\n    SINGLE = auto()\n    TWO = auto()\n\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n    system: str\n    roles: List[str]\n    messages: List[List[str]]\n    offset: int\n    sep_style: SeparatorStyle = SeparatorStyle.SINGLE\n    sep: str = \"###\"\n    sep2: str = None","source_hash":"26bc22f298f041b21f9c0877b66425f314dd0c1e0222e349779be6e1de949cf6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/global_configs.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/global_configs.py","kind":"file","name":"gorilla/finetune/global_configs.py","path":"gorilla/finetune/global_configs.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":4,"code":"tokenizer_path = '/data1/llma/tokenizer.model'\npetrel_conf = \"/mnt/petrelfs/share_data/gaopeng/ldy/petreloss_all.conf\"\npetrel_prefix = \"cluster_p_ssd:s3://falcon-refinedweb/data\"\ndata_meta_path = \"/mnt/petrelfs/share_data/gaopeng/ldy/falcon_list.json\"","source_hash":"ef7e1236d31fb389e2d529063d6ad7f7ba242fde9261f8943c12bd4b200d9e88","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/data_preprocess.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/data_preprocess.py","kind":"file","name":"gorilla/finetune/data_preprocess.py","path":"gorilla/finetune/data_preprocess.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import json\n\nPROMPT_DICT = {\n    \"prompt_input\": (\n        \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"\n    ),\n    \"prompt_no_input\": (\n        \"Below is an instruction that describes a task. \"\n        \"Write a response that appropriately completes the request.\\n\\n\"\n        \"### Instruction:\\n{instruction}\\n\\n### Response:\"\n    ),\n}\nimport pdb\npdb.set_trace()\n\ndatas = json.load(open('/home/pgao/stanford_alpaca/stanford_alpaca/alpaca_data.json'))\nprompt_input, prompt_no_input = PROMPT_DICT[\"prompt_input\"], PROMPT_DICT[\"prompt_no_input\"]\nsources = [\n        prompt_input.format_map(example) if example.get(\"input\", \"\") != \"\" else prompt_no_input.format_map(example)","source_hash":"4456043a285aeef35fe923702414b326d7cce06887487ce4942834dfe370e279","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/main_pretrain.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/main_pretrain.py","kind":"file","name":"gorilla/finetune/main_pretrain.py","path":"gorilla/finetune/main_pretrain.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\nimport functools\nfrom functools import partial\n\nimport torch","source_hash":"efb38fa3fda169a2a9cbde46b53e5cb3ad6bbf5b2f4d2c4aa8dc86da164f5ee6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/engine_finetune.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/engine_finetune.py","kind":"file","name":"gorilla/finetune/engine_finetune.py","path":"gorilla/finetune/engine_finetune.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\nimport json\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport math\nimport sys\nimport os\nfrom typing import Iterable\nimport contextlib\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\nimport pdb","source_hash":"04485c8d845a90473030686c62d742ae5b1837a84fdc72b8d89c5a51516f94e7","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/engine_pretrain.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/engine_pretrain.py","kind":"file","name":"gorilla/finetune/engine_pretrain.py","path":"gorilla/finetune/engine_pretrain.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\nimport json\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport math\nimport sys\nimport os\nfrom typing import Iterable\nimport contextlib\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\nimport pdb","source_hash":"04b42f9156ec322fcfc9789802a9bb6c60279c84ac1fa0ae1f941a8349a4260b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/submitit_pretrain.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/submitit_pretrain.py","kind":"file","name":"gorilla/finetune/submitit_pretrain.py","path":"gorilla/finetune/submitit_pretrain.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# A script to run multinode training with submitit.\n# --------------------------------------------------------\n\nimport argparse\nimport os\nimport uuid\nfrom pathlib import Path\n\nimport main_pretrain as trainer\nimport submitit\n\n\ndef parse_args():\n    trainer_parser = trainer.get_args_parser()\n    parser = argparse.ArgumentParser(\"Submitit for MAE pretrain\", parents=[trainer_parser])","source_hash":"ace6d14b743db09df699a998a1fec85bab2584ad8c7114ee591f912aea8101e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/util/misc.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/util/misc.py","kind":"file","name":"gorilla/finetune/util/misc.py","path":"gorilla/finetune/util/misc.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\nimport subprocess\n\nimport torch\nimport torch.distributed as dist","source_hash":"7f22be12988fb8a42e631aa2fd1019531d98ed7a323f98df46ac6ccd66a3a311","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/util/lr_decay.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/util/lr_decay.py","kind":"file","name":"gorilla/finetune/util/lr_decay.py","path":"gorilla/finetune/util/lr_decay.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# ELECTRA https://github.com/google-research/electra\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport json\n\n\ndef param_groups_lrd(model, weight_decay=0.05, no_weight_decay_list=[], layer_decay=.75):\n    \"\"\"\n    Parameter groups for layer-wise lr decay\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L58\n    \"\"\"\n    param_group_names = {}\n    param_groups = {}","source_hash":"ff0ea0cf26230819d56c2c839e6c8e1b8c4f9834c46a8da226d601042bd2cdd5","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/util/crop.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/util/crop.py","kind":"file","name":"gorilla/finetune/util/crop.py","path":"gorilla/finetune/util/crop.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\nimport torch\n\nfrom torchvision import transforms\nfrom torchvision.transforms import functional as F\n\n\nclass RandomResizedCrop(transforms.RandomResizedCrop):\n    \"\"\"\n    RandomResizedCrop for matching TF/TPU implementation: no for-loop is used.\n    This may lead to results different with torchvision's version.\n    Following BYOL's TF code:\n    https://github.com/deepmind/deepmind-research/blob/master/byol/utils/dataset.py#L206\n    \"\"\"","source_hash":"494b97fecddf698fa1089b9c16fae97a4fa22a809c1a4f908dcb09387d2b36ff","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/util/lr_sched.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/util/lr_sched.py","kind":"file","name":"gorilla/finetune/util/lr_sched.py","path":"gorilla/finetune/util/lr_sched.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\ndef adjust_learning_rate(optimizer, it, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if it < args.warmup_iters: # 1) linear warmup for warmup_iters steps\n        lr = args.lr * it / args.warmup_iters\n    elif it > args.lr_decay_iters: # 2) if it > lr_decay_iters, return min learning rate\n        lr = args.min_lr\n    else: # 3) in between, use cosine decay down to min learning rate\n        decay_ratio = (it - args.warmup_iters) / (args.lr_decay_iters - args.warmup_iters)\n        assert 0 <= decay_ratio <= 1\n        coeff = 0.5 * (1.0 + math.cos(math.pi * decay_ratio))  # coeff ranges 0..1\n        lr = args.min_lr + (args.lr - args.min_lr) * coeff\n\n    for param_group in optimizer.param_groups:","source_hash":"57ea96156d0b62bb6af56bda534e9f6f906d1e1b7939799c9b1402a9c659ab43","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/util/pos_embed.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/util/pos_embed.py","kind":"file","name":"gorilla/finetune/util/pos_embed.py","path":"gorilla/finetune/util/pos_embed.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# Position embedding utils\n# --------------------------------------------------------\n\nimport numpy as np\n\nimport torch\n\n# --------------------------------------------------------\n# 2D sine-cosine position embedding\n# References:\n# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py\n# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"","source_hash":"b50f20cb689db0bc58b40aec97a3767f124e9ae1c747efbb4f2c51a0b744a67f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/data/alpaca.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/data/alpaca.py","kind":"file","name":"gorilla/finetune/data/alpaca.py","path":"gorilla/finetune/data/alpaca.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport yaml\nfrom torch.utils.data import Dataset\nfrom PIL import Image\nimport json\nfrom model.tokenizer import Tokenizer\nimport copy\nimport torchvision.transforms as transforms\nimport pdb\n\ntry:\n    from torchvision.transforms import InterpolationMode\n    BICUBIC = InterpolationMode.BICUBIC\nexcept ImportError:\n    BICUBIC = Image.BICUBIC\n\n\ndef format_prompt(instruction, input=None):\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"","source_hash":"f1c356152d9979a79d2368bf2a2d3d8b55031e165d0ac81b96535f11fb44ddd1","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/model/meta.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/model/meta.py","kind":"file","name":"gorilla/finetune/model/meta.py","path":"gorilla/finetune/model/meta.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport torch.nn as nn\nimport json\nfrom .tokenizer import Tokenizer\nfrom . import LLM\nfrom global_configs import tokenizer_path\n\n\nclass MetaModel(nn.Module):\n    \"\"\" Masked Autoencoder with VisionTransformer backbone\n    \"\"\"\n    def __init__(self, llama_type, reversible_grad: bool, llama_config):\n        super().__init__()\n\n        self.criterion = torch.nn.CrossEntropyLoss(ignore_index=0)\n\n        ModelArgs = LLM.__dict__[llama_type].ModelArgs\n        Transformer = LLM.__dict__[llama_type].Transformer\n\n        with open(llama_config, \"r\") as f:\n            params = json.loads(f.read())","source_hash":"5ae0a461056ec2a2ee50a5c7efde1e77ef8c57d3751d844da3bc8629b25863a8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/model/tokenizer.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/model/tokenizer.py","kind":"file","name":"gorilla/finetune/model/tokenizer.py","path":"gorilla/finetune/model/tokenizer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/model/LLM/revllama.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/model/LLM/revllama.py","kind":"file","name":"gorilla/finetune/model/LLM/revllama.py","path":"gorilla/finetune/model/LLM/revllama.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\nimport copy\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\nimport sys\n\nimport torch\nfrom torch import nn\nimport torch.nn.functional as F\nfrom torch.nn import Embedding, Linear\nfrom torch.autograd import Function\n\nfrom apex.normalization import FusedRMSNorm as RMSNorm\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512","source_hash":"79d5e077c07b5695350f01cbb2e04cc59e0914040c249464fd1848d514fedcfb","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/model/LLM/llama.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/model/LLM/llama.py","kind":"file","name":"gorilla/finetune/model/LLM/llama.py","path":"gorilla/finetune/model/LLM/llama.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\nimport functools\n\nimport torch\nfrom torch import nn\nimport torch.nn.functional as F\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nfrom fairscale.nn.model_parallel.layers import (\n    ParallelEmbedding,\n    RowParallelLinear,\n    ColumnParallelLinear,\n)\n\nfrom apex.normalization import FusedRMSNorm as RMSNorm\n","source_hash":"6fccd06d56837aa7ee12bba38e8cd75c55439616bdc46b9612a55aa8e01bc8d6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/model/LLM/__init__.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/model/LLM/__init__.py","kind":"file","name":"gorilla/finetune/model/LLM/__init__.py","path":"gorilla/finetune/model/LLM/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":2,"code":"from . import llama\nfrom . import revllama","source_hash":"21ed032324d65aca518f464b9437d16cb0fec00c3ffcac3d384a2bc82c55d741","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/finetune/tools/get_consolidated_ckpt.py","uri":"program://LLaMA-Adapter/file/gorilla/finetune/tools/get_consolidated_ckpt.py","kind":"file","name":"gorilla/finetune/tools/get_consolidated_ckpt.py","path":"gorilla/finetune/tools/get_consolidated_ckpt.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\nimport sys\nimport os\nsys.path.append(os.path.abspath(__file__).rsplit('/', 2)[0])\n\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport time\nfrom pathlib import Path","source_hash":"9ec2fc9de26ef16ab90e936f2ae235df21297a616bde0935aca78b9e9a92421c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/setup.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/setup.py","kind":"file","name":"gorilla/inference/setup.py","path":"gorilla/inference/setup.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom setuptools import setup, find_packages\n\nsetup(name=\"llama\", version=\"0.0.0\", packages=find_packages())","source_hash":"550bf5cf8205e38353070d61d3a01af1e7174f35a40c625c1e976d1f1f4133e8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/example.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/example.py","kind":"file","name":"gorilla/inference/example.py","path":"gorilla/inference/example.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Tuple\nimport os\nimport sys\nimport torch\nimport fire\nimport time\nimport json\n\nfrom pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))\n    world_size = int(os.environ.get(\"WORLD_SIZE\", -1))","source_hash":"fe4f54077b8d1bfb1c05199e82ffcabb0dcae6414514e6447b296190105e9e7e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/gorilla_inference_llama_adapter_v1.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/gorilla_inference_llama_adapter_v1.py","kind":"file","name":"gorilla/inference/gorilla_inference_llama_adapter_v1.py","path":"gorilla/inference/gorilla_inference_llama_adapter_v1.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Tuple\nimport os\nimport sys\nimport torch\nimport fire\nimport time\nimport json\nfrom tqdm import tqdm\n\nfrom pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama_for_adapter import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))","source_hash":"d75f8b7173d048be89a4f376b8c4d45f18e70f843861578d687ac52a8192d849","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/gorilla_inference_full_finetune.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/gorilla_inference_full_finetune.py","kind":"file","name":"gorilla/inference/gorilla_inference_full_finetune.py","path":"gorilla/inference/gorilla_inference_full_finetune.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Tuple\nimport os\nimport sys\nimport torch\nimport fire\nimport time\nimport json\nfrom tqdm import tqdm\n\nfrom pathlib import Path\n\nfrom fairscale.nn.model_parallel.initialize import initialize_model_parallel\n\nfrom llama import ModelArgs, Transformer, Tokenizer, LLaMA\n\n\ndef setup_model_parallel() -> Tuple[int, int]:\n    local_rank = int(os.environ.get(\"LOCAL_RANK\", -1))","source_hash":"657c95387fb4aa14fdcfcf7f8422a42dd1e59b5f6f72e347c10196fcb98cbe01","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/llama_for_adapter/generation.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/llama_for_adapter/generation.py","kind":"file","name":"gorilla/inference/llama_for_adapter/generation.py","path":"gorilla/inference/llama_for_adapter/generation.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,","source_hash":"dec4599b5612695f005e136e918ffb9911eb81fa4a7fd3d37bb887a1a9cbee8f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/llama_for_adapter/model.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/llama_for_adapter/model.py","kind":"file","name":"gorilla/inference/llama_for_adapter/model.py","path":"gorilla/inference/llama_for_adapter/model.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.nn.functional as F\nfrom fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2","source_hash":"dab5112d6db00556409c66d24b2f4d2b25bbcb8ae286a282053457681c7d5682","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/llama_for_adapter/__init__.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/llama_for_adapter/__init__.py","kind":"file","name":"gorilla/inference/llama_for_adapter/__init__.py","path":"gorilla/inference/llama_for_adapter/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom .generation import LLaMA\nfrom .model import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer","source_hash":"1b13d2a1a3d2443f4b3d9bf4c6fe3e08c34a0a0e03cc72ba4fbee16bea46f992","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/llama_for_adapter/tokenizer.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/llama_for_adapter/tokenizer.py","kind":"file","name":"gorilla/inference/llama_for_adapter/tokenizer.py","path":"gorilla/inference/llama_for_adapter/tokenizer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/llama/generation.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/llama/generation.py","kind":"file","name":"gorilla/inference/llama/generation.py","path":"gorilla/inference/llama/generation.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.tokenizer import Tokenizer\nfrom llama.model import Transformer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,","source_hash":"ccf9877e509c2f5fa350117dd9385fce9f055329d36d43a5f4fe70a7b7b24b81","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/llama/model.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/llama/model.py","kind":"file","name":"gorilla/inference/llama/model.py","path":"gorilla/inference/llama/model.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\n\nimport torch\nfrom torch import nn\nimport torch.nn.functional as F\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nfrom fairscale.nn.model_parallel.layers import (\n    ParallelEmbedding,\n    RowParallelLinear,\n    ColumnParallelLinear,\n)\n\n\n@dataclass\nclass ModelArgs:","source_hash":"8bfe7d71eed0037c690eb09b8530b7599cfe938df26adce4de5d36638933a5bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/llama/__init__.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/llama/__init__.py","kind":"file","name":"gorilla/inference/llama/__init__.py","path":"gorilla/inference/llama/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom .generation import LLaMA\nfrom .model import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer","source_hash":"26c15dffa63a2ca509d1ff0857669f94e23f197b5504863e06af02f4f6393db5","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/inference/llama/tokenizer.py","uri":"program://LLaMA-Adapter/file/gorilla/inference/llama/tokenizer.py","kind":"file","name":"gorilla/inference/llama/tokenizer.py","path":"gorilla/inference/llama/tokenizer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()","source_hash":"4ffa0c212242ba914acf2698853250bc3cf682876fa6fcf7b9bceedf66cc2d5d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/extract_adapter_from_checkpoint.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/extract_adapter_from_checkpoint.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/extract_adapter_from_checkpoint.py","path":"gorilla/alpaca_finetuning_v1/extract_adapter_from_checkpoint.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport argparse\n\nargs = argparse.ArgumentParser(\"extract\", add_help=False)\n\nargs.add_argument(\"--model_path\", type=str)\n\nargs = args.parse_args()\n\nmodel = torch.load(args.model_path, map_location=\"cpu\")\nnew_model = dict()\nweight_list = [\"layers.\" + str(i) + \".attention.gate\" for i in range(32)]\nold_weight_list = [\"layers.\" + str(i) + \".attention.gate\" for i in range(32)]\nweight_list = weight_list + [\"adapter_query.weight\"]\n\nprint(weight_list)\nprint(model[\"model\"][\"adapter_query.weight\"].shape)\n\nfor i in range(len(weight_list)):\n    new_model[weight_list[i]] = model[\"model\"][weight_list[i]]\n","source_hash":"df2af3a9757f423ca8e8f05de35694c99a88a4a334f8c787e5481766a6e0f7f3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/models_llama_adapter.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/models_llama_adapter.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/models_llama_adapter.py","path":"gorilla/alpaca_finetuning_v1/models_llama_adapter.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import json\n\nimport torch\n\nfrom llama import ModelArgs, Tokenizer, Transformer\n\n\ndef Llama7B_adapter(args, **kwargs):\n\n    llama_model_path = args.llama_model_path\n    model_name = \"7B\"\n\n    checkpoint = torch.load(llama_model_path + \"/consolidated.00.pth\", map_location=\"cpu\")\n    print(llama_model_path + \"/consolidated.00.pth\")\n\n    with open(llama_model_path + \"/params.json\", \"r\") as f:\n        params = json.loads(f.read())\n\n    model_args: ModelArgs = ModelArgs(\n        max_seq_len=args.max_seq_len,\n        max_batch_size=32,","source_hash":"4e63b524896d6fd5c98dda7936b21db83f656d8eab48fbc3261b6709eaf69261","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/engine_finetuning.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/engine_finetuning.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/engine_finetuning.py","path":"gorilla/alpaca_finetuning_v1/engine_finetuning.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\nimport util.lr_sched as lr_sched\nimport util.misc as misc\n\n\ndef train_one_epoch(\n    model: torch.nn.Module,\n    data_loader: Iterable,\n    optimizer: torch.optim.Optimizer,\n    device: torch.device,\n    epoch: int,\n    loss_scaler,\n    log_writer=None,\n    args=None,\n):\n\n    model.train(True)","source_hash":"e90ffdf70e0a5bc1217c32533638a96c91df1ec263f2805e7eecde3789ceb5a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/finetuning.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/finetuning.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/finetuning.py","path":"gorilla/alpaca_finetuning_v1/finetuning.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import argparse\nimport copy\nimport datetime\nimport json\nimport os\nimport time\nfrom pathlib import Path\n\nimport models_llama_adapter\nimport numpy as np\nimport timm.optim.optim_factory as optim_factory\nimport torch\nimport torch.backends.cudnn as cudnn\nimport util.misc as misc\nfrom engine_finetuning import train_one_epoch, val_one_epoch\nfrom torch.utils.data import Dataset\nfrom torch.utils.tensorboard import SummaryWriter\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\n\nfrom llama import Tokenizer\n","source_hash":"9b86155ada31fbcad2a44259b26833df69bf9b5674e2f9924e077e6e06fc5474","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/util/misc.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/util/misc.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/util/misc.py","path":"gorilla/alpaca_finetuning_v1/util/misc.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\n\nimport torch\nimport torch.distributed as dist\nfrom torch import inf","source_hash":"20d3c767622be9a2bb9634feec27bf63815c5c58d0f3bac9e6fb9c115fa524d9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/util/datasets.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/util/datasets.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/util/datasets.py","path":"gorilla/alpaca_finetuning_v1/util/datasets.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# --------------------------------------------------------\n\nimport os\n\nimport PIL\nfrom timm.data import create_transform\nfrom timm.data.constants import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD\nfrom torchvision import datasets, transforms\n\n\ndef build_dataset(is_train, args):\n    transform = build_transform(is_train, args)\n","source_hash":"b40d68d3bfae8ebdf0b0c35884f609d3d8a46dc73c5e6397d8f537a389ee9b1d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/util/lr_decay.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/util/lr_decay.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/util/lr_decay.py","path":"gorilla/alpaca_finetuning_v1/util/lr_decay.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# ELECTRA https://github.com/google-research/electra\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport json\n\n\ndef param_groups_lrd(model, weight_decay=0.05, no_weight_decay_list=[], layer_decay=.75):\n    \"\"\"\n    Parameter groups for layer-wise lr decay\n    Following BEiT: https://github.com/microsoft/unilm/blob/master/beit/optim_factory.py#L58\n    \"\"\"\n    param_group_names = {}\n    param_groups = {}","source_hash":"c21f7ac4070e7dc3ef9fe3a904b9ab59b80082d136b70d99cf7853c5ceb62dba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/util/lr_sched.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/util/lr_sched.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/util/lr_sched.py","path":"gorilla/alpaca_finetuning_v1/util/lr_sched.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs\n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * (\n            1.0 + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs))\n        )\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:","source_hash":"136a8dba00b53af7934e7377cc6a0dce1e84fef604e55673ce8460b109aa6d0d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/util/pos_embed.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/util/pos_embed.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/util/pos_embed.py","path":"gorilla/alpaca_finetuning_v1/util/pos_embed.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# Position embedding utils\n# --------------------------------------------------------\n\nimport numpy as np\nimport torch\n\n\n# --------------------------------------------------------\n# 2D sine-cosine position embedding\n# References:\n# Transformer: https://github.com/tensorflow/models/blob/master/official/nlp/transformer/model_utils.py\n# MoCo v3: https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\ndef get_2d_sincos_pos_embed(embed_dim, grid_size, cls_token=False):\n    \"\"\"","source_hash":"c7ba54a63ba9439d2019a1c390d05f8a104e252ca5584dd768d3ee6550e34b5a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/util/lars.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/util/lars.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/util/lars.py","path":"gorilla/alpaca_finetuning_v1/util/lars.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# LARS optimizer, implementation from MoCo v3:\n# https://github.com/facebookresearch/moco-v3\n# --------------------------------------------------------\n\nimport torch\n\n\nclass LARS(torch.optim.Optimizer):\n    \"\"\"\n    LARS optimizer, no rate scaling or weight decay for parameters <= 1D.\n    \"\"\"\n    def __init__(self, params, lr=0, weight_decay=0, momentum=0.9, trust_coefficient=0.001):\n        defaults = dict(lr=lr, weight_decay=weight_decay, momentum=momentum, trust_coefficient=trust_coefficient)\n        super().__init__(params, defaults)\n","source_hash":"1bb75085eecf93b8d7ed7972c507495586e71a8f483bcdcd4f37c354f88a0fc0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/llama/generation.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/llama/generation.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/llama/generation.py","path":"gorilla/alpaca_finetuning_v1/llama/generation.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer):\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,","source_hash":"bb3e748b010935282107a61aeb52781e32f1bac71f8fa4115a226cb7db530951","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/llama/model.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/llama/model.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/llama/model.py","path":"gorilla/alpaca_finetuning_v1/llama/model.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport torch\nimport torch.nn.functional as F\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5","source_hash":"cc896836c087208d4f8dad83b5a61cbbb7b3b977fbb4648313a2ebee3a54e25e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/llama/__init__.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/llama/__init__.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/llama/__init__.py","path":"gorilla/alpaca_finetuning_v1/llama/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom .generation import LLaMA\nfrom .model import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer","source_hash":"1b13d2a1a3d2443f4b3d9bf4c6fe3e08c34a0a0e03cc72ba4fbee16bea46f992","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/alpaca_finetuning_v1/llama/tokenizer.py","uri":"program://LLaMA-Adapter/file/gorilla/alpaca_finetuning_v1/llama/tokenizer.py","kind":"file","name":"gorilla/alpaca_finetuning_v1/llama/tokenizer.py","path":"gorilla/alpaca_finetuning_v1/llama/tokenizer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/inference/apply_delta.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/inference/apply_delta.py","kind":"file","name":"gorilla/gorilla-main/inference/apply_delta.py","path":"gorilla/gorilla-main/inference/apply_delta.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"\"\"\"\nApply the delta weights on top of a base model.\n\nUsage:\npython3 apply_delta.py --base-model-path path/to/hf_llama/ --target-model-path path/to/gorilla-7b-hf-v0 --delta-path path/to/models--gorilla-llm--gorilla-7b-hf-delta-v0\n\nThanks to LMSYS for the template of this code.\n\"\"\"\nimport argparse\nimport gc\nimport glob\nimport json\nimport os\nimport shutil\nimport tempfile\n\nfrom huggingface_hub import snapshot_download\nimport torch\nfrom torch import nn\nfrom tqdm import tqdm\nfrom transformers import AutoTokenizer, AutoModelForCausalLM, AutoConfig","source_hash":"b33fef0c645fd573827306a76bf3f767a0fadbdbeb9f30cbd4e599fbc5a095c8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/inference/gorilla_eval.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/inference/gorilla_eval.py","kind":"file","name":"gorilla/gorilla-main/inference/gorilla_eval.py","path":"gorilla/gorilla-main/inference/gorilla_eval.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import json\nimport argparse\nimport os\nfrom tqdm import tqdm\nimport torch\nfrom transformers import (\n    AutoConfig,\n    AutoModel,\n    AutoModelForCausalLM,\n    AutoModelForSeq2SeqLM,\n    AutoTokenizer,\n    LlamaTokenizer,\n    LlamaForCausalLM,\n    T5Tokenizer,\n)\n\n# Load Gorilla Model from HF\ndef load_model(\n        model_path: str,\n        device: str,\n        num_gpus: int,","source_hash":"a5c70bd40bf9560b1db239ce916a8a3f7f4b2a798fed884fa623942ffcfca781","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/inference/serve/conv_template.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/inference/serve/conv_template.py","kind":"file","name":"gorilla/gorilla-main/inference/serve/conv_template.py","path":"gorilla/gorilla-main/inference/serve/conv_template.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"\"\"\"\nConversation prompt templates.\n\nThanks to LMSYS for the template of this code.\n\"\"\"\n\nimport dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Any, Dict\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Separator styles.\"\"\"\n\n    ADD_COLON_SINGLE = auto()\n    ADD_COLON_TWO = auto()\n    ADD_COLON_SPACE_SINGLE = auto()\n    NO_COLON_SINGLE = auto()\n    ADD_NEW_LINE_SINGLE = auto()\n    DOLLY = auto()\n    RWKV = auto()","source_hash":"44c9e37846637b52e24da58568da85acd590a857ed77bba33e3f34dbc59fae52","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","kind":"file","name":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","path":"gorilla/gorilla-main/inference/serve/gorilla_falcon_cli.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"\"\"\"\nChat with a model with command line interface.\n\nUsage:\npython3 -m gorilla_cli --model path/to/gorilla-7b-hf-v0\n\nThanks to LMSYS for the template of this code.\n\"\"\"\nimport argparse\nimport gc\nimport os\nimport re\nimport sys\nimport abc\nimport torch\nfrom transformers import (\n    AutoConfig,\n    AutoModel,\n    AutoModelForCausalLM,\n    AutoModelForSeq2SeqLM,\n    AutoTokenizer,","source_hash":"3bdbae2592db9c84ea5e72251871a044816e35c09c9b7c59adee9337373a14ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/inference/serve/gorilla_cli.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/inference/serve/gorilla_cli.py","kind":"file","name":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","path":"gorilla/gorilla-main/inference/serve/gorilla_cli.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"\"\"\"\nChat with a model with command line interface.\n\nUsage:\npython3 gorilla_cli.py --model-path path/to/gorilla-7b-hf-v0\n\nThanks to LMSYS for the template of this code.\n\"\"\"\nimport argparse\nimport gc\nimport os\nimport re\nimport sys\nimport abc\nimport torch\nfrom transformers import (\n    AutoConfig,\n    AutoModel,\n    AutoModelForCausalLM,\n    AutoModelForSeq2SeqLM,\n    AutoTokenizer,","source_hash":"f68b20c6df3c862339b802afa8fbaccb89d17ccf41018abb747b0aac353e3356","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/get_llm_responses.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/get_llm_responses.py","kind":"file","name":"gorilla/gorilla-main/eval/get_llm_responses.py","path":"gorilla/gorilla-main/eval/get_llm_responses.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import argparse\nimport sys\nimport json\nimport openai\nimport anthropic\nimport multiprocessing as mp\nimport time\n\ndef encode_question(question, api_name):\n    \"\"\"Encode multiple prompt instructions into a single string.\"\"\"\n    \n    prompts = []\n    if api_name == \"torchhub\":\n        domains = \"1. $DOMAIN is inferred from the task description and should include one of {Classification, Semantic Segmentation, Object Detection, Audio Separation, Video Classification, Text-to-Speech}.\"\n    elif api_name == \"huggingface\":\n        domains = \"1. $DOMAIN should include one of {Multimodal Feature Extraction, Multimodal Text-to-Image, Multimodal Image-to-Text, Multimodal Text-to-Video, \\\n        Multimodal Visual Question Answering, Multimodal Document Question Answer, Multimodal Graph Machine Learning, Computer Vision Depth Estimation,\\\n        Computer Vision Image Classification, Computer Vision Object Detection, Computer Vision Image Segmentation, Computer Vision Image-to-Image, \\\n        Computer Vision Unconditional Image Generation, Computer Vision Video Classification, Computer Vision Zero-Shor Image Classification, \\\n        Natural Language Processing Text Classification, Natural Language Processing Token Classification, Natural Language Processing Table Question Answering, \\\n        Natural Language Processing Question Answering, Natural Language Processing Zero-Shot Classification, Natural Language Processing Translation, \\","source_hash":"a1f2396797292461fb9aa31463a8180134e82de030722fa8f65f8ae18b9f957c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_th.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import argparse\nimport json\nfrom codebleu.parser import (\n    DFG_python,\n    DFG_java,\n    DFG_ruby,\n    DFG_go,\n    DFG_php,\n    DFG_javascript,\n    DFG_csharp,\n)\nfrom codebleu.parser import (\n    remove_comments_and_docstrings,\n    tree_to_token_index,\n    index_to_code_token,\n    tree_to_variable_index,\n)\nfrom tree_sitter import Language, Parser\nimport concurrent.futures\n\ndfg_function = {","source_hash":"2ef51deb15484e775819fcaadf0d4372048d1371641d7f1025bed961f7a664e0","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_hf.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import argparse\nimport json \nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\n# Get all the subtrees given a root_node\ndef get_all_sub_trees(root_node):","source_hash":"b8f8c1fd383b1a114370357504ffacbb0af6dfec89b4435f7bea9fcdee0775ee","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","path":"gorilla/gorilla-main/eval/eval-scripts/ast_eval_tf.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import argparse\nimport json \nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\n# Get all the subtrees given a root_node\ndef get_all_sub_trees(root_node):","source_hash":"fc0147c13cab79eb5d58dd85c2f4bdc7ab72c92c4613339a42993caaf32523ed","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_check.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_check.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_check.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_check.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\ndef calc_syntax_match(references, candidate, lang):","source_hash":"411e527290a836e47f9a7956d71dab64588928d88b05466da1cb880e8e917fea","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/weighted_ngram_match.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# -*- coding: utf-8 -*-\n# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\n# Natural Language Toolkit: BLEU Score\n#\n# Copyright (C) 2001-2020 NLTK Project\n# Authors: Chin Yee Lee, Hengfeng Li, Ruxin Hou, Calvin Tanujaya Lim\n# Contributors: Björn Mattsson, Dmitrijs Milajevs, Liling Tan\n# URL: <http://nltk.org/>\n# For license information, see LICENSE.TXT\n\n\"\"\"BLEU score implementation.\"\"\"\n\nimport math\nimport sys\nfrom fractions import Fraction\nimport warnings\nfrom collections import Counter\n\nfrom codebleu.utils import ngrams","source_hash":"81fe783f9745a36a9fb1e6c556e17937b7669a3e75bfd16d461cbf0eecdceb56","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/dataflow_match.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/dataflow_match.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/dataflow_match.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/dataflow_match.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom codebleu.parser import DFG_python\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\nimport pdb\n\ndfg_function={\n    'python':DFG_python\n}\n\ndef calc_dataflow_match(references, candidate, lang):\n    return corpus_dataflow_match([references], [candidate], lang)\n\ndef corpus_dataflow_match(references, candidates, lang):   \n    LANGUAGE = Language('codebleu/parser/my-languages.so', lang)\n    parser = Parser()","source_hash":"72304a13041dd77d7b16691cc48d96d584b13818680f9f8b08b94d38eb898841","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/utils.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/utils.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/utils.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/utils.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Natural Language Toolkit: Utility functions\n#\n# Copyright (C) 2001-2020 NLTK Project\n# Author: Steven Bird <stevenbird1@gmail.com>\n# URL: <http://nltk.org/>\n# For license information, see LICENSE.TXT\n\nfrom itertools import chain\n\ndef pad_sequence(\n    sequence,\n    n,\n    pad_left=False,\n    pad_right=False,\n    left_pad_symbol=None,\n    right_pad_symbol=None,\n):\n    \"\"\"\n    Returns a padded sequence of items before ngram extraction.\n        >>> list(pad_sequence([1,2,3,4,5], 2, pad_left=True, pad_right=True, left_pad_symbol='<s>', right_pad_symbol='</s>'))\n        ['<s>', 1, 2, 3, 4, 5, '</s>']","source_hash":"264e39c7dfb7355617861fed173838a0d4434365662f86b9e9abc2a31b61c455","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_match.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_match.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_match.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/syntax_match.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom codebleu.parser import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp\nfrom codebleu.parser import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom tree_sitter import Language, Parser\n\ndfg_function={\n    'python':DFG_python,\n    'java':DFG_java,\n    'ruby':DFG_ruby,\n    'go':DFG_go,\n    'php':DFG_php,\n    'javascript':DFG_javascript,\n    'c_sharp':DFG_csharp,\n}\n\ndef calc_syntax_match(references, candidate, lang):","source_hash":"81d088ac5225c2f00cbb49515483052cb3b2e032666654349f0c701caee43556","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/bleu.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# -*- coding: utf-8 -*-\n# Natural Language Toolkit: BLEU Score\n#\n# Copyright (C) 2001-2020 NLTK Project\n# Authors: Chin Yee Lee, Hengfeng Li, Ruxin Hou, Calvin Tanujaya Lim\n# Contributors: Björn Mattsson, Dmitrijs Milajevs, Liling Tan\n# URL: <http://nltk.org/>\n# For license information, see LICENSE.TXT\n\n\"\"\"BLEU score implementation.\"\"\"\n\nimport math\nimport sys\nfrom fractions import Fraction\nimport warnings\nfrom collections import Counter\n\nfrom codebleu.utils import ngrams\nimport pdb\n\n","source_hash":"11482645892c0f789e566617c016031c3c73cb7d666289eaf2535b0dc8bed2a4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/DFG.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom tree_sitter import Language, Parser\nfrom .utils import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\n\n\ndef DFG_python(root_node,index_to_code,states):\n    assignment=['assignment','augmented_assignment','for_in_clause']\n    if_statement=['if_statement']\n    for_statement=['for_statement']\n    while_statement=['while_statement']\n    do_first_statement=['for_in_clause'] \n    def_statement=['default_parameter']\n    states=states.copy() \n    if (len(root_node.children)==0 or root_node.type in ['string_literal','string','character_literal']) and root_node.type!='comment':        \n        idx,code=index_to_code[(root_node.start_point,root_node.end_point)]\n        if root_node.type==code:","source_hash":"aa834b9452568da78d6e37907ac83c94801d9d9630febced21d6c19ec3e414f4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/__init__.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/__init__.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/__init__.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":8,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom .utils import (remove_comments_and_docstrings,\n                   tree_to_token_index,\n                   index_to_code_token,\n                   tree_to_variable_index)\nfrom .DFG import DFG_python,DFG_java,DFG_ruby,DFG_go,DFG_php,DFG_javascript,DFG_csharp","source_hash":"d8cd00fc694868c34545c456819396015e07744a7c9163c480162213bd12daf9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/utils.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/utils.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/utils.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/utils.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nimport re\nfrom io import StringIO\nimport  tokenize\n\ndef remove_comments_and_docstrings(source,lang):\n    if lang in ['python']:\n        \"\"\"\n        Returns 'source' minus comments and docstrings.\n        \"\"\"\n        io_obj = StringIO(source)\n        out = \"\"\n        prev_toktype = tokenize.INDENT\n        last_lineno = -1\n        last_col = 0\n        for tok in tokenize.generate_tokens(io_obj.readline):\n            token_type = tok[0]\n            token_string = tok[1]\n            start_line, start_col = tok[2]","source_hash":"2ecb4ed83e010e05ab32e112e2cdae4eb0c050a5e1b8ecf0eb0533565c17c479","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/build.sh","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/build.sh","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/build.sh","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/build.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":14,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom tree_sitter import Language, Parser\n\nLanguage.build_library(\n  # Store the library in the `build` directory\n  'my-languages.so',\n\n  # Include one or more languages\n  [ 'tree-sitter-python',\n  ]\n)\n","source_hash":"2bbff97589bc4dae551a425cfad854eb5cf1275bd35c236866208c13c8e52187","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/build.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/build.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/build.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/build.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":14,"code":"# Copyright (c) Microsoft Corporation. \n# Licensed under the MIT license.\n\nfrom tree_sitter import Language, Parser\n\nLanguage.build_library(\n  # Store the library in the `build` directory\n  'my-languages.so',\n\n  # Include one or more languages\n  [ 'tree-sitter-python',\n  ]\n)\n","source_hash":"2bbff97589bc4dae551a425cfad854eb5cf1275bd35c236866208c13c8e52187","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/parameters.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/parameters.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/parameters.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/parameters.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":4,"code":"def g(h, i, /, j, *, k=100, **kwarg):\n    #       ^ operator\n    #             ^ operator\n    pass","source_hash":"71aeb9a20f6752310f9256cb0c5a7abf29a8e01aeb49f4857c6af0362ee0c072","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/keywords.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/keywords.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/keywords.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/keywords.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"if foo():\n# <- keyword\n    pass\n    # <- keyword\nelif bar():\n# <- keyword\n    pass\nelse:\n# <- keyword\n    foo\n\nreturn\n# ^ keyword\nraise e\n# ^ keyword\n\nfor i in foo():\n# <- keyword\n#   ^ variable\n#     ^ operator\n#        ^ function","source_hash":"5209cfe62baefa489f21f00ca457880a0908c54e3613ab5f73fb0d568e3d9d69","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/pattern_matching.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/pattern_matching.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/pattern_matching.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/test/highlight/pattern_matching.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"match command.split():\n# ^ keyword\n    case [\"quit\"]:\n    # ^ keyword\n        print(\"Goodbye!\")\n        quit_game()\n    case [\"look\"]:\n    # ^ keyword\n        current_room.describe()\n    case [\"get\", obj]:\n    # ^ keyword\n        character.get(obj, current_room)\n    case [\"go\", direction]:\n    # ^ keyword\n        current_room = current_room.neighbor(direction)\n    # The rest of your commands go here\n\nmatch command.split():\n# ^ keyword\n    case [\"drop\", *objects]:\n    # ^ keyword","source_hash":"9f5c10ed608893ffcbc0d85ea94ab67e50f903c25aee2982ad29f806f8298998","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/multiple-newlines.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/multiple-newlines.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/multiple-newlines.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/multiple-newlines.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"def hi():\n\n\n\n    print \"hi\"\n\n\ndef bye():\n    print \"bye\"\n\n\n\n\n\n\n\n\n\n\n\n","source_hash":"43b9ea50f59013b0204eac91c14ebccd6e8507780923516f7f0d40146adcbb68","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python2-grammar-crlf.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python2-grammar-crlf.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python2-grammar-crlf.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python2-grammar-crlf.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.test_support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\","source_hash":"42e9014edfae60326c9e26568e1ce04ca266616410b81ac5bcf09ab1424649bd","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/crlf-line-endings.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/crlf-line-endings.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/crlf-line-endings.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/crlf-line-endings.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"print a\n\nif b:    \n    if c:\n        d\n    e","source_hash":"cce5aa47eb68b83f709a86766ddfd75a7141d3d9708beeadc1a465acd8441796","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/trailing-whitespace.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/trailing-whitespace.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/trailing-whitespace.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/trailing-whitespace.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"print a    \n\nif b:    \n    if c:    \n        d\n    e     ","source_hash":"7d63af3a7bf0ad5f851005a540d9c0ee6bb92c07f90969361047ab0632a09eb3","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/simple-statements-without-trailing-newline.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/simple-statements-without-trailing-newline.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/simple-statements-without-trailing-newline.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/simple-statements-without-trailing-newline.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":1,"code":"pass; print \"hi\"","source_hash":"35b1cefdabf4ddfe171073aed03d903af42713ac1aab854e94e4c3ac09ee42dc","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/tabs.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/tabs.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/tabs.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/tabs.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"def set_password(args):\n\tpassword = args.password\n\twhile not password  :\n\t\tpassword1 = getpass(\"\" if args.quiet else \"Provide password: \")\n\t\tpassword_repeat = getpass(\"\" if args.quiet else \"Repeat password:  \")\n\t\tif password1 != password_repeat:\n\t\t\tprint(\"Passwords do not match, try again\")\n\t\telif len(password1) < 4:\n\t\t\tprint(\"Please provide at least 4 characters\")\n\t\telse:\n\t\t\tpassword = password1\n\n\tpassword_hash = passwd(password)\n\tcfg = BaseJSONConfigManager(config_dir=jupyter_config_dir())\n\tcfg.update('jupyter_notebook_config', {\n\t\t'NotebookApp': {\n\t\t\t'password': password_hash,\n\t\t}\n\t})\n\tif not args.quiet:\n\t\tprint(\"password stored in config dir: %s\" % jupyter_config_dir())","source_hash":"6f3c435bec254bbaff69f0835685f018db23d21e6cae8cb77aaaaff030a08075","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/compound-statement-without-trailing-newline.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/compound-statement-without-trailing-newline.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/compound-statement-without-trailing-newline.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/compound-statement-without-trailing-newline.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":4,"code":"\nclass Foo:\n  def bar():\n    print \"hi\"","source_hash":"8296cca55337701b6da541e9a622df9928f784620326d45bc62f32675cc185e8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\","source_hash":"eddc388af6d77394197d51276263b90a8f722bc12c58330829a22b1ce4baae38","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/mixed-spaces-tabs.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/mixed-spaces-tabs.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/mixed-spaces-tabs.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/mixed-spaces-tabs.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":4,"code":"def main():\n\tprint \"hello\"\n\t# 1 tab = 8 spaces in Python 2\n        return","source_hash":"2fe77b01e6e989bc120dd4985bbe06beb15ab84e8fb81350fc53b480f5f1cd9b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python3-grammar-crlf.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\","source_hash":"3ca5acc15ae82d4f5f380e4d4ea830873d22582e8530149c14cda2b24005374d","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python2-grammar.py","uri":"program://LLaMA-Adapter/file/gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python2-grammar.py","kind":"file","name":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python2-grammar.py","path":"gorilla/gorilla-main/eval/eval-scripts/codebleu/parser/tree-sitter-python/examples/python2-grammar.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Python test set -- part 1, grammar.\n# This just tests whether the parser accepts them all.\n\n# NOTE: When you run this test as a script from the command line, you\n# get warnings about certain hex/oct constants.  Since those are\n# issued by the parser, you can't suppress them by adding a\n# filterwarnings() call to this module.  Therefore, to shut up the\n# regression test, the filterwarnings() call has been added to\n# regrtest.py.\n\nfrom test.test_support import run_unittest, check_syntax_error\nimport unittest\nimport sys\n# testing import *\nfrom sys import *\n\nclass TokenTests(unittest.TestCase):\n\n    def testBackslash(self):\n        # Backslash means line continuation:\n        x = 1 \\","source_hash":"5008d0357851f744321f0ec3e8c6568214d5dd88e251c9a861ad533b1557d92a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/demo.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/demo.py","kind":"file","name":"imagebind_LLM/demo.py","path":"imagebind_LLM/demo.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import ImageBind.data as data\nimport llama\n\n\nllama_dir = \"/path/to/LLaMA\"\n\nmodel = llama.load(\"7B\", llama_dir, knn=True)\nmodel.eval()\n\ninputs = {}\nimage = data.load_and_transform_vision_data([\"examples/girl.jpg\"], device='cuda')\ninputs['Image'] = [image, 1]\naudio = data.load_and_transform_audio_data(['examples/girl_bgm.wav'], device='cuda')\ninputs['Audio'] = [audio, 1]\n\nresults = model.generate(\n    inputs,\n    [llama.format_prompt(\"Guess the girl's mood based on the background music and explain the reason?\")],\n    max_gen_len=256\n)\nresult = results[0].strip()","source_hash":"520a739d1e57efbfcb4b801c864db1ad9e97743e74ad4d5bd72c263157ecf4de","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/demo_3d.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/demo_3d.py","kind":"file","name":"imagebind_LLM/demo_3d.py","path":"imagebind_LLM/demo_3d.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":20,"code":"import ImageBind.data as data\nimport llama\n\n\nllama_dir = \"/path/to/LLaMA\"\n\nmodel = llama.load(\"7B\", llama_dir, knn=True)\nmodel.eval()\n\ninputs = {}\npoint = data.load_and_transform_point_cloud_data([\"examples/airplane.pt\"], device='cuda')\ninputs['Point'] = [point, 1]\n\nresults = model.generate(\n    inputs,\n    [llama.format_prompt(\"Describe the 3D object in detail.\")],\n    max_gen_len=256\n)\nresult = results[0].strip()\nprint(result)","source_hash":"5d4916dbfe01372c3c782cc75a980c7b5dc75c0af139a13d3dd22145241514a8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/main_finetune.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/main_finetune.py","kind":"file","name":"imagebind_LLM/main_finetune.py","path":"imagebind_LLM/main_finetune.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import FinetuneDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_finetune import train_one_epoch\n","source_hash":"ff1670518e18fa13f63278df89d074094d85ceb7f336ff9bb7b123e2abb8e9aa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/main_pretrain.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/main_pretrain.py","kind":"file","name":"imagebind_LLM/main_pretrain.py","path":"imagebind_LLM/main_pretrain.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport torch.backends.cudnn as cudnn\nfrom torch.utils.tensorboard import SummaryWriter\nfrom torch.utils.data import Dataset\n\nimport util.misc as misc\nfrom util.misc import NativeScalerWithGradNormCount as NativeScaler\nfrom llama.llama_adapter import LLaMA_adapter\n\nfrom data.dataset import PretrainDataset, transform_train\n\nimport argparse\nimport datetime\nimport json\nimport numpy as np\nimport os\nimport time\nfrom pathlib import Path\n\nfrom engine_pretrain import train_one_epoch\n","source_hash":"4a687bac8852db4420d7421898270cd7629d02950989a8cdbd34fba84d19402c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/engine_finetune.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/engine_finetune.py","kind":"file","name":"imagebind_LLM/engine_finetune.py","path":"imagebind_LLM/engine_finetune.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/gradio_app.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/gradio_app.py","kind":"file","name":"imagebind_LLM/gradio_app.py","path":"imagebind_LLM/gradio_app.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import os\nimport argparse\n\nimport gradio as gr\nimport plotly.graph_objects as go\nimport torch, numpy, random\nimport torch.cuda\n\nimport ImageBind.data as data\nfrom diffusers import StableUnCLIPImg2ImgPipeline\nfrom image_generate import image_generate\n\nimport llama\n\nparser = argparse.ArgumentParser()\nparser.add_argument(\n    \"--model\", default=\"7B\", type=str,\n    help=\"Name of or path to ImageBind-LLM pretrained checkpoint\",\n)\nparser.add_argument(\n    \"--llama_type\", default=\"7B_chinese\", type=str,","source_hash":"cd48ac1e298f556bfd6ff9dbfc3d5a25d76532dfee22a81cdf1f3c22b3843f39","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/image_generate.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/image_generate.py","kind":"file","name":"imagebind_LLM/image_generate.py","path":"imagebind_LLM/image_generate.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import llama\nimport torch\nimport numpy as np\n\n\n@torch.inference_mode()\ndef image_generate(inputs, model: llama.LLaMA_adapter, pipe, prompt, cache_size, cache_t, cache_weight, knn=True, point_scale=5.):\n\n    embeddings = []\n    embeddings_weights = []\n\n    for input_type, (input, input_weight) in inputs.items():\n        if input_type in ['Image', 'Video']:\n            type = 'vision'\n        else:\n            type = input_type.lower()\n        embedding = model.image_bind({type : input}, prenorm=True)[1][type]\n        if type == 'point':\n            embedding = embedding / point_scale\n        embeddings.append(embedding)\n        embeddings_weights.append(input_weight)","source_hash":"4545f651278425e7793206653469106f5cd06367171233b17e7ee097fa2b155e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/convert_ckpt.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/convert_ckpt.py","kind":"file","name":"imagebind_LLM/convert_ckpt.py","path":"imagebind_LLM/convert_ckpt.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nfrom collections import OrderedDict\nimport argparse\nfrom pathlib import Path\n\nparser = argparse.ArgumentParser()\nparser.add_argument(\n    \"--ori\", required=True, type=str,\n    help=\"Name of or path to LLaMAAdapter pretrained checkpoint\",\n)\nparser.add_argument(\n    \"--target\", default=None,\n    help=\"target position for the ckpt\",\n)\nargs = parser.parse_args()\n\nori_ckpt_path = Path(args.ori)\ntarget_ckpt_path = ori_ckpt_path.with_stem(\"converted_\" + ori_ckpt_path.stem)\n\nckpt = torch.load(ori_ckpt_path, map_location='cpu')\n","source_hash":"9d4665a5d6453dfdc2e5f3622a37bfdcaaa771818947bff9644a614f03166c42","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/engine_pretrain.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/engine_pretrain.py","kind":"file","name":"imagebind_LLM/engine_pretrain.py","path":"imagebind_LLM/engine_pretrain.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import math\nimport sys\nfrom typing import Iterable\n\nimport torch\n\nimport util.misc as misc\nimport util.lr_sched as lr_sched\n\nfrom llama import LLaMA_adapter\n\ndef train_one_epoch(model: LLaMA_adapter,\n                    data_loader: Iterable, optimizer: torch.optim.Optimizer,\n                    device: torch.device, epoch: int, loss_scaler,\n                    log_writer=None,\n                    args=None):\n    model.train(True)\n    # model.module.set_default_trainability()\n\n    metric_logger = misc.MetricLogger(delimiter=\"  \")\n    metric_logger.add_meter('lr', misc.SmoothedValue(window_size=1, fmt='{value:.6f}'))","source_hash":"4bd5ffee4e7bc84d0e247e543e4ca5b4f1ea7442cca41dc2e8dd358b211b2687","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/util/misc.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/util/misc.py","kind":"file","name":"imagebind_LLM/util/misc.py","path":"imagebind_LLM/util/misc.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\nimport urllib\nfrom tqdm import tqdm\n\nimport torch","source_hash":"172187ea806ce56512fa2a2afb1c99545a79ce348713b804f2d7948aacdcfbc6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/util/lr_sched.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/util/lr_sched.py","kind":"file","name":"imagebind_LLM/util/lr_sched.py","path":"imagebind_LLM/util/lr_sched.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport math\n\ndef adjust_learning_rate(optimizer, epoch, args):\n    \"\"\"Decay the learning rate with half-cycle cosine after warmup\"\"\"\n    if epoch < args.warmup_epochs:\n        lr = args.lr * epoch / args.warmup_epochs \n    else:\n        lr = args.min_lr + (args.lr - args.min_lr) * 0.5 * \\\n            (1. + math.cos(math.pi * (epoch - args.warmup_epochs) / (args.epochs - args.warmup_epochs)))\n    for param_group in optimizer.param_groups:\n        if \"lr_scale\" in param_group:\n            param_group[\"lr\"] = lr * param_group[\"lr_scale\"]\n        else:\n            param_group[\"lr\"] = lr\n    return lr","source_hash":"4ab5d5633bda0be9173ec91570bb3050326d942582ded2267702b53c3ac87c2c","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/data/dataset.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/data/dataset.py","kind":"file","name":"imagebind_LLM/data/dataset.py","path":"imagebind_LLM/data/dataset.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport yaml\nfrom torch.utils.data import Dataset\nfrom PIL import Image\nimport json\nimport llama.utils\nfrom llama import Tokenizer\nimport copy\nimport torchvision.transforms as transforms\nimport pandas as pd\nimport random\n\ntry:\n    from torchvision.transforms import InterpolationMode\n    BICUBIC = InterpolationMode.BICUBIC\nexcept ImportError:\n    BICUBIC = Image.BICUBIC\n\n\nPROMPT_DICT = {\n    \"prompt_input\": (","source_hash":"21dd21acbfa1466236388e074f1d48472af0435a1e92452970f870bcaa21eae6","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/demo.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/demo.py","kind":"file","name":"imagebind_LLM/ImageBind/demo.py","path":"imagebind_LLM/ImageBind/demo.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import data\nimport torch\nfrom models import imagebind_model\nfrom models.imagebind_model import ModalityType\n\ntext_list=[\"A dog.\", \"A car\", \"A bird\"]\nimage_paths=[\".assets/dog_image.jpg\", \".assets/car_image.jpg\", \".assets/bird_image.jpg\"]\naudio_paths=[\".assets/dog_audio.wav\", \".assets/car_audio.wav\", \".assets/bird_audio.wav\"]\n\ndevice = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n\n# Instantiate model\nmodel = imagebind_model.imagebind_huge(pretrained=True)\nmodel.eval()\nmodel.to(device)\n\n# Load data\ninputs = {\n    ModalityType.TEXT: data.load_and_transform_text(text_list, device),\n    ModalityType.VISION: data.load_and_transform_vision_data(image_paths, device),\n    ModalityType.AUDIO: data.load_and_transform_audio_data(audio_paths, device),","source_hash":"6164353bb6650d6e39c854828cd571d1ec6f55a9eaf83fd62c423394ef2f1375","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/data.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/data.py","kind":"file","name":"imagebind_LLM/ImageBind/data.py","path":"imagebind_LLM/ImageBind/data.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport logging\nimport math\n\nimport torch\nimport torch.nn as nn\nimport torchaudio\nfrom PIL import Image\nimport open3d as o3d\nimport numpy\nfrom pytorchvideo import transforms as pv_transforms\nfrom pytorchvideo.data.clip_sampling import ConstantClipsPerVideoSampler\nfrom pytorchvideo.data.encoded_video import EncodedVideo\nfrom torchvision import transforms\nfrom torchvision.transforms._transforms_video import NormalizeVideo","source_hash":"1380ed40bba37b2e3ca4b22ad2c155f898c576b5ff5d1063fe11eed6d78a130b","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/models/helpers.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/models/helpers.py","kind":"file","name":"imagebind_LLM/ImageBind/models/helpers.py","path":"imagebind_LLM/ImageBind/models/helpers.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\n\nimport einops\nimport numpy as np\nimport torch\nimport torch.nn as nn\n\n\nclass Normalize(nn.Module):\n    def __init__(self, dim: int) -> None:\n        super().__init__()\n        self.dim = dim\n\n    def forward(self, x):\n        return torch.nn.functional.normalize(x, dim=self.dim, p=2)","source_hash":"ee5d71dde0c9c2016272ee461edb7c9f4acb3449df829aa56757963c893a8396","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/models/transformer.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/models/transformer.py","kind":"file","name":"imagebind_LLM/ImageBind/models/transformer.py","path":"imagebind_LLM/ImageBind/models/transformer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\n# Code modified from\n# https://github.com/rwightman/pytorch-image-models/blob/master/timm/models/vision_transformer.py ;\n# https://github.com/facebookresearch/deit/blob/main/models.py\n# and https://github.com/facebookresearch/vissl/blob/main/vissl/models/trunks/vision_transformer.py\n\n\nfrom functools import partial\nfrom typing import Callable, List, Optional\n\nimport torch\nimport torch.nn as nn\nimport torch.utils.checkpoint as checkpoint\nfrom timm.models.layers import DropPath, trunc_normal_\n","source_hash":"b361fe13bf39758fc10d772aa3d948bc1261c31bd34bac07663c336c674b340a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","kind":"file","name":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","path":"imagebind_LLM/ImageBind/models/multimodal_preprocessors.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\nimport gzip\nimport html\nimport io\nimport math\nfrom functools import lru_cache\nfrom typing import Callable, List, Optional, Tuple\n\nimport ftfy\nimport numpy as np\nimport regex as re\nimport torch\nimport torch.nn as nn\nfrom iopath.common.file_io import g_pathmgr\nfrom timm.models.layers import trunc_normal_","source_hash":"fab69ddfe599b8ca471086b24db2c9c171622fa61478a10d73ef1641e4764930","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/models/imagebind_model.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/models/imagebind_model.py","kind":"file","name":"imagebind_LLM/ImageBind/models/imagebind_model.py","path":"imagebind_LLM/ImageBind/models/imagebind_model.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"#!/usr/bin/env python3\n# Portions Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n\n\nimport os\nfrom functools import partial\nfrom types import SimpleNamespace\nfrom collections import OrderedDict\nfrom .pointbert.point_encoder import PointTransformerBind\nfrom .pointbert.checkpoint import get_missing_parameters_message, get_unexpected_parameters_message\n\nimport torch\nimport torch.nn as nn\n\nfrom .helpers import (EinOpsRearrange, LearnableLogitScaling, Normalize,\n                            SelectElement, SelectEOSAndProject)\nfrom .multimodal_preprocessors import (AudioPreprocessor,","source_hash":"288839d634d71027dbe353a790b6419c26d3fbf7f527e82331765a8018863885","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/models/pointbert/misc.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/models/pointbert/misc.py","kind":"file","name":"imagebind_LLM/ImageBind/models/pointbert/misc.py","path":"imagebind_LLM/ImageBind/models/pointbert/misc.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import numpy as np\nimport matplotlib.pyplot as plt\nfrom mpl_toolkits.mplot3d import Axes3D\nimport random\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nimport os\nfrom collections import abc\n\n\n# def fps(data, number):\n#     '''\n#         data B N 3\n#         number int\n#     '''\n#     fps_idx = pointnet2_utils.furthest_point_sample(data, number)\n#     fps_data = pointnet2_utils.gather_operation(data.transpose(1, 2).contiguous(), fps_idx).transpose(1,2).contiguous()\n#     return fps_data\n\ndef index_points(points, idx):","source_hash":"950d72385684f399ac659c950761e793ef232bc1d55403890758489c00745aaa","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","kind":"file","name":"imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","path":"imagebind_LLM/ImageBind/models/pointbert/checkpoint.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"from collections import defaultdict\nimport torch.nn as nn\n\nfrom typing import Any\nfrom typing import Optional, List, Dict, NamedTuple, Tuple, Iterable\n\nfrom termcolor import colored\n\ndef get_missing_parameters_message(keys: List[str]) -> str:\n    \"\"\"\n    Get a logging-friendly message to report parameter names (keys) that are in\n    the model but not found in a checkpoint.\n    Args:\n        keys (list[str]): List of keys that were not found in the checkpoint.\n    Returns:\n        str: message.\n    \"\"\"\n    groups = _group_checkpoint_keys(keys)\n    msg = \"Some model parameters or buffers are not found in the checkpoint:\\n\"\n    msg += \"\\n\".join(\n        \"  \" + colored(k + _group_to_str(v), \"blue\") for k, v in groups.items()","source_hash":"bd95d3c00499acc67b69555c7d9022731b3f53633dda6789380f32e41a084946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","kind":"file","name":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","path":"imagebind_LLM/ImageBind/models/pointbert/point_encoder.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nfrom timm.models.layers import DropPath\nfrom .dvae import Group\nfrom .dvae import Encoder\nfrom .logger import print_log\nimport yaml\nfrom easydict import EasyDict\nfrom .checkpoint import get_missing_parameters_message, get_unexpected_parameters_message\n\ndef merge_new_config(config, new_config):\n    for key, val in new_config.items():\n        if not isinstance(val, dict):\n            if key == '_base_':\n                with open(new_config['_base_'], 'r') as f:\n                    try:\n                        val = yaml.load(f, Loader=yaml.FullLoader)\n                    except:\n                        val = yaml.load(f)\n                config[key] = EasyDict()","source_hash":"96d82d83275cecf97e375b85c15a75d70c6617967408bff1a9f658ad22c99df2","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/models/pointbert/logger.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/models/pointbert/logger.py","kind":"file","name":"imagebind_LLM/ImageBind/models/pointbert/logger.py","path":"imagebind_LLM/ImageBind/models/pointbert/logger.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import logging\nimport torch.distributed as dist\n\nlogger_initialized = {}\n\ndef get_root_logger(log_file=None, log_level=logging.INFO, name='main'):\n    \"\"\"Get root logger and add a keyword filter to it.\n    The logger will be initialized if it has not been initialized. By default a\n    StreamHandler will be added. If `log_file` is specified, a FileHandler will\n    also be added. The name of the root logger is the top-level package name,\n    e.g., \"mmdet3d\".\n    Args:\n        log_file (str, optional): File path of log. Defaults to None.\n        log_level (int, optional): The level of logger.\n            Defaults to logging.INFO.\n        name (str, optional): The name of the root logger, also used as a\n            filter keyword. Defaults to 'mmdet3d'.\n    Returns:\n        :obj:`logging.Logger`: The obtained logger\n    \"\"\"\n    logger = get_logger(name=name, log_file=log_file, log_level=log_level)","source_hash":"d5ce96284512f59bfe386c857c9e3587fbd9afd605d4140f3383bfe7d0a03f69","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/ImageBind/models/pointbert/dvae.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/ImageBind/models/pointbert/dvae.py","kind":"file","name":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","path":"imagebind_LLM/ImageBind/models/pointbert/dvae.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch.nn as nn\nimport torch\nimport torch.nn.functional as F\nfrom . import misc\n\nfrom knn_cuda import KNN\n\nknn = KNN(k=4, transpose_mode=False)\n\n\nclass DGCNN(nn.Module):\n    def __init__(self, encoder_channel, output_channel):\n        super().__init__()\n        '''\n        K has to be 16\n        '''\n        self.input_trans = nn.Conv1d(encoder_channel, 128, 1)\n\n        self.layer1 = nn.Sequential(nn.Conv2d(256, 256, kernel_size=1, bias=False),\n                                    nn.GroupNorm(4, 256),\n                                    nn.LeakyReLU(negative_slope=0.2)","source_hash":"0024f3399c550048c690796284117675ab9b4221a20764db6a96b837cbd7357e","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/llama/llama_adapter.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/llama/llama_adapter.py","kind":"file","name":"imagebind_LLM/llama/llama_adapter.py","path":"imagebind_LLM/llama/llama_adapter.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import json\nimport os\nfrom pathlib import Path\nimport numpy as np\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\n\nfrom .llama import Transformer, ModelArgs, RMSNorm\nfrom .tokenizer import Tokenizer\nfrom util.misc import download\nfrom .utils import sample_top_p\n\nfrom ImageBind.models import imagebind_model\n\n\nclass LLaMA_adapter(nn.Module):\n    \"\"\" Masked Autoencoder with VisionTransformer backbone\n    \"\"\"\n    def __init__(self, llama_ckpt_dir, llama_tokenizer, knn=False, phase=\"finetune\", legacy_bridge=False):","source_hash":"f5f5c793ebdea0858c63dcd6909095493a4eead3cf499c25e64f1597437e2c46","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/llama/llama.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/llama/llama.py","kind":"file","name":"imagebind_LLM/llama/llama.py","path":"imagebind_LLM/llama/llama.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import Optional, Tuple\nfrom dataclasses import dataclass\nimport math\n\nimport torch\nfrom torch import nn\nfrom torch.nn import Embedding, Linear\nimport torch.nn.functional as F\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2\n    norm_eps: float = 1e-5","source_hash":"65a4017f599abdd80d31ae38fa47501180e8ae3076bc15cf72efd09a6b1f2b63","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/llama/__init__.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/llama/__init__.py","kind":"file","name":"imagebind_LLM/llama/__init__.py","path":"imagebind_LLM/llama/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":4,"code":"from .llama import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer\nfrom .llama_adapter import *\nfrom .utils import format_prompt","source_hash":"1f2d46064701de2419f628100b108dfddf2dbe8f21fa4773c2027ada9b3729e4","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/llama/utils.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/llama/utils.py","kind":"file","name":"imagebind_LLM/llama/utils.py","path":"imagebind_LLM/llama/utils.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import torch\n\n\ndef sample_top_p(probs, p):\n    probs_sort, probs_idx = torch.sort(probs, dim=-1, descending=True)\n    probs_sum = torch.cumsum(probs_sort, dim=-1)\n    mask = probs_sum - probs_sort > p\n    probs_sort[mask] = 0.0\n    probs_sort.div_(probs_sort.sum(dim=-1, keepdim=True))\n    next_token = torch.multinomial(probs_sort, num_samples=1)\n    next_token = torch.gather(probs_idx, -1, next_token)\n    return next_token\n\n\ndef format_prompt(instruction, input=None):\n\n    PROMPT_DICT = {\n        \"prompt_input\": (\n            \"Below is an instruction that describes a task, paired with an input that provides further context. \"\n            \"Write a response that appropriately completes the request.\\n\\n\"\n            \"### Instruction:\\n{instruction}\\n\\n### Input:\\n{input}\\n\\n### Response:\"","source_hash":"e4a4149bd7b1c9a320c9eec061ba9666d770f9cf901685563ae4f0486aa245ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/llama/tokenizer.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/llama/tokenizer.py","kind":"file","name":"imagebind_LLM/llama/tokenizer.py","path":"imagebind_LLM/llama/tokenizer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom sentencepiece import SentencePieceProcessor\nfrom logging import getLogger\nfrom typing import List\nimport os\n\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()","source_hash":"7fc25e9bd3dab330649c296f0eec882d33f203205c752818c49b9daddd84a95f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:imagebind_LLM/tools/get_chinese_llama.py","uri":"program://LLaMA-Adapter/file/imagebind_LLM/tools/get_chinese_llama.py","kind":"file","name":"imagebind_LLM/tools/get_chinese_llama.py","path":"imagebind_LLM/tools/get_chinese_llama.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Script for obtaining Chinese LLaMA weights from the OpenChineseLLaMA project (https://github.com/OpenLMLab/OpenChineseLLaMA)\n# Due to the License of LLaMA, we only provide a delta-version patch\n# Adding the patch to the original LLaMA weights makes the Chinese LLaMA weights\nimport os\nimport sys\nsys.path.append(os.path.abspath(__file__).rsplit('/', 2)[0])\nimport shutil\nimport torch\nimport argparse\nfrom util.misc import download\n\nparser = argparse.ArgumentParser()\nparser.add_argument(\n    \"--llama_dir\", default=\"/path/to/llama\", type=str,\n    help=\"Path to official LLaMA weights\",\n)\nargs = parser.parse_args()\n\nori_path = os.path.join(args.llama_dir, '7B')\ndelta_path = os.path.join(args.llama_dir, '7B_chinese_delta')\nnew_path = os.path.join(args.llama_dir, '7B_chinese')","source_hash":"50b5ab89be0b84d50efc629d1b5d2adb5f3cb5dd09f8ec97bd9ea6f3eab5b10f","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_chat65b/conversation.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_chat65b/conversation.py","kind":"file","name":"llama_adapter_v2_chat65b/conversation.py","path":"llama_adapter_v2_chat65b/conversation.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import dataclasses\nfrom enum import auto, Enum\nfrom typing import List, Tuple\n\n\nclass SeparatorStyle(Enum):\n    \"\"\"Different separator style.\"\"\"\n    SINGLE = auto()\n    TWO = auto()\n\n\n@dataclasses.dataclass\nclass Conversation:\n    \"\"\"A class that keeps all conversation history.\"\"\"\n    system: str\n    roles: List[str]\n    messages: List[List[str]]\n    offset: int\n    sep_style: SeparatorStyle = SeparatorStyle.SINGLE\n    sep: str = \"###\"\n    sep2: str = None","source_hash":"b49b07dda1d3c118b5df73393388b8ee70c6639a9b2a3cf7e11ad5abda3bb0ba","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_chat65b/chat_demo.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_chat65b/chat_demo.py","kind":"file","name":"llama_adapter_v2_chat65b/chat_demo.py","path":"llama_adapter_v2_chat65b/chat_demo.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import argparse\nimport os\nimport sys\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport models_llama_adapter\nimport torch\nimport torch.distributed as dist\nfrom conversation import conv_templates, SeparatorStyle\nfrom util import misc\n\nfrom llama import LLaMA, Tokenizer\n\n\ndef load_model(args, load_8bit=False):\n    model = models_llama_adapter.__dict__[args.model_name](args)\n    model.eval()\n    if args.model_path is None:\n        print(\"Warning: not loading instruct tuned weights.\")\n    else:\n        print(\"Using instruct tuned weights from:\", args.model_path)","source_hash":"6220deba345664385d054b284402ca14848a15aa842d58c5ac568bf678052946","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_chat65b/models_llama_adapter.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_chat65b/models_llama_adapter.py","kind":"file","name":"llama_adapter_v2_chat65b/models_llama_adapter.py","path":"llama_adapter_v2_chat65b/models_llama_adapter.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"import functools\nimport json\nfrom pathlib import Path\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.distributed as dist\n\nfrom llama import ModelArgs, Tokenizer, Transformer\n\n\ndef _load_and_redistribute_checkpoint(llama_model_path, model_name):\n    with open(Path(llama_model_path) / model_name / \"params.json\") as f:\n        params = json.load(f)\n    tokenizer = Tokenizer(model_path=str(Path(llama_model_path) / \"tokenizer.model\"))\n    print(\"Using model path: %s, model_name: %s\" % (llama_model_path, model_name))\n\n    checkpoints = (Path(llama_model_path) / model_name).glob(\"*.pth\")\n    checkpoints = sorted(checkpoints)\n\n    mp_world_size = fs_init.get_model_parallel_world_size()","source_hash":"febc64422e500736302f0044a0038e15ccd6a033ca0713a0b0b4e848a839caf8","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_chat65b/util/misc.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_chat65b/util/misc.py","kind":"file","name":"llama_adapter_v2_chat65b/util/misc.py","path":"llama_adapter_v2_chat65b/util/misc.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# All rights reserved.\n\n# This source code is licensed under the license found in the\n# LICENSE file in the root directory of this source tree.\n# --------------------------------------------------------\n# References:\n# DeiT: https://github.com/facebookresearch/deit\n# BEiT: https://github.com/microsoft/unilm/tree/master/beit\n# --------------------------------------------------------\n\nimport builtins\nimport datetime\nimport os\nimport subprocess\nimport time\nfrom collections import defaultdict, deque\nfrom pathlib import Path\n\nimport torch\nimport torch.distributed as dist","source_hash":"2f630ff84d9f226ac697afc5476ae1c7d34832a2ed5f32f7776ec4dfed5242f9","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_chat65b/llama/generation.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_chat65b/llama/generation.py","kind":"file","name":"llama_adapter_v2_chat65b/llama/generation.py","path":"llama_adapter_v2_chat65b/llama/generation.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom typing import List\n\nimport torch\n\nfrom llama.model import Transformer\nfrom llama.tokenizer import Tokenizer\n\n\nclass LLaMA:\n    def __init__(self, model: Transformer, tokenizer: Tokenizer) -> None:\n        self.model = model\n        self.tokenizer = tokenizer\n\n    def generate(\n        self,\n        prompts: List[str],\n        max_gen_len: int,\n        temperature: float = 0.8,","source_hash":"8cbad0b94018145da71f990d707bdf3f9a90a5c9dead3b7406b19e940356a09a","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_chat65b/llama/model.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_chat65b/llama/model.py","kind":"file","name":"llama_adapter_v2_chat65b/llama/model.py","path":"llama_adapter_v2_chat65b/llama/model.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport math\nfrom dataclasses import dataclass\nfrom typing import Optional, Tuple\n\nimport fairscale.nn.model_parallel.initialize as fs_init\nimport torch\nimport torch.nn.functional as F\nfrom fairscale.nn.model_parallel.layers import ColumnParallelLinear, ParallelEmbedding, RowParallelLinear\nfrom torch import nn\n\n\n@dataclass\nclass ModelArgs:\n    dim: int = 512\n    n_layers: int = 8\n    n_heads: int = 8\n    vocab_size: int = -1  # defined later by tokenizer\n    multiple_of: int = 256  # make SwiGLU hidden layer size multiple of large power of 2","source_hash":"be4869009b18d7d97339c648fc884844c98a38921374b236dad3a69a311e4868","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_chat65b/llama/__init__.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_chat65b/llama/__init__.py","kind":"file","name":"llama_adapter_v2_chat65b/llama/__init__.py","path":"llama_adapter_v2_chat65b/llama/__init__.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":6,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nfrom .generation import LLaMA\nfrom .model import ModelArgs, Transformer\nfrom .tokenizer import Tokenizer","source_hash":"1b13d2a1a3d2443f4b3d9bf4c6fe3e08c34a0a0e03cc72ba4fbee16bea46f992","truncated":false}
{"repo_id":"LLaMA-Adapter","entity_id":"file:llama_adapter_v2_chat65b/llama/tokenizer.py","uri":"program://LLaMA-Adapter/file/llama_adapter_v2_chat65b/llama/tokenizer.py","kind":"file","name":"llama_adapter_v2_chat65b/llama/tokenizer.py","path":"llama_adapter_v2_chat65b/llama/tokenizer.py","language":"python","start_line":1,"end_line":1,"context_start_line":1,"context_end_line":21,"code":"# Copyright (c) Meta Platforms, Inc. and affiliates.\n# This software may be used and distributed according to the terms of the GNU General Public License version 3.\n\nimport os\nfrom logging import getLogger\nfrom typing import List\n\nfrom sentencepiece import SentencePieceProcessor\n\nlogger = getLogger()\n\n\nclass Tokenizer:\n    def __init__(self, model_path: str):\n        # reload tokenizer\n        assert os.path.isfile(model_path), model_path\n        self.sp_model = SentencePieceProcessor(model_file=model_path)\n        logger.info(f\"Reloaded SentencePiece model from {model_path}\")\n\n        # BOS / EOS token IDs\n        self.n_words: int = self.sp_model.vocab_size()","source_hash":"8bd26934e9a4b3ef78aeeaa959cb34bfd2dc8737d983320dc14ea9940d346aae","truncated":false}