diff --git a/3b_only_dpo/3b_only_dpo_ckpt_100.log b/3b_only_dpo/3b_only_dpo_ckpt_100.log new file mode 100644 index 0000000000000000000000000000000000000000..126d3ba2243784b053c51dc9fa9676eec595ce2b --- /dev/null +++ b/3b_only_dpo/3b_only_dpo_ckpt_100.log @@ -0,0 +1,25 @@ +2025-02-08:22:34:04,653 INFO [__main__.py:284] Verbosity set to INFO +2025-02-08:22:34:16,782 INFO [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama'] +2025-02-08:22:34:16,783 INFO [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234 +2025-02-08:22:34:16,784 WARNING [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding! +2025-02-08:22:34:16,784 INFO [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Llama-3.2-3B-Instruct', 'peft': 'JaqueMate/new_config_3b_only_dpo_checkpoint-100'} +2025-02-08:22:34:16,825 INFO [huggingface.py:135] Using device 'cuda:0' +2025-02-08:22:34:17,519 INFO [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'} + Loading checkpoint shards: 0%| | 0/2 [00:00 + cli_evaluate() + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 387, in cli_evaluate + results = evaluator.simple_evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 205, in simple_evaluate + lm = lm_eval.api.registry.get_model(model).create_from_arg_string( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/api/model.py", line 151, in create_from_arg_string + return cls(**args, **args2) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 190, in __init__ + self._create_model( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 647, in _create_model + self._model = PeftModel.from_pretrained( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 554, in from_pretrained + model.load_adapter( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 1122, in load_adapter + adapters_weights = load_peft_weights(model_id, device=torch_device, **hf_hub_download_kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/utils/save_and_load.py", line 507, in load_peft_weights + raise ValueError( +ValueError: Can't find weights for JaqueMate/new_config_3b_only_dpo_checkpoint-1100 in JaqueMate/new_config_3b_only_dpo_checkpoint-1100 or in the Hugging Face Hub. Please check that the file adapter_model.bin or adapter_model.safetensors is present at JaqueMate/new_config_3b_only_dpo_checkpoint-1100. diff --git a/3b_only_dpo/3b_only_dpo_ckpt_1200.log b/3b_only_dpo/3b_only_dpo_ckpt_1200.log new file mode 100644 index 0000000000000000000000000000000000000000..25bd8b59adbe0cbae442d095b4ec49017f00bd9c --- /dev/null +++ b/3b_only_dpo/3b_only_dpo_ckpt_1200.log @@ -0,0 +1,25 @@ +2025-02-09:11:17:21,116 INFO [__main__.py:284] Verbosity set to INFO +2025-02-09:11:17:32,402 INFO [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama'] +2025-02-09:11:17:32,403 INFO [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234 +2025-02-09:11:17:32,403 WARNING [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding! +2025-02-09:11:17:32,403 INFO [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Llama-3.2-3B-Instruct', 'peft': 'JaqueMate/new_config_3b_only_dpo_checkpoint-1200'} +2025-02-09:11:17:32,441 INFO [huggingface.py:135] Using device 'cuda:0' +2025-02-09:11:17:33,093 INFO [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'} + Loading checkpoint shards: 0%| | 0/2 [00:00 + cli_evaluate() + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 387, in cli_evaluate + results = evaluator.simple_evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 205, in simple_evaluate + lm = lm_eval.api.registry.get_model(model).create_from_arg_string( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/api/model.py", line 151, in create_from_arg_string + return cls(**args, **args2) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 190, in __init__ + self._create_model( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 647, in _create_model + self._model = PeftModel.from_pretrained( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 554, in from_pretrained + model.load_adapter( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 1122, in load_adapter + adapters_weights = load_peft_weights(model_id, device=torch_device, **hf_hub_download_kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/utils/save_and_load.py", line 507, in load_peft_weights + raise ValueError( +ValueError: Can't find weights for JaqueMate/new_config_3b_only_dpo_checkpoint-600 in JaqueMate/new_config_3b_only_dpo_checkpoint-600 or in the Hugging Face Hub. Please check that the file adapter_model.bin or adapter_model.safetensors is present at JaqueMate/new_config_3b_only_dpo_checkpoint-600. diff --git a/3b_only_dpo/3b_only_dpo_ckpt_700.log b/3b_only_dpo/3b_only_dpo_ckpt_700.log new file mode 100644 index 0000000000000000000000000000000000000000..024946a0e6d65a397a729f1f6892e2092e5cdc60 --- /dev/null +++ b/3b_only_dpo/3b_only_dpo_ckpt_700.log @@ -0,0 +1,25 @@ +2025-02-09:00:30:10,322 INFO [__main__.py:284] Verbosity set to INFO +2025-02-09:00:30:21,749 INFO [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama'] +2025-02-09:00:30:21,750 INFO [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234 +2025-02-09:00:30:21,750 WARNING [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding! +2025-02-09:00:30:21,750 INFO [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Llama-3.2-3B-Instruct', 'peft': 'JaqueMate/new_config_3b_only_dpo_checkpoint-700'} +2025-02-09:00:30:21,790 INFO [huggingface.py:135] Using device 'cuda:0' +2025-02-09:00:30:22,384 INFO [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'} + Loading checkpoint shards: 0%| | 0/2 [00:00 + cli_evaluate() + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 387, in cli_evaluate + results = evaluator.simple_evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 205, in simple_evaluate + lm = lm_eval.api.registry.get_model(model).create_from_arg_string( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/api/model.py", line 151, in create_from_arg_string + return cls(**args, **args2) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 190, in __init__ + self._create_model( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 647, in _create_model + self._model = PeftModel.from_pretrained( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 554, in from_pretrained + model.load_adapter( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 1122, in load_adapter + adapters_weights = load_peft_weights(model_id, device=torch_device, **hf_hub_download_kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/utils/save_and_load.py", line 507, in load_peft_weights + raise ValueError( +ValueError: Can't find weights for JaqueMate/shuffle_3b_checkpoint-400 in JaqueMate/shuffle_3b_checkpoint-400 or in the Hugging Face Hub. Please check that the file adapter_model.bin or adapter_model.safetensors is present at JaqueMate/shuffle_3b_checkpoint-400. diff --git a/3b_shuffle/solo_dpo_model_ckpt_ckpt_500.log b/3b_shuffle/solo_dpo_model_ckpt_ckpt_500.log new file mode 100644 index 0000000000000000000000000000000000000000..da8a1ac3f20f274dd2c9cc497c5a9704eb675957 --- /dev/null +++ b/3b_shuffle/solo_dpo_model_ckpt_ckpt_500.log @@ -0,0 +1,25 @@ +2025-02-12:16:00:07,205 INFO [__main__.py:284] Verbosity set to INFO +2025-02-12:16:00:18,326 INFO [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama'] +2025-02-12:16:00:18,327 INFO [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234 +2025-02-12:16:00:18,328 WARNING [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding! +2025-02-12:16:00:18,328 INFO [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Llama-3.2-3B-Instruct', 'peft': 'JaqueMate/shuffle_3b_checkpoint-500'} +2025-02-12:16:00:18,367 INFO [huggingface.py:135] Using device 'cuda:0' +2025-02-12:16:00:18,959 INFO [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'} + Loading checkpoint shards: 0%| | 0/2 [00:00 + cli_evaluate() + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 387, in cli_evaluate + results = evaluator.simple_evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 205, in simple_evaluate + lm = lm_eval.api.registry.get_model(model).create_from_arg_string( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/api/model.py", line 151, in create_from_arg_string + return cls(**args, **args2) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 190, in __init__ + self._create_model( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 586, in _create_model + self._model = self.AUTO_MODEL_CLASS.from_pretrained( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/auto/auto_factory.py", line 564, in from_pretrained + return model_class.from_pretrained( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/modeling_utils.py", line 3961, in from_pretrained + ) = cls._load_pretrained_model( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/modeling_utils.py", line 4435, in _load_pretrained_model + new_error_msgs, offload_index, state_dict_index = _load_state_dict_into_meta_model( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/modeling_utils.py", line 962, in _load_state_dict_into_meta_model + set_module_tensor_to_device(model, param_name, param_device, **set_module_kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/accelerate/utils/modeling.py", line 416, in set_module_tensor_to_device + new_value = value.to(device) +KeyboardInterrupt diff --git a/8b_only_dpo/solo_dpo_model_ckpt_ckpt_1000.log b/8b_only_dpo/solo_dpo_model_ckpt_ckpt_1000.log new file mode 100644 index 0000000000000000000000000000000000000000..933ff1299dec10107c10fe1833508ee1200e340f --- /dev/null +++ b/8b_only_dpo/solo_dpo_model_ckpt_ckpt_1000.log @@ -0,0 +1,25 @@ +2025-02-09:22:41:45,858 INFO [__main__.py:284] Verbosity set to INFO +2025-02-09:22:41:56,962 INFO [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama'] +2025-02-09:22:41:56,963 INFO [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234 +2025-02-09:22:41:56,963 WARNING [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding! +2025-02-09:22:41:56,963 INFO [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/new_config_checkpoint-1000'} +2025-02-09:22:41:57,002 INFO [huggingface.py:135] Using device 'cuda:0' +2025-02-09:22:41:57,612 INFO [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'} + Loading checkpoint shards: 0%| | 0/4 [00:00 + cli_evaluate() + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 387, in cli_evaluate + results = evaluator.simple_evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 304, in simple_evaluate + results = evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 524, in evaluate + resps = getattr(lm, reqtype)(cloned_reqs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 1354, in generate_until + cont = self._model_generate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 897, in _model_generate + return self.model.generate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 1646, in generate + outputs = self.base_model.generate(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context + return func(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/utils.py", line 1996, in generate + result = self._sample( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/utils.py", line 2923, in _sample + outputs = self(**model_inputs, return_dict=True) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 1182, in forward + outputs = self.model( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 994, in forward + layer_outputs = decoder_layer( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 727, in forward + hidden_states, self_attn_weights, present_key_value = self.self_attn( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 612, in forward + value_states = self.v_proj(hidden_states) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/tuners/lora/layer.py", line 556, in forward + result = result + lora_B(lora_A(dropout(x))) * scaling + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 116, in forward + return F.linear(input, self.weight, self.bias) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1696, in __getattr__ + def __getattr__(self, name: str) -> Any: +KeyboardInterrupt + Running generate_until requests: 52%|█████▏ | 690/1319 [28:29<25:58, 2.48s/it] diff --git a/8b_shuffle/solo_dpo_model_ckpt_ckpt_1700.log b/8b_shuffle/solo_dpo_model_ckpt_ckpt_1700.log new file mode 100644 index 0000000000000000000000000000000000000000..85943239d6d05724512ab5dfe78e7e5d84995edd --- /dev/null +++ b/8b_shuffle/solo_dpo_model_ckpt_ckpt_1700.log @@ -0,0 +1,56 @@ +2025-02-11:10:41:59,764 INFO [__main__.py:284] Verbosity set to INFO +2025-02-11:10:42:10,657 INFO [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama'] +2025-02-11:10:42:10,658 INFO [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234 +2025-02-11:10:42:10,658 WARNING [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding! +2025-02-11:10:42:10,658 INFO [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/shuffle_checkpoint-1700'} +2025-02-11:10:42:10,696 INFO [huggingface.py:135] Using device 'cuda:0' +2025-02-11:10:42:11,280 INFO [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'} + Loading checkpoint shards: 0%| | 0/4 [00:00 + cli_evaluate() + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 387, in cli_evaluate + results = evaluator.simple_evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 304, in simple_evaluate + results = evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 524, in evaluate + resps = getattr(lm, reqtype)(cloned_reqs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 1354, in generate_until + cont = self._model_generate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 897, in _model_generate + return self.model.generate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 1646, in generate + outputs = self.base_model.generate(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context + return func(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/utils.py", line 1996, in generate + result = self._sample( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/utils.py", line 2923, in _sample + outputs = self(**model_inputs, return_dict=True) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 1201, in forward + logits = self.lm_head(hidden_states) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 116, in forward + return F.linear(input, self.weight, self.bias) +torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 1.71 GiB. GPU + Running generate_until requests: 0%| | 0/1319 [00:04 + cli_evaluate() + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 387, in cli_evaluate + results = evaluator.simple_evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 304, in simple_evaluate + results = evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 524, in evaluate + resps = getattr(lm, reqtype)(cloned_reqs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 1354, in generate_until + cont = self._model_generate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 897, in _model_generate + return self.model.generate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 1646, in generate + outputs = self.base_model.generate(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context + return func(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/utils.py", line 1996, in generate + result = self._sample( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/generation/utils.py", line 2923, in _sample + outputs = self(**model_inputs, return_dict=True) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/transformers/models/llama/modeling_llama.py", line 1201, in forward + logits = self.lm_head(hidden_states) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl + return self._call_impl(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl + return forward_call(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/torch/nn/modules/linear.py", line 116, in forward + return F.linear(input, self.weight, self.bias) +torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 1.71 GiB. GPU + Running generate_until requests: 0%| | 0/1319 [00:04 + cli_evaluate() + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/__main__.py", line 387, in cli_evaluate + results = evaluator.simple_evaluate( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/utils.py", line 402, in _wrapper + return fn(*args, **kwargs) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/evaluator.py", line 205, in simple_evaluate + lm = lm_eval.api.registry.get_model(model).create_from_arg_string( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/api/model.py", line 151, in create_from_arg_string + return cls(**args, **args2) + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 190, in __init__ + self._create_model( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/lm-evaluation-harness/lm_eval/models/huggingface.py", line 647, in _create_model + self._model = PeftModel.from_pretrained( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/peft_model.py", line 462, in from_pretrained + PeftConfig._get_peft_type( + File "/home/luciano/Documents/Tesis Ezequiel/Tesis/env/lib/python3.10/site-packages/peft/config.py", line 211, in _get_peft_type + raise ValueError(f"Can't find '{CONFIG_NAME}' at '{model_id}'") +ValueError: Can't find 'adapter_config.json' at 'JaqueMate/shuffle_checkpoint-1900' diff --git a/8b_shuffle/solo_dpo_model_ckpt_ckpt_200.log b/8b_shuffle/solo_dpo_model_ckpt_ckpt_200.log new file mode 100644 index 0000000000000000000000000000000000000000..cdec377478eca0d41e0d83064813c9dd798c487f --- /dev/null +++ b/8b_shuffle/solo_dpo_model_ckpt_ckpt_200.log @@ -0,0 +1,25 @@ +2025-02-11:12:24:00,850 INFO [__main__.py:284] Verbosity set to INFO +2025-02-11:12:24:12,047 INFO [__main__.py:381] Selected Tasks: ['gsm8k_cot_llama'] +2025-02-11:12:24:12,048 INFO [evaluator.py:165] Setting random seed to 0 | Setting numpy seed to 1234 | Setting torch manual seed to 1234 | Setting fewshot manual seed to 1234 +2025-02-11:12:24:12,048 WARNING [evaluator.py:176] generation_kwargs specified through cli, these settings will update set parameters in yaml tasks. Ensure 'do_sample=True' for non-greedy decoding! +2025-02-11:12:24:12,048 INFO [evaluator.py:202] Initializing hf model, with arguments: {'pretrained': 'meta-llama/Meta-Llama-3-8B-Instruct', 'peft': 'JaqueMate/shuffle_checkpoint-200'} +2025-02-11:12:24:12,086 INFO [huggingface.py:135] Using device 'cuda:0' +2025-02-11:12:24:12,707 INFO [huggingface.py:375] Model parallel was set to False, max memory was not set, and device map was set to {'': 'cuda:0'} + Loading checkpoint shards: 0%| | 0/4 [00:00