Upload ElysiumForCausalLM

README.md

@@ -0,0 +1,201 @@
+---
+library_name: transformers
+tags: []
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure 
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+
+

build_vit.py

@@ -0,0 +1,92 @@
+import re
+
+from transformers import CLIPVisionModel
+import torch
+import torch.nn as nn
+
+
+class ClipVisionTransformer(CLIPVisionModel):
+    def forward(
+        self,
+        pixel_values=None,
+        output_attentions=None,
+        output_hidden_states=True,
+        return_dict=None,
+    ):
+        r"""
+        Returns:
+
+        Examples:
+
+        ```python
+        >>> from PIL import Image
+        >>> import requests
+        >>> from transformers import AutoProcessor, CLIPVisionModel
+
+        >>> model = CLIPVisionModel.from_pretrained("openai/clip-vit-base-patch32")
+        >>> processor = AutoProcessor.from_pretrained("openai/clip-vit-base-patch32")
+
+        >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
+        >>> image = Image.open(requests.get(url, stream=True).raw)
+
+        >>> inputs = processor(images=image, return_tensors="pt")
+
+        >>> outputs = model(**inputs)
+        >>> last_hidden_state = outputs.last_hidden_state
+        >>> pooled_output = outputs.pooler_output  # pooled CLS states
+        ```"""
+        return_dict = (
+            return_dict if return_dict is not None else self.config.use_return_dict
+        )
+
+        image_forward_outs = self.vision_model(
+            pixel_values=pixel_values,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        return image_forward_outs.hidden_states[-2][:, 1:]  # Use second to last layer as in LLaVA
+
+
+def create_clip_vit(
+    precision="fp16", pretrained_model_name_or_path: str = "", low_cpu_mem_usage=False, **kwargs
+):
+    dtype = torch.float16 if "16" in precision else torch.float32
+    model = ClipVisionTransformer.from_pretrained(
+        pretrained_model_name_or_path,
+        torch_dtype=dtype,
+        low_cpu_mem_usage=low_cpu_mem_usage,
+        ignore_mismatched_sizes=True,
+    ).cuda()
+    return model
+
+
+def build_projector(
+    type: str = "linear", input_hidden_size: int = 1024, output_hidden_size: int = 1024
+):
+    """build vision projector
+    Args:
+        type: projector type (linear, mlp2x_gelu, identity)
+        input_hidden_size: input hidden size from adaptor
+        output_hidden_size: output hidden size to llm
+    Returns:
+        vision projector module(nn.Module)
+    """
+
+    if type == "linear":
+        return nn.Linear(input_hidden_size, output_hidden_size)
+
+    mlp_gelu_match = re.match(r"^mlp(\d+)x_gelu$", type)
+    if mlp_gelu_match:
+        mlp_depth = int(mlp_gelu_match.group(1))
+        modules = [nn.Linear(input_hidden_size, output_hidden_size)]
+        for _ in range(1, mlp_depth):
+            modules.append(nn.GELU())
+            modules.append(nn.Linear(output_hidden_size, output_hidden_size))
+        return nn.Sequential(*modules)
+
+    if type == "identity":
+        return nn.Identity()
+
+    raise ValueError(f"Unknown projector type: {type}")

config.json

@@ -0,0 +1,36 @@
+{
+  "_name_or_path": "elysium_7b",
+  "adapter_config": {
+    "fix_random": false,
+    "freeze_adapter": false,
+    "in_token_num": 576,
+    "max_video_tokens": 2048,
+    "min_out_token_num": 1,
+    "num_features": 1024,
+    "out_token_num": 108
+  },
+  "architectures": [
+    "ElysiumForCausalLM"
+  ],
+  "auto_map": {
+    "AutoConfig": "modeling_elysium.ElysiumConfig",
+    "AutoModelForCausalLM": "modeling_elysium.ElysiumForCausalLM"
+  },
+  "gradient_checkpointing_enable": false,
+  "hidden_size": 4096,
+  "llm_config": {
+    "freeze_llm": true,
+    "pretrained_model_name_or_path": "lmsys/vicuna-7b-v1.5"
+  },
+  "model_type": "elysium",
+  "projector_config": {
+    "type": "mlp2x_gelu"
+  },
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.37.2",
+  "use_flash_attention": false,
+  "visual_config": {
+    "freeze_vit": true,
+    "pretrained_model_name_or_path": "openai/clip-vit-large-patch14-336"
+  }
+}

generation_config.json

@@ -0,0 +1,4 @@
+{
+  "_from_model_config": true,
+  "transformers_version": "4.37.2"
+}

llama_flash_attn_monkey_patch.py

@@ -0,0 +1,115 @@
+from typing import Optional, Tuple
+import warnings
+
+import torch
+
+import transformers
+from transformers.models.llama.modeling_llama import apply_rotary_pos_emb, repeat_kv
+
+try:
+    from flash_attn.flash_attn_interface import flash_attn_unpadded_qkvpacked_func
+except ImportError:
+    from flash_attn.flash_attn_interface import flash_attn_varlen_qkvpacked_func as flash_attn_unpadded_qkvpacked_func
+from flash_attn.bert_padding import unpad_input, pad_input
+
+
+def forward(
+    self,
+    hidden_states: torch.Tensor,
+    attention_mask: Optional[torch.Tensor] = None,
+    position_ids: Optional[torch.Tensor] = None,
+    past_key_value: Optional[Tuple[torch.Tensor]] = None,
+    output_attentions: bool = False,
+    use_cache: bool = False,
+) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+    if output_attentions:
+        warnings.warn(
+            "Output attentions is not supported for patched `LlamaAttention`, returning `None` instead."
+        )
+
+    bsz, q_len, _ = hidden_states.size()
+
+    query_states = (
+        self.q_proj(hidden_states)
+        .view(bsz, q_len, self.num_heads, self.head_dim)
+        .transpose(1, 2)
+    )
+    key_states = (
+        self.k_proj(hidden_states)
+        .view(bsz, q_len, self.num_key_value_heads, self.head_dim)
+        .transpose(1, 2)
+    )
+    value_states = (
+        self.v_proj(hidden_states)
+        .view(bsz, q_len, self.num_key_value_heads, self.head_dim)
+        .transpose(1, 2)
+    )  # shape: (b, num_heads, s, head_dim)
+
+    kv_seq_len = key_states.shape[-2]
+    if past_key_value is not None:
+        kv_seq_len += past_key_value[0].shape[-2]
+
+    cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
+    query_states, key_states = apply_rotary_pos_emb(
+        query_states, key_states, cos, sin, position_ids
+    )
+
+    if past_key_value is not None:
+        # reuse k, v
+        key_states = torch.cat([past_key_value[0], key_states], dim=2)
+        value_states = torch.cat([past_key_value[1], value_states], dim=2)
+
+    past_key_value = (key_states, value_states) if use_cache else None
+
+    # repeat k/v heads if n_kv_heads < n_heads
+    key_states = repeat_kv(key_states, self.num_key_value_groups)
+    value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+    # Transform the data into the format required by flash attention
+    qkv = torch.stack([query_states, key_states, value_states], dim=2)
+    qkv = qkv.transpose(1, 3)  # shape: [b, s, 3, num_heads, head_dim]
+    key_padding_mask = attention_mask
+
+    if key_padding_mask is None:
+        qkv = qkv.reshape(-1, 3, self.num_heads, self.head_dim)
+        cu_q_lens = torch.arange(
+            0, (bsz + 1) * q_len, step=q_len, dtype=torch.int32, device=qkv.device
+        )
+        max_s = q_len
+        output = flash_attn_unpadded_qkvpacked_func(
+            qkv, cu_q_lens, max_s, 0.0, softmax_scale=None, causal=True
+        )
+        output = output.view(bsz, q_len, -1)
+    else:
+        qkv = qkv.reshape(bsz, q_len, -1)
+        qkv, indices, cu_q_lens, max_s = unpad_input(qkv, key_padding_mask)
+        qkv = qkv.view(-1, 3, self.num_heads, self.head_dim)
+        output_unpad = flash_attn_unpadded_qkvpacked_func(
+            qkv, cu_q_lens, max_s, 0.0, softmax_scale=None, causal=True
+        )
+        output_unpad = output_unpad.reshape(-1, self.num_heads * self.head_dim)
+        output = pad_input(output_unpad, indices, bsz, q_len)
+
+    return self.o_proj(output), None, past_key_value
+
+
+# Disable the transformation of the attention mask in LlamaModel as the flash attention
+# requires the attention mask to be the same as the key_padding_mask
+def _prepare_decoder_attention_mask(
+    self, attention_mask, input_shape, inputs_embeds, past_key_values_length
+):
+    # [bsz, seq_len]
+    return attention_mask
+
+
+def replace_llama_attn_with_flash_attn():
+    cuda_major, cuda_minor = torch.cuda.get_device_capability()
+    if cuda_major < 8:
+        warnings.warn(
+            "Flash attention is only supported on A100 or H100 GPU during training due to head dim > 64 backward."
+            "ref: https://github.com/HazyResearch/flash-attention/issues/190#issuecomment-1523359593"
+        )
+    transformers.models.llama.modeling_llama.LlamaModel._prepare_decoder_attention_mask = (
+        _prepare_decoder_attention_mask
+    )
+    transformers.models.llama.modeling_llama.LlamaAttention.forward = forward

model-00001-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52d47491d2bb58837a52947303c7d42e590f5198b11d125b1a6cf646d19791af
+size 4938985768

model-00002-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8468ab5b6dad76e193aaac8cfe0b1bdb6dddd6f8dfcc21bb068ee5c15bf87c1b
+size 4947391312

model-00003-of-00003.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b4386832fc54a0a04460f8a1895c1a186e73a77777fa2744098fada5594e203
+size 4240310466

model.safetensors.index.json

@@ -0,0 +1,699 @@
+{
+  "metadata": {
+    "total_size": 14126596954
+  },
+  "weight_map": {
+    "adapter.pre_proj.bias": "model-00003-of-00003.safetensors",
+    "adapter.pre_proj.weight": "model-00003-of-00003.safetensors",
+    "adapter.score_proj.bias": "model-00003-of-00003.safetensors",
+    "adapter.score_proj.weight": "model-00003-of-00003.safetensors",
+    "adapter.trans_bias": "model-00003-of-00003.safetensors",
+    "adapter.trans_weight": "model-00003-of-00003.safetensors",
+    "llm.lm_head.weight": "model-00003-of-00003.safetensors",
+    "llm.model.embed_tokens.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.0.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.0.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.0.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.1.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.1.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.1.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.10.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.10.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.10.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.11.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.11.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.11.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.11.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.12.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.12.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.12.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.12.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.12.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.12.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.12.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.12.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.12.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.13.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.13.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.13.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.13.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.13.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.13.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.13.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.13.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.13.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.14.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.14.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.14.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.14.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.14.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.14.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.14.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.14.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.14.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.15.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.15.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.15.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.15.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.15.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.15.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.15.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.15.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.15.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.16.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.16.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.16.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.16.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.16.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.16.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.16.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.16.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.16.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.17.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.17.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.17.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.17.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.17.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.17.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.17.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.17.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.17.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.18.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.18.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.18.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.18.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.18.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.18.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.18.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.18.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.18.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.19.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.19.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.19.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.19.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.19.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.19.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.19.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.19.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.19.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.2.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.2.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.2.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.20.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.20.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.20.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.20.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.20.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.20.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.20.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.20.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.20.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.21.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.21.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.21.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.21.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.21.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.21.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.21.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.21.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.21.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.22.input_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.22.mlp.down_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.22.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.22.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.22.post_attention_layernorm.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.22.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.22.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.22.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.22.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.23.input_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.23.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.23.mlp.gate_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.23.mlp.up_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.23.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.23.self_attn.k_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.23.self_attn.o_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.23.self_attn.q_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.23.self_attn.v_proj.weight": "model-00002-of-00003.safetensors",
+    "llm.model.layers.24.input_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.24.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.24.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.24.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.24.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.24.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.24.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.24.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.24.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.25.input_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.25.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.25.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.25.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.25.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.25.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.25.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.25.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.25.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.26.input_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.26.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.26.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.26.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.26.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.26.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.26.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.26.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.26.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.27.input_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.27.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.27.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.27.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.27.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.27.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.27.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.27.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.27.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.28.input_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.28.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.28.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.28.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.28.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.28.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.28.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.28.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.28.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.29.input_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.29.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.29.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.29.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.29.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.29.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.29.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.29.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.29.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.3.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.3.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.3.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.30.input_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.30.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.30.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.30.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.30.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.30.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.30.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.30.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.30.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.31.input_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.31.mlp.down_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.31.mlp.gate_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.31.mlp.up_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.31.post_attention_layernorm.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.31.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.31.self_attn.o_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.31.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.31.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "llm.model.layers.4.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.4.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.4.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.5.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.5.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.5.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.6.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.6.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.6.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.7.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.7.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.7.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.8.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.8.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.8.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.9.input_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.9.mlp.down_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.9.mlp.up_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00003.safetensors",
+    "llm.model.norm.weight": "model-00003-of-00003.safetensors",
+    "llm_proj.0.bias": "model-00003-of-00003.safetensors",
+    "llm_proj.0.weight": "model-00003-of-00003.safetensors",
+    "llm_proj.2.bias": "model-00003-of-00003.safetensors",
+    "llm_proj.2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.embeddings.class_embedding": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.embeddings.patch_embedding.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.embeddings.position_embedding.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.layer_norm1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.layer_norm1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.layer_norm2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.layer_norm2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.post_layernorm.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.post_layernorm.weight": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.pre_layrnorm.bias": "model-00003-of-00003.safetensors",
+    "visual_encoder.vision_model.pre_layrnorm.weight": "model-00003-of-00003.safetensors"
+  }
+}

modeling_elysium.py

@@ -0,0 +1,296 @@
+from typing import Dict, List, Optional, Tuple, Union
+
+import torch
+from transformers import (
+    AutoTokenizer,
+    AutoModelForCausalLM,
+    PretrainedConfig,
+    PreTrainedModel
+)
+from transformers.utils import logging
+from transformers.modeling_outputs import CausalLMOutputWithPast
+
+from .build_vit import build_projector, create_clip_vit
+from .rantselector import build_adapter
+
+logging.set_verbosity_info()  # Turn on this for debug mode
+logger = logging.get_logger(__name__)
+
+
+DTYPE_MAPPING = {
+    "float16": torch.float16,
+    "fp16": torch.float16,
+    "bf16": torch.bfloat16,
+    "bfloat16": torch.bfloat16,
+    "float": torch.float32,
+    "fp32": torch.float32,
+}
+
+# image level
+IMAGE_TOKEN_INDEX = -200
+DEFAULT_IMAGE_TOKEN = "<image>"
+
+# video level
+VIDEO_TOKEN_INDEX = -201
+DEFAULT_VIDEO_TOKEN = "<video>"
+
+
+class ElysiumConfig(PretrainedConfig):
+    model_type = "elysium"
+    _auto_class = "AutoConfig"
+
+    def __init__(
+        self,
+        use_flash_attention: bool = False,
+        gradient_checkpointing_enable: bool = False,
+        torch_dtype: str = "bf16",
+        llm_config: Dict = None,
+        visual_config: Dict = None,
+        adapter_config: Dict = None,
+        projector_config: Dict = None,
+        **kwargs,
+    ):
+        self.use_flash_attention = use_flash_attention
+        self.gradient_checkpointing_enable = gradient_checkpointing_enable
+        self.torch_dtype = torch_dtype
+        self.llm_config = llm_config
+        self.visual_config = visual_config
+        self.adapter_config = adapter_config
+        self.projector_config = projector_config
+
+        super().__init__(**kwargs)
+
+
+class ElysiumForCausalLM(PreTrainedModel):
+    _auto_class = 'AutoModelForCausalLM'
+    supports_gradient_checkpointing = True
+
+    def __init__(self, config: PretrainedConfig = ElysiumConfig()):
+        super().__init__(config)
+
+        # setup llm
+        self.flash_attn_monkey_patch()
+        self._setup_llm()
+
+        self._setup_visual_encoder()
+        self._setup_adapter()
+        self._setup_projector()
+
+        if self.config.torch_dtype:
+            logger.info(f"Converting model to {DTYPE_MAPPING[self.torch_dtype]}.")
+            self.to(DTYPE_MAPPING[self.torch_dtype])
+            logger.info("Done.")
+
+        if self.config.gradient_checkpointing_enable:
+            self._enable_gradient_checkpointing()
+
+    def flash_attn_monkey_patch(self):
+        if self.config.use_flash_attention:
+            # use flash attention
+            from .llama_flash_attn_monkey_patch import (
+                replace_llama_attn_with_flash_attn,
+            )
+            logger.info("Flash attention is availiable, patching.")
+            replace_llama_attn_with_flash_attn()
+
+    def _setup_visual_encoder(self):
+        self.visual_encoder = create_clip_vit(**self.config.visual_config)
+        if self.config.visual_config["freeze_vit"]:
+            for _, param in self.visual_encoder.named_parameters():
+                param.requires_grad = False
+            self.visual_encoder = self.visual_encoder.eval()
+            self.visual_encoder.train = lambda self, mode=True: self
+            logger.info("freeze vision encoder")
+
+    def _setup_llm(self):
+        # text encoder & load pretrained model
+        self.tokenizer = AutoTokenizer.from_pretrained(self.config.llm_config["pretrained_model_name_or_path"])
+        self.llm = AutoModelForCausalLM.from_pretrained(self.config.llm_config["pretrained_model_name_or_path"])
+
+        # freeze llm if needed
+        if self.config.llm_config["freeze_llm"]:
+            for _, param in self.llm.named_parameters():
+                param.requires_grad = False
+            logger.info("freeze llm")
+        
+        if hasattr(self.llm.config, "hidden_size"):
+            self.config.hidden_size = self.llm.config.hidden_size
+        if hasattr(self.llm.config, "hidden_sizes"):
+            self.config.hidden_sizes = self.llm.config.hidden_sizes
+
+    def _setup_adapter(self):
+        self.adapter = build_adapter(self.config.adapter_config)
+        if self.config.adapter_config["freeze_adapter"]:
+            for _, param in self.adapter.named_parameters():
+                param.requires_grad = False
+            self.adapter = self.adapter.eval()
+            self.adapter.train = lambda self, mode=True: self
+            logger.info("freeze adapter")
+
+    def _setup_projector(self):
+        self.llm_proj = build_projector(
+            output_hidden_size=self.llm.config.hidden_size,
+            input_hidden_size=self.adapter.hidden_size,
+            **self.config.projector_config,
+        )
+
+    def _encode_vision(self, images, n_frames):
+        image_embeds = self.visual_encoder(images)
+        adapter_out = self.adapter(image_embeds, n_frames=n_frames)
+        vision_embeds = [self.llm_proj(feature) for feature in adapter_out]
+
+        attention_mask = [
+            torch.ones(feature.size()[:-1], dtype=torch.long).to(feature.device) for feature in vision_embeds]
+        vision_targets = [
+            torch.ones(feature.size(), dtype=torch.long).to(feature.device).fill_(-100) for feature in attention_mask]
+        return vision_embeds, attention_mask, vision_targets
+
+    def _concat_embedding(self, vision_encode_out, input_ids, attention_mask, labels=None, left_padding=False):
+        """ concat vision and text
+        """
+        vision_embeds, vision_atts, vision_targets = vision_encode_out
+
+        input_embeds = []
+        attention_masks = []
+        targets = []
+
+        for cur_batch_idx, cur_input_ids in enumerate(input_ids):
+            cur_vision_embeds = vision_embeds[cur_batch_idx]
+            cur_vision_attn = vision_atts[cur_batch_idx]
+            cur_vision_targets = vision_targets[cur_batch_idx]
+            cur_attn_masks = attention_mask[cur_batch_idx]
+
+            image_token_indices = torch.where(cur_input_ids == IMAGE_TOKEN_INDEX)[0]
+
+            cur_image_num = len(image_token_indices)
+            image_token_indices = list(image_token_indices) + [cur_input_ids.shape[0]]
+
+            cur_input_embeds = []
+            cur_attention_mask = []
+            cur_target = []
+
+            # convert text before 1st <image> to embedding
+            image_token_index = image_token_indices[0]
+
+            cur_input_embeds.append(
+                self.llm.get_input_embeddings()(cur_input_ids[:image_token_index]),
+            )
+            cur_attention_mask.append(
+                cur_attn_masks[:image_token_index],
+            )
+            if labels is not None:
+                cur_target.append(labels[cur_batch_idx, :image_token_index])
+
+            assert cur_image_num == len(cur_vision_embeds), \
+                f"Size mismatch! cur_image_num: {cur_image_num}, len(cur_vision_embeds): {len(cur_vision_embeds)}"
+            # convert each <image> xxx group into embedding
+            for i in range(0, cur_image_num):
+                image_token_index = image_token_indices[i]
+                cur_input_embeds.append(torch.cat([
+                    cur_vision_embeds[i],
+                    self.llm.get_input_embeddings()(cur_input_ids[image_token_index+1:image_token_indices[i+1]])
+                ]))
+                cur_attention_mask.append(torch.cat([
+                    cur_vision_attn[i],
+                    cur_attn_masks[image_token_index+1:image_token_indices[i+1]]
+                ]))
+                if labels is not None:
+                    cur_target.append(torch.cat([
+                        cur_vision_targets[i],
+                        labels[cur_batch_idx, image_token_index+1:image_token_indices[i+1]],
+                    ]))
+
+            input_embeds.append(torch.cat(cur_input_embeds))
+            attention_masks.append(torch.cat(cur_attention_mask))
+            if labels is not None:
+                targets.append(torch.cat(cur_target))
+
+        # padding
+        n_tokens = [embed.shape[0] for embed in input_embeds]
+        max_token = max(n_tokens)
+        for i in range(len(input_embeds)):
+            if max_token > n_tokens[i]:
+                self.pad_id = self.tokenizer.pad_token_id or self.tokenizer.eos_token_id
+                pad_token = torch.tensor([self.pad_id] * (max_token - n_tokens[i]))
+                pad_embedding = self.llm.get_input_embeddings()(pad_token.to(vision_embeds[i].device))
+                pad_attention = torch.zeros(pad_embedding.shape[0], dtype=torch.long).to(vision_embeds[i].device)
+                pad_targets = torch.ones(pad_attention.size(), dtype=torch.long).to(vision_embeds[i].device).fill_(-100)
+
+                if left_padding:
+                    input_embeds[i] = torch.cat([pad_embedding, input_embeds[i]])
+                    attention_masks[i] = torch.cat([pad_attention, attention_masks[i]])
+                    if labels is not None:
+                        targets[i] = torch.cat([pad_targets, targets[i]])
+                else:
+                    input_embeds[i] = torch.cat([input_embeds[i], pad_embedding])
+                    attention_masks[i] = torch.cat([attention_masks[i], pad_attention])
+                    if labels is not None:
+                        targets[i] = torch.cat([targets[i], pad_targets])
+
+        inputs_embeds = torch.stack(input_embeds, dim=0).type(self.llm.dtype)
+        attention_masks = torch.stack(attention_masks, dim=0)
+
+        if len(targets) > 0:
+            targets = torch.stack(targets, dim=0)
+
+        return inputs_embeds, attention_masks, targets
+
+    def forward(self,
+                input_ids: torch.LongTensor = None,
+                attention_mask: Optional[torch.Tensor] = None,
+                frames: torch.LongTensor = None,
+                n_frames: List[int] = None,
+                labels: Optional[torch.LongTensor] = None,
+                **kwargs) -> Union[Tuple, CausalLMOutputWithPast]:
+        # get vision features
+        vision_encode_out = self._encode_vision(frames, n_frames)
+
+        inputs_embeds, attention_mask, targets = self._concat_embedding(
+            vision_encode_out, input_ids, attention_mask, labels)
+
+        # input to llm
+        outputs = self.llm(
+            inputs_embeds=inputs_embeds,
+            attention_mask=attention_mask,
+            labels=targets,
+            return_dict=True,
+        )
+        return outputs
+
+    def _enable_gradient_checkpointing(self):
+        for model in (self.visual_encoder, self.llm):
+            if hasattr(model, "gradient_checkpointing_enable"):
+                model.gradient_checkpointing_enable()
+                if hasattr(model, "enable_input_require_grads"):
+                    model.enable_input_require_grads()
+                else:
+                    def make_inputs_require_grad(module, input, output):
+                        output.requires_grad_(True)
+                    model.get_input_embeddings().register_forward_hook(make_inputs_require_grad)
+
+    def generate(self,
+                 frames: torch.Tensor,
+                 n_frames: List[int],
+                 input_ids: torch.LongTensor,
+                 attention_mask: torch.Tensor,
+                 **kwargs):
+        with torch.cuda.amp.autocast(
+            enabled=(self.device != torch.device("cpu"))
+        ):
+            vision_encode_out = self._encode_vision(frames, n_frames)
+            inputs_embeds, attention_mask, _ = self._concat_embedding(
+                vision_encode_out, input_ids, attention_mask)
+
+        outputs = self.llm.generate(
+            inputs_embeds=inputs_embeds,
+            attention_mask=attention_mask,
+            eos_token_id=self.tokenizer.eos_token_id,
+            **kwargs
+        )
+
+        # parse result text
+        output_text = self.tokenizer.batch_decode(
+            outputs, skip_special_tokens=True
+        )
+
+        return output_text

rantselector.py

@@ -0,0 +1,80 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class RanTSelecor(nn.Module):
+    def __init__(
+        self,
+        num_features=1408,
+        in_token_num=576,
+        out_token_num=144,
+        min_out_token_num=16,
+        max_video_tokens=2048,
+        fix_random=False,
+        **kwargs,
+    ):
+        super().__init__()
+        self.in_token_num = in_token_num
+        self.out_token_num = out_token_num
+        self.min_out_token_num = min_out_token_num
+        self.max_video_tokens = max_video_tokens
+        self.pre_proj = nn.Linear(num_features, 1)
+        self.score_proj = nn.Linear(in_token_num, out_token_num)
+
+        self.trans_weight = nn.Parameter(
+            torch.zeros(in_token_num, in_token_num), requires_grad=True
+        )
+        self.trans_bias = nn.Parameter(self._init_bias(), requires_grad=True)
+
+        self.hidden_size = num_features
+        self._init_bias()
+        self.fix_random = fix_random
+
+    def _init_bias(self):
+        tensor = torch.zeros(self.in_token_num)
+        m = self.out_token_num
+        n = tensor.numel()
+        if m > n:
+            m = n
+        count = 0
+
+        n_group = n // m
+        interval = max(4, n_group)
+
+        for j in range(0, interval):
+            for i in range(j, n, interval):
+                tensor[i] = 0.02
+                count += 1
+                if count >= m:
+                    break
+            if count >= m:
+                break
+        return tensor
+
+    def forward(self, image_embeds, n_frames, noise_epsilon=0.001):
+        image_embeds_list = image_embeds.split(n_frames, dim=0)
+        # Compute temporal tokens as the mean along the time axis
+        ret_tokens = []
+        for image_embeds_per_video in image_embeds_list:
+            video_raw_token = image_embeds_per_video
+
+            video_raw_token_trans = self.pre_proj(video_raw_token).mT
+            video_token_logits = (
+                video_raw_token_trans @ self.trans_weight + self.trans_bias
+            )
+
+            video_token_logits = video_token_logits.squeeze(1)
+
+            video_token_scores = F.softmax(video_token_logits)
+            topk_indices = torch.argsort(video_token_scores, descending=True)[:, :self.out_token_num]
+            topk_indices, _ = torch.sort(topk_indices)
+            video_topk_token = video_raw_token[
+                torch.arange(video_raw_token.size(0)).unsqueeze(1), topk_indices
+            ]
+            ret_tokens.append(video_topk_token)
+        return ret_tokens
+
+
+def build_adapter(config):
+    return RanTSelecor(**config)