Upload folder using huggingface_hub
Browse files- .eval_results/olmocrbench.yaml +72 -0
- .gitattributes +1 -0
- README.md +36 -0
- chat_template.jinja +2 -0
- chat_template.json +3 -0
- config.json +65 -0
- configuration_dots.py +78 -0
- generation_config.json +8 -0
- model.safetensors +3 -0
- model.safetensors.index.json +1046 -0
- modeling_dots_ocr.py +131 -0
- modeling_dots_vision.py +404 -0
- preprocessor_config.json +22 -0
- processor_config.json +34 -0
- special_tokens_map.json +25 -0
- tokenizer.json +3 -0
- tokenizer_config.json +30 -0
- vocab.json +0 -0
.eval_results/olmocrbench.yaml
ADDED
|
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
- dataset:
|
| 2 |
+
id: allenai/olmOCR-bench
|
| 3 |
+
task_id: overall
|
| 4 |
+
value: 83.9
|
| 5 |
+
source:
|
| 6 |
+
url: https://huggingface.co/papers/2603.13032
|
| 7 |
+
name: dots.mocr technical report
|
| 8 |
+
user: nielsr
|
| 9 |
+
- dataset:
|
| 10 |
+
id: allenai/olmOCR-bench
|
| 11 |
+
task_id: arxiv_math
|
| 12 |
+
value: 85.9
|
| 13 |
+
source:
|
| 14 |
+
url: https://huggingface.co/papers/2603.13032
|
| 15 |
+
name: dots.mocr technical report
|
| 16 |
+
user: nielsr
|
| 17 |
+
- dataset:
|
| 18 |
+
id: allenai/olmOCR-bench
|
| 19 |
+
task_id: old_scans_math
|
| 20 |
+
value: 85.5
|
| 21 |
+
source:
|
| 22 |
+
url: https://huggingface.co/papers/2603.13032
|
| 23 |
+
name: dots.mocr technical report
|
| 24 |
+
user: nielsr
|
| 25 |
+
- dataset:
|
| 26 |
+
id: allenai/olmOCR-bench
|
| 27 |
+
task_id: table_tests
|
| 28 |
+
value: 90.7
|
| 29 |
+
source:
|
| 30 |
+
url: https://huggingface.co/papers/2603.13032
|
| 31 |
+
name: dots.mocr technical report
|
| 32 |
+
user: nielsr
|
| 33 |
+
- dataset:
|
| 34 |
+
id: allenai/olmOCR-bench
|
| 35 |
+
task_id: old_scans
|
| 36 |
+
value: 48.2
|
| 37 |
+
source:
|
| 38 |
+
url: https://huggingface.co/papers/2603.13032
|
| 39 |
+
name: dots.mocr technical report
|
| 40 |
+
user: nielsr
|
| 41 |
+
- dataset:
|
| 42 |
+
id: allenai/olmOCR-bench
|
| 43 |
+
task_id: multi_column
|
| 44 |
+
value: 85.3
|
| 45 |
+
source:
|
| 46 |
+
url: https://huggingface.co/papers/2603.13032
|
| 47 |
+
name: dots.mocr technical report
|
| 48 |
+
user: nielsr
|
| 49 |
+
- dataset:
|
| 50 |
+
id: allenai/olmOCR-bench
|
| 51 |
+
task_id: long_tiny_text
|
| 52 |
+
value: 81.6
|
| 53 |
+
source:
|
| 54 |
+
url: https://huggingface.co/papers/2603.13032
|
| 55 |
+
name: dots.mocr technical report
|
| 56 |
+
user: nielsr
|
| 57 |
+
- dataset:
|
| 58 |
+
id: allenai/olmOCR-bench
|
| 59 |
+
task_id: headers_footers
|
| 60 |
+
value: 94.0
|
| 61 |
+
source:
|
| 62 |
+
url: https://huggingface.co/papers/2603.13032
|
| 63 |
+
name: dots.mocr technical report
|
| 64 |
+
user: nielsr
|
| 65 |
+
- dataset:
|
| 66 |
+
id: allenai/olmOCR-bench
|
| 67 |
+
task_id: baseline
|
| 68 |
+
value: 99.7
|
| 69 |
+
source:
|
| 70 |
+
url: https://huggingface.co/papers/2603.13032
|
| 71 |
+
name: dots.mocr technical report
|
| 72 |
+
user: nielsr
|
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
library_name: mlx
|
| 4 |
+
pipeline_tag: image-text-to-text
|
| 5 |
+
tags:
|
| 6 |
+
- image-to-text
|
| 7 |
+
- ocr
|
| 8 |
+
- document-parse
|
| 9 |
+
- layout
|
| 10 |
+
- table
|
| 11 |
+
- formula
|
| 12 |
+
- transformers
|
| 13 |
+
- custom_code
|
| 14 |
+
- mlx
|
| 15 |
+
language:
|
| 16 |
+
- en
|
| 17 |
+
- zh
|
| 18 |
+
- multilingual
|
| 19 |
+
base_model: rednote-hilab/dots.mocr
|
| 20 |
+
---
|
| 21 |
+
|
| 22 |
+
# mlx-community/dots.mocr-8bit
|
| 23 |
+
|
| 24 |
+
This model was converted to MLX format from [`rednote-hilab/dots.mocr`](https://huggingface.co/rednote-hilab/dots.mocr)
|
| 25 |
+
using mlx-vlm version **0.4.1**.
|
| 26 |
+
Refer to the [original model card](https://huggingface.co/rednote-hilab/dots.mocr) for more details on the model.
|
| 27 |
+
|
| 28 |
+
## Use with mlx
|
| 29 |
+
|
| 30 |
+
```bash
|
| 31 |
+
pip install -U mlx-vlm
|
| 32 |
+
```
|
| 33 |
+
|
| 34 |
+
```bash
|
| 35 |
+
python -m mlx_vlm.generate --model mlx-community/dots.mocr-8bit --max-tokens 100 --temperature 0.0 --prompt "Describe this image." --image <path_to_image>
|
| 36 |
+
```
|
chat_template.jinja
ADDED
|
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{% set image_count = namespace(value=0) %}{% set video_count = namespace(value=0) %}{%- for m in messages %}{%- if m.role == 'system' %}{{- '<|system|>' + m.content + '<|endofsystem|>
|
| 2 |
+
' }}{%- elif m.role == 'user' %}{% if m.content is string %}{{- '<|user|>' + m.content + '<|endofuser|>' }}{% else %}{{- '<|user|>' }}{% for content in m.content %}{% if content['type'] == 'image' or 'image' in content or 'image_url' in content %}{% set image_count.value = image_count.value + 1 %}{% if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<|img|><|imgpad|><|endofimg|>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}{% if add_vision_id %}Video {{ video_count.value }}: {% endif %}<|img|><|video_pad|><|endofimg|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}{{- '<|endofuser|>' }}{%- endif %}{%- elif m.role == 'assistant' %}{{- '<|assistant|>' + m.content }}{%- if not loop.last %}{{- '<|endofassistant|>' }}{%- endif %}{%- endif %}{%- endfor %}{%- if messages[-1].role != 'assistant' %}{{- '<|assistant|>' }}{%- endif %}
|
chat_template.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"chat_template": "{% set image_count = namespace(value=0) %}{% set video_count = namespace(value=0) %}{%- for m in messages %}{%- if m.role == 'system' %}{{- '<|system|>' + m.content + '<|endofsystem|>\n' }}{%- elif m.role == 'user' %}{% if m.content is string %}{{- '<|user|>' + m.content + '<|endofuser|>' }}{% else %}{{- '<|user|>' }}{% for content in m.content %}{% if content['type'] == 'image' or 'image' in content or 'image_url' in content %}{% set image_count.value = image_count.value + 1 %}{% if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<|img|><|imgpad|><|endofimg|>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}{% if add_vision_id %}Video {{ video_count.value }}: {% endif %}<|img|><|video_pad|><|endofimg|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}{{- '<|endofuser|>' }}{%- endif %}{%- elif m.role == 'assistant' %}{{- '<|assistant|>' + m.content }}{%- if not loop.last %}{{- '<|endofassistant|>' }}{%- endif %}{%- endif %}{%- endfor %}{%- if messages[-1].role != 'assistant' %}{{- '<|assistant|>' }}{%- endif %}"
|
| 3 |
+
}
|
config.json
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"DotsOCRForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": true,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"auto_map": {
|
| 8 |
+
"AutoConfig": "configuration_dots.DotsOCRConfig",
|
| 9 |
+
"AutoModelForCausalLM": "modeling_dots_ocr.DotsOCRForCausalLM"
|
| 10 |
+
},
|
| 11 |
+
"eos_token_id": [
|
| 12 |
+
151643,
|
| 13 |
+
151672,
|
| 14 |
+
151673
|
| 15 |
+
],
|
| 16 |
+
"hidden_act": "silu",
|
| 17 |
+
"hidden_size": 1536,
|
| 18 |
+
"image_token_id": 151665,
|
| 19 |
+
"initializer_range": 0.02,
|
| 20 |
+
"intermediate_size": 8960,
|
| 21 |
+
"max_position_embeddings": 131072,
|
| 22 |
+
"max_window_layers": 28,
|
| 23 |
+
"model_type": "dots_ocr",
|
| 24 |
+
"num_attention_heads": 12,
|
| 25 |
+
"num_hidden_layers": 28,
|
| 26 |
+
"num_key_value_heads": 2,
|
| 27 |
+
"quantization": {
|
| 28 |
+
"group_size": 64,
|
| 29 |
+
"bits": 8,
|
| 30 |
+
"mode": "affine"
|
| 31 |
+
},
|
| 32 |
+
"quantization_config": {
|
| 33 |
+
"group_size": 64,
|
| 34 |
+
"bits": 8,
|
| 35 |
+
"mode": "affine"
|
| 36 |
+
},
|
| 37 |
+
"rms_norm_eps": 1e-06,
|
| 38 |
+
"rope_scaling": null,
|
| 39 |
+
"rope_theta": 1000000,
|
| 40 |
+
"sliding_window": 131072,
|
| 41 |
+
"tie_word_embeddings": false,
|
| 42 |
+
"transformers_version": "4.51.0",
|
| 43 |
+
"use_cache": true,
|
| 44 |
+
"use_sliding_window": false,
|
| 45 |
+
"video_token_id": 151656,
|
| 46 |
+
"vision_config": {
|
| 47 |
+
"embed_dim": 1536,
|
| 48 |
+
"hidden_size": 1536,
|
| 49 |
+
"intermediate_size": 4224,
|
| 50 |
+
"num_hidden_layers": 42,
|
| 51 |
+
"num_attention_heads": 12,
|
| 52 |
+
"num_channels": 3,
|
| 53 |
+
"patch_size": 14,
|
| 54 |
+
"post_norm": true,
|
| 55 |
+
"rms_norm_eps": 1e-05,
|
| 56 |
+
"spatial_merge_size": 2,
|
| 57 |
+
"temporal_patch_size": 1,
|
| 58 |
+
"use_bias": false,
|
| 59 |
+
"attn_implementation": "flash_attention_2",
|
| 60 |
+
"init_merger_std": 0.02,
|
| 61 |
+
"initializer_range": 0.02,
|
| 62 |
+
"is_causal": false
|
| 63 |
+
},
|
| 64 |
+
"vocab_size": 151936
|
| 65 |
+
}
|
configuration_dots.py
ADDED
|
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from typing import Any, Optional
|
| 2 |
+
from transformers.configuration_utils import PretrainedConfig
|
| 3 |
+
from transformers.models.qwen2 import Qwen2Config
|
| 4 |
+
from transformers import Qwen2_5_VLProcessor, AutoProcessor
|
| 5 |
+
from transformers.models.auto.configuration_auto import CONFIG_MAPPING
|
| 6 |
+
|
| 7 |
+
|
| 8 |
+
class DotsVisionConfig(PretrainedConfig):
|
| 9 |
+
model_type: str = "dots_vit"
|
| 10 |
+
|
| 11 |
+
def __init__(
|
| 12 |
+
self,
|
| 13 |
+
embed_dim: int = 1536, # vision encoder embed size
|
| 14 |
+
hidden_size: int = 1536, # after merger hidden size
|
| 15 |
+
intermediate_size: int = 4224,
|
| 16 |
+
num_hidden_layers: int = 42,
|
| 17 |
+
num_attention_heads: int = 12,
|
| 18 |
+
num_channels: int = 3,
|
| 19 |
+
patch_size: int = 14,
|
| 20 |
+
spatial_merge_size: int = 2,
|
| 21 |
+
temporal_patch_size: int = 1,
|
| 22 |
+
rms_norm_eps: float = 1e-5,
|
| 23 |
+
use_bias: bool = False,
|
| 24 |
+
attn_implementation="flash_attention_2", # "eager","sdpa","flash_attention_2"
|
| 25 |
+
initializer_range=0.02,
|
| 26 |
+
init_merger_std=0.02,
|
| 27 |
+
is_causal=False, # ve causal forward
|
| 28 |
+
post_norm=True,
|
| 29 |
+
gradient_checkpointing=False,
|
| 30 |
+
**kwargs: Any,
|
| 31 |
+
):
|
| 32 |
+
super().__init__(**kwargs)
|
| 33 |
+
self.embed_dim = embed_dim
|
| 34 |
+
self.hidden_size = hidden_size
|
| 35 |
+
self.intermediate_size = intermediate_size
|
| 36 |
+
self.num_hidden_layers = num_hidden_layers
|
| 37 |
+
self.num_attention_heads = num_attention_heads
|
| 38 |
+
self.num_channels = num_channels
|
| 39 |
+
self.patch_size = patch_size
|
| 40 |
+
self.spatial_merge_size = spatial_merge_size
|
| 41 |
+
self.temporal_patch_size = temporal_patch_size
|
| 42 |
+
self.rms_norm_eps = rms_norm_eps
|
| 43 |
+
self.use_bias = use_bias
|
| 44 |
+
self.attn_implementation = attn_implementation
|
| 45 |
+
self.initializer_range = initializer_range
|
| 46 |
+
self.init_merger_std = init_merger_std
|
| 47 |
+
self.is_causal = is_causal
|
| 48 |
+
self.post_norm = post_norm
|
| 49 |
+
self.gradient_checkpointing = gradient_checkpointing
|
| 50 |
+
|
| 51 |
+
|
| 52 |
+
|
| 53 |
+
class DotsOCRConfig(Qwen2Config):
|
| 54 |
+
model_type = "dots_ocr"
|
| 55 |
+
def __init__(self,
|
| 56 |
+
image_token_id = 151665,
|
| 57 |
+
video_token_id = 151656,
|
| 58 |
+
vision_config: Optional[dict] = None, *args, **kwargs):
|
| 59 |
+
super().__init__(*args, **kwargs)
|
| 60 |
+
self.image_token_id = image_token_id
|
| 61 |
+
self.video_token_id = video_token_id
|
| 62 |
+
self.vision_config = DotsVisionConfig(**(vision_config or {}))
|
| 63 |
+
|
| 64 |
+
def save_pretrained(self, save_directory, **kwargs):
|
| 65 |
+
self._auto_class = None
|
| 66 |
+
super().save_pretrained(save_directory, **kwargs)
|
| 67 |
+
|
| 68 |
+
|
| 69 |
+
class DotsVLProcessor(Qwen2_5_VLProcessor):
|
| 70 |
+
def __init__(self, image_processor=None, tokenizer=None, video_processor=None, chat_template=None, **kwargs):
|
| 71 |
+
super().__init__(image_processor, tokenizer, video_processor, chat_template=chat_template)
|
| 72 |
+
self.image_token = "<|imgpad|>" if not hasattr(tokenizer, "image_token") else tokenizer.image_token
|
| 73 |
+
self.image_token_id = 151665
|
| 74 |
+
self.video_token = "<|video_pad|>" if not hasattr(tokenizer, "video_token") else tokenizer.video_token
|
| 75 |
+
self.video_token_id = 151656
|
| 76 |
+
|
| 77 |
+
AutoProcessor.register("dots_ocr", DotsVLProcessor)
|
| 78 |
+
CONFIG_MAPPING.register("dots_ocr", DotsOCRConfig)
|
generation_config.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"max_length": 32768,
|
| 3 |
+
"eos_token_id": [
|
| 4 |
+
151643,
|
| 5 |
+
151672,
|
| 6 |
+
151673
|
| 7 |
+
]
|
| 8 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9d8c0ed06334ed486db9aeb7ea8e1b22e807693c5aad629e93679fc5381da757
|
| 3 |
+
size 4412602873
|
model.safetensors.index.json
ADDED
|
@@ -0,0 +1,1046 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"metadata": {
|
| 3 |
+
"total_size": 4412474368
|
| 4 |
+
},
|
| 5 |
+
"weight_map": {
|
| 6 |
+
"language_model.model.embed_tokens.biases": "model.safetensors",
|
| 7 |
+
"language_model.model.embed_tokens.scales": "model.safetensors",
|
| 8 |
+
"language_model.model.embed_tokens.weight": "model.safetensors",
|
| 9 |
+
"language_model.model.layers.0.input_layernorm.weight": "model.safetensors",
|
| 10 |
+
"language_model.model.layers.0.mlp.down_proj.biases": "model.safetensors",
|
| 11 |
+
"language_model.model.layers.0.mlp.down_proj.scales": "model.safetensors",
|
| 12 |
+
"language_model.model.layers.0.mlp.down_proj.weight": "model.safetensors",
|
| 13 |
+
"language_model.model.layers.0.mlp.gate_proj.biases": "model.safetensors",
|
| 14 |
+
"language_model.model.layers.0.mlp.gate_proj.scales": "model.safetensors",
|
| 15 |
+
"language_model.model.layers.0.mlp.gate_proj.weight": "model.safetensors",
|
| 16 |
+
"language_model.model.layers.0.mlp.up_proj.biases": "model.safetensors",
|
| 17 |
+
"language_model.model.layers.0.mlp.up_proj.scales": "model.safetensors",
|
| 18 |
+
"language_model.model.layers.0.mlp.up_proj.weight": "model.safetensors",
|
| 19 |
+
"language_model.model.layers.0.post_attention_layernorm.weight": "model.safetensors",
|
| 20 |
+
"language_model.model.layers.0.self_attn.k_proj.bias": "model.safetensors",
|
| 21 |
+
"language_model.model.layers.0.self_attn.k_proj.biases": "model.safetensors",
|
| 22 |
+
"language_model.model.layers.0.self_attn.k_proj.scales": "model.safetensors",
|
| 23 |
+
"language_model.model.layers.0.self_attn.k_proj.weight": "model.safetensors",
|
| 24 |
+
"language_model.model.layers.0.self_attn.o_proj.biases": "model.safetensors",
|
| 25 |
+
"language_model.model.layers.0.self_attn.o_proj.scales": "model.safetensors",
|
| 26 |
+
"language_model.model.layers.0.self_attn.o_proj.weight": "model.safetensors",
|
| 27 |
+
"language_model.model.layers.0.self_attn.q_proj.bias": "model.safetensors",
|
| 28 |
+
"language_model.model.layers.0.self_attn.q_proj.biases": "model.safetensors",
|
| 29 |
+
"language_model.model.layers.0.self_attn.q_proj.scales": "model.safetensors",
|
| 30 |
+
"language_model.model.layers.0.self_attn.q_proj.weight": "model.safetensors",
|
| 31 |
+
"language_model.model.layers.0.self_attn.v_proj.bias": "model.safetensors",
|
| 32 |
+
"language_model.model.layers.0.self_attn.v_proj.biases": "model.safetensors",
|
| 33 |
+
"language_model.model.layers.0.self_attn.v_proj.scales": "model.safetensors",
|
| 34 |
+
"language_model.model.layers.0.self_attn.v_proj.weight": "model.safetensors",
|
| 35 |
+
"language_model.model.layers.1.input_layernorm.weight": "model.safetensors",
|
| 36 |
+
"language_model.model.layers.1.mlp.down_proj.biases": "model.safetensors",
|
| 37 |
+
"language_model.model.layers.1.mlp.down_proj.scales": "model.safetensors",
|
| 38 |
+
"language_model.model.layers.1.mlp.down_proj.weight": "model.safetensors",
|
| 39 |
+
"language_model.model.layers.1.mlp.gate_proj.biases": "model.safetensors",
|
| 40 |
+
"language_model.model.layers.1.mlp.gate_proj.scales": "model.safetensors",
|
| 41 |
+
"language_model.model.layers.1.mlp.gate_proj.weight": "model.safetensors",
|
| 42 |
+
"language_model.model.layers.1.mlp.up_proj.biases": "model.safetensors",
|
| 43 |
+
"language_model.model.layers.1.mlp.up_proj.scales": "model.safetensors",
|
| 44 |
+
"language_model.model.layers.1.mlp.up_proj.weight": "model.safetensors",
|
| 45 |
+
"language_model.model.layers.1.post_attention_layernorm.weight": "model.safetensors",
|
| 46 |
+
"language_model.model.layers.1.self_attn.k_proj.bias": "model.safetensors",
|
| 47 |
+
"language_model.model.layers.1.self_attn.k_proj.biases": "model.safetensors",
|
| 48 |
+
"language_model.model.layers.1.self_attn.k_proj.scales": "model.safetensors",
|
| 49 |
+
"language_model.model.layers.1.self_attn.k_proj.weight": "model.safetensors",
|
| 50 |
+
"language_model.model.layers.1.self_attn.o_proj.biases": "model.safetensors",
|
| 51 |
+
"language_model.model.layers.1.self_attn.o_proj.scales": "model.safetensors",
|
| 52 |
+
"language_model.model.layers.1.self_attn.o_proj.weight": "model.safetensors",
|
| 53 |
+
"language_model.model.layers.1.self_attn.q_proj.bias": "model.safetensors",
|
| 54 |
+
"language_model.model.layers.1.self_attn.q_proj.biases": "model.safetensors",
|
| 55 |
+
"language_model.model.layers.1.self_attn.q_proj.scales": "model.safetensors",
|
| 56 |
+
"language_model.model.layers.1.self_attn.q_proj.weight": "model.safetensors",
|
| 57 |
+
"language_model.model.layers.1.self_attn.v_proj.bias": "model.safetensors",
|
| 58 |
+
"language_model.model.layers.1.self_attn.v_proj.biases": "model.safetensors",
|
| 59 |
+
"language_model.model.layers.1.self_attn.v_proj.scales": "model.safetensors",
|
| 60 |
+
"language_model.model.layers.1.self_attn.v_proj.weight": "model.safetensors",
|
| 61 |
+
"language_model.model.layers.10.input_layernorm.weight": "model.safetensors",
|
| 62 |
+
"language_model.model.layers.10.mlp.down_proj.biases": "model.safetensors",
|
| 63 |
+
"language_model.model.layers.10.mlp.down_proj.scales": "model.safetensors",
|
| 64 |
+
"language_model.model.layers.10.mlp.down_proj.weight": "model.safetensors",
|
| 65 |
+
"language_model.model.layers.10.mlp.gate_proj.biases": "model.safetensors",
|
| 66 |
+
"language_model.model.layers.10.mlp.gate_proj.scales": "model.safetensors",
|
| 67 |
+
"language_model.model.layers.10.mlp.gate_proj.weight": "model.safetensors",
|
| 68 |
+
"language_model.model.layers.10.mlp.up_proj.biases": "model.safetensors",
|
| 69 |
+
"language_model.model.layers.10.mlp.up_proj.scales": "model.safetensors",
|
| 70 |
+
"language_model.model.layers.10.mlp.up_proj.weight": "model.safetensors",
|
| 71 |
+
"language_model.model.layers.10.post_attention_layernorm.weight": "model.safetensors",
|
| 72 |
+
"language_model.model.layers.10.self_attn.k_proj.bias": "model.safetensors",
|
| 73 |
+
"language_model.model.layers.10.self_attn.k_proj.biases": "model.safetensors",
|
| 74 |
+
"language_model.model.layers.10.self_attn.k_proj.scales": "model.safetensors",
|
| 75 |
+
"language_model.model.layers.10.self_attn.k_proj.weight": "model.safetensors",
|
| 76 |
+
"language_model.model.layers.10.self_attn.o_proj.biases": "model.safetensors",
|
| 77 |
+
"language_model.model.layers.10.self_attn.o_proj.scales": "model.safetensors",
|
| 78 |
+
"language_model.model.layers.10.self_attn.o_proj.weight": "model.safetensors",
|
| 79 |
+
"language_model.model.layers.10.self_attn.q_proj.bias": "model.safetensors",
|
| 80 |
+
"language_model.model.layers.10.self_attn.q_proj.biases": "model.safetensors",
|
| 81 |
+
"language_model.model.layers.10.self_attn.q_proj.scales": "model.safetensors",
|
| 82 |
+
"language_model.model.layers.10.self_attn.q_proj.weight": "model.safetensors",
|
| 83 |
+
"language_model.model.layers.10.self_attn.v_proj.bias": "model.safetensors",
|
| 84 |
+
"language_model.model.layers.10.self_attn.v_proj.biases": "model.safetensors",
|
| 85 |
+
"language_model.model.layers.10.self_attn.v_proj.scales": "model.safetensors",
|
| 86 |
+
"language_model.model.layers.10.self_attn.v_proj.weight": "model.safetensors",
|
| 87 |
+
"language_model.model.layers.11.input_layernorm.weight": "model.safetensors",
|
| 88 |
+
"language_model.model.layers.11.mlp.down_proj.biases": "model.safetensors",
|
| 89 |
+
"language_model.model.layers.11.mlp.down_proj.scales": "model.safetensors",
|
| 90 |
+
"language_model.model.layers.11.mlp.down_proj.weight": "model.safetensors",
|
| 91 |
+
"language_model.model.layers.11.mlp.gate_proj.biases": "model.safetensors",
|
| 92 |
+
"language_model.model.layers.11.mlp.gate_proj.scales": "model.safetensors",
|
| 93 |
+
"language_model.model.layers.11.mlp.gate_proj.weight": "model.safetensors",
|
| 94 |
+
"language_model.model.layers.11.mlp.up_proj.biases": "model.safetensors",
|
| 95 |
+
"language_model.model.layers.11.mlp.up_proj.scales": "model.safetensors",
|
| 96 |
+
"language_model.model.layers.11.mlp.up_proj.weight": "model.safetensors",
|
| 97 |
+
"language_model.model.layers.11.post_attention_layernorm.weight": "model.safetensors",
|
| 98 |
+
"language_model.model.layers.11.self_attn.k_proj.bias": "model.safetensors",
|
| 99 |
+
"language_model.model.layers.11.self_attn.k_proj.biases": "model.safetensors",
|
| 100 |
+
"language_model.model.layers.11.self_attn.k_proj.scales": "model.safetensors",
|
| 101 |
+
"language_model.model.layers.11.self_attn.k_proj.weight": "model.safetensors",
|
| 102 |
+
"language_model.model.layers.11.self_attn.o_proj.biases": "model.safetensors",
|
| 103 |
+
"language_model.model.layers.11.self_attn.o_proj.scales": "model.safetensors",
|
| 104 |
+
"language_model.model.layers.11.self_attn.o_proj.weight": "model.safetensors",
|
| 105 |
+
"language_model.model.layers.11.self_attn.q_proj.bias": "model.safetensors",
|
| 106 |
+
"language_model.model.layers.11.self_attn.q_proj.biases": "model.safetensors",
|
| 107 |
+
"language_model.model.layers.11.self_attn.q_proj.scales": "model.safetensors",
|
| 108 |
+
"language_model.model.layers.11.self_attn.q_proj.weight": "model.safetensors",
|
| 109 |
+
"language_model.model.layers.11.self_attn.v_proj.bias": "model.safetensors",
|
| 110 |
+
"language_model.model.layers.11.self_attn.v_proj.biases": "model.safetensors",
|
| 111 |
+
"language_model.model.layers.11.self_attn.v_proj.scales": "model.safetensors",
|
| 112 |
+
"language_model.model.layers.11.self_attn.v_proj.weight": "model.safetensors",
|
| 113 |
+
"language_model.model.layers.12.input_layernorm.weight": "model.safetensors",
|
| 114 |
+
"language_model.model.layers.12.mlp.down_proj.biases": "model.safetensors",
|
| 115 |
+
"language_model.model.layers.12.mlp.down_proj.scales": "model.safetensors",
|
| 116 |
+
"language_model.model.layers.12.mlp.down_proj.weight": "model.safetensors",
|
| 117 |
+
"language_model.model.layers.12.mlp.gate_proj.biases": "model.safetensors",
|
| 118 |
+
"language_model.model.layers.12.mlp.gate_proj.scales": "model.safetensors",
|
| 119 |
+
"language_model.model.layers.12.mlp.gate_proj.weight": "model.safetensors",
|
| 120 |
+
"language_model.model.layers.12.mlp.up_proj.biases": "model.safetensors",
|
| 121 |
+
"language_model.model.layers.12.mlp.up_proj.scales": "model.safetensors",
|
| 122 |
+
"language_model.model.layers.12.mlp.up_proj.weight": "model.safetensors",
|
| 123 |
+
"language_model.model.layers.12.post_attention_layernorm.weight": "model.safetensors",
|
| 124 |
+
"language_model.model.layers.12.self_attn.k_proj.bias": "model.safetensors",
|
| 125 |
+
"language_model.model.layers.12.self_attn.k_proj.biases": "model.safetensors",
|
| 126 |
+
"language_model.model.layers.12.self_attn.k_proj.scales": "model.safetensors",
|
| 127 |
+
"language_model.model.layers.12.self_attn.k_proj.weight": "model.safetensors",
|
| 128 |
+
"language_model.model.layers.12.self_attn.o_proj.biases": "model.safetensors",
|
| 129 |
+
"language_model.model.layers.12.self_attn.o_proj.scales": "model.safetensors",
|
| 130 |
+
"language_model.model.layers.12.self_attn.o_proj.weight": "model.safetensors",
|
| 131 |
+
"language_model.model.layers.12.self_attn.q_proj.bias": "model.safetensors",
|
| 132 |
+
"language_model.model.layers.12.self_attn.q_proj.biases": "model.safetensors",
|
| 133 |
+
"language_model.model.layers.12.self_attn.q_proj.scales": "model.safetensors",
|
| 134 |
+
"language_model.model.layers.12.self_attn.q_proj.weight": "model.safetensors",
|
| 135 |
+
"language_model.model.layers.12.self_attn.v_proj.bias": "model.safetensors",
|
| 136 |
+
"language_model.model.layers.12.self_attn.v_proj.biases": "model.safetensors",
|
| 137 |
+
"language_model.model.layers.12.self_attn.v_proj.scales": "model.safetensors",
|
| 138 |
+
"language_model.model.layers.12.self_attn.v_proj.weight": "model.safetensors",
|
| 139 |
+
"language_model.model.layers.13.input_layernorm.weight": "model.safetensors",
|
| 140 |
+
"language_model.model.layers.13.mlp.down_proj.biases": "model.safetensors",
|
| 141 |
+
"language_model.model.layers.13.mlp.down_proj.scales": "model.safetensors",
|
| 142 |
+
"language_model.model.layers.13.mlp.down_proj.weight": "model.safetensors",
|
| 143 |
+
"language_model.model.layers.13.mlp.gate_proj.biases": "model.safetensors",
|
| 144 |
+
"language_model.model.layers.13.mlp.gate_proj.scales": "model.safetensors",
|
| 145 |
+
"language_model.model.layers.13.mlp.gate_proj.weight": "model.safetensors",
|
| 146 |
+
"language_model.model.layers.13.mlp.up_proj.biases": "model.safetensors",
|
| 147 |
+
"language_model.model.layers.13.mlp.up_proj.scales": "model.safetensors",
|
| 148 |
+
"language_model.model.layers.13.mlp.up_proj.weight": "model.safetensors",
|
| 149 |
+
"language_model.model.layers.13.post_attention_layernorm.weight": "model.safetensors",
|
| 150 |
+
"language_model.model.layers.13.self_attn.k_proj.bias": "model.safetensors",
|
| 151 |
+
"language_model.model.layers.13.self_attn.k_proj.biases": "model.safetensors",
|
| 152 |
+
"language_model.model.layers.13.self_attn.k_proj.scales": "model.safetensors",
|
| 153 |
+
"language_model.model.layers.13.self_attn.k_proj.weight": "model.safetensors",
|
| 154 |
+
"language_model.model.layers.13.self_attn.o_proj.biases": "model.safetensors",
|
| 155 |
+
"language_model.model.layers.13.self_attn.o_proj.scales": "model.safetensors",
|
| 156 |
+
"language_model.model.layers.13.self_attn.o_proj.weight": "model.safetensors",
|
| 157 |
+
"language_model.model.layers.13.self_attn.q_proj.bias": "model.safetensors",
|
| 158 |
+
"language_model.model.layers.13.self_attn.q_proj.biases": "model.safetensors",
|
| 159 |
+
"language_model.model.layers.13.self_attn.q_proj.scales": "model.safetensors",
|
| 160 |
+
"language_model.model.layers.13.self_attn.q_proj.weight": "model.safetensors",
|
| 161 |
+
"language_model.model.layers.13.self_attn.v_proj.bias": "model.safetensors",
|
| 162 |
+
"language_model.model.layers.13.self_attn.v_proj.biases": "model.safetensors",
|
| 163 |
+
"language_model.model.layers.13.self_attn.v_proj.scales": "model.safetensors",
|
| 164 |
+
"language_model.model.layers.13.self_attn.v_proj.weight": "model.safetensors",
|
| 165 |
+
"language_model.model.layers.14.input_layernorm.weight": "model.safetensors",
|
| 166 |
+
"language_model.model.layers.14.mlp.down_proj.biases": "model.safetensors",
|
| 167 |
+
"language_model.model.layers.14.mlp.down_proj.scales": "model.safetensors",
|
| 168 |
+
"language_model.model.layers.14.mlp.down_proj.weight": "model.safetensors",
|
| 169 |
+
"language_model.model.layers.14.mlp.gate_proj.biases": "model.safetensors",
|
| 170 |
+
"language_model.model.layers.14.mlp.gate_proj.scales": "model.safetensors",
|
| 171 |
+
"language_model.model.layers.14.mlp.gate_proj.weight": "model.safetensors",
|
| 172 |
+
"language_model.model.layers.14.mlp.up_proj.biases": "model.safetensors",
|
| 173 |
+
"language_model.model.layers.14.mlp.up_proj.scales": "model.safetensors",
|
| 174 |
+
"language_model.model.layers.14.mlp.up_proj.weight": "model.safetensors",
|
| 175 |
+
"language_model.model.layers.14.post_attention_layernorm.weight": "model.safetensors",
|
| 176 |
+
"language_model.model.layers.14.self_attn.k_proj.bias": "model.safetensors",
|
| 177 |
+
"language_model.model.layers.14.self_attn.k_proj.biases": "model.safetensors",
|
| 178 |
+
"language_model.model.layers.14.self_attn.k_proj.scales": "model.safetensors",
|
| 179 |
+
"language_model.model.layers.14.self_attn.k_proj.weight": "model.safetensors",
|
| 180 |
+
"language_model.model.layers.14.self_attn.o_proj.biases": "model.safetensors",
|
| 181 |
+
"language_model.model.layers.14.self_attn.o_proj.scales": "model.safetensors",
|
| 182 |
+
"language_model.model.layers.14.self_attn.o_proj.weight": "model.safetensors",
|
| 183 |
+
"language_model.model.layers.14.self_attn.q_proj.bias": "model.safetensors",
|
| 184 |
+
"language_model.model.layers.14.self_attn.q_proj.biases": "model.safetensors",
|
| 185 |
+
"language_model.model.layers.14.self_attn.q_proj.scales": "model.safetensors",
|
| 186 |
+
"language_model.model.layers.14.self_attn.q_proj.weight": "model.safetensors",
|
| 187 |
+
"language_model.model.layers.14.self_attn.v_proj.bias": "model.safetensors",
|
| 188 |
+
"language_model.model.layers.14.self_attn.v_proj.biases": "model.safetensors",
|
| 189 |
+
"language_model.model.layers.14.self_attn.v_proj.scales": "model.safetensors",
|
| 190 |
+
"language_model.model.layers.14.self_attn.v_proj.weight": "model.safetensors",
|
| 191 |
+
"language_model.model.layers.15.input_layernorm.weight": "model.safetensors",
|
| 192 |
+
"language_model.model.layers.15.mlp.down_proj.biases": "model.safetensors",
|
| 193 |
+
"language_model.model.layers.15.mlp.down_proj.scales": "model.safetensors",
|
| 194 |
+
"language_model.model.layers.15.mlp.down_proj.weight": "model.safetensors",
|
| 195 |
+
"language_model.model.layers.15.mlp.gate_proj.biases": "model.safetensors",
|
| 196 |
+
"language_model.model.layers.15.mlp.gate_proj.scales": "model.safetensors",
|
| 197 |
+
"language_model.model.layers.15.mlp.gate_proj.weight": "model.safetensors",
|
| 198 |
+
"language_model.model.layers.15.mlp.up_proj.biases": "model.safetensors",
|
| 199 |
+
"language_model.model.layers.15.mlp.up_proj.scales": "model.safetensors",
|
| 200 |
+
"language_model.model.layers.15.mlp.up_proj.weight": "model.safetensors",
|
| 201 |
+
"language_model.model.layers.15.post_attention_layernorm.weight": "model.safetensors",
|
| 202 |
+
"language_model.model.layers.15.self_attn.k_proj.bias": "model.safetensors",
|
| 203 |
+
"language_model.model.layers.15.self_attn.k_proj.biases": "model.safetensors",
|
| 204 |
+
"language_model.model.layers.15.self_attn.k_proj.scales": "model.safetensors",
|
| 205 |
+
"language_model.model.layers.15.self_attn.k_proj.weight": "model.safetensors",
|
| 206 |
+
"language_model.model.layers.15.self_attn.o_proj.biases": "model.safetensors",
|
| 207 |
+
"language_model.model.layers.15.self_attn.o_proj.scales": "model.safetensors",
|
| 208 |
+
"language_model.model.layers.15.self_attn.o_proj.weight": "model.safetensors",
|
| 209 |
+
"language_model.model.layers.15.self_attn.q_proj.bias": "model.safetensors",
|
| 210 |
+
"language_model.model.layers.15.self_attn.q_proj.biases": "model.safetensors",
|
| 211 |
+
"language_model.model.layers.15.self_attn.q_proj.scales": "model.safetensors",
|
| 212 |
+
"language_model.model.layers.15.self_attn.q_proj.weight": "model.safetensors",
|
| 213 |
+
"language_model.model.layers.15.self_attn.v_proj.bias": "model.safetensors",
|
| 214 |
+
"language_model.model.layers.15.self_attn.v_proj.biases": "model.safetensors",
|
| 215 |
+
"language_model.model.layers.15.self_attn.v_proj.scales": "model.safetensors",
|
| 216 |
+
"language_model.model.layers.15.self_attn.v_proj.weight": "model.safetensors",
|
| 217 |
+
"language_model.model.layers.16.input_layernorm.weight": "model.safetensors",
|
| 218 |
+
"language_model.model.layers.16.mlp.down_proj.biases": "model.safetensors",
|
| 219 |
+
"language_model.model.layers.16.mlp.down_proj.scales": "model.safetensors",
|
| 220 |
+
"language_model.model.layers.16.mlp.down_proj.weight": "model.safetensors",
|
| 221 |
+
"language_model.model.layers.16.mlp.gate_proj.biases": "model.safetensors",
|
| 222 |
+
"language_model.model.layers.16.mlp.gate_proj.scales": "model.safetensors",
|
| 223 |
+
"language_model.model.layers.16.mlp.gate_proj.weight": "model.safetensors",
|
| 224 |
+
"language_model.model.layers.16.mlp.up_proj.biases": "model.safetensors",
|
| 225 |
+
"language_model.model.layers.16.mlp.up_proj.scales": "model.safetensors",
|
| 226 |
+
"language_model.model.layers.16.mlp.up_proj.weight": "model.safetensors",
|
| 227 |
+
"language_model.model.layers.16.post_attention_layernorm.weight": "model.safetensors",
|
| 228 |
+
"language_model.model.layers.16.self_attn.k_proj.bias": "model.safetensors",
|
| 229 |
+
"language_model.model.layers.16.self_attn.k_proj.biases": "model.safetensors",
|
| 230 |
+
"language_model.model.layers.16.self_attn.k_proj.scales": "model.safetensors",
|
| 231 |
+
"language_model.model.layers.16.self_attn.k_proj.weight": "model.safetensors",
|
| 232 |
+
"language_model.model.layers.16.self_attn.o_proj.biases": "model.safetensors",
|
| 233 |
+
"language_model.model.layers.16.self_attn.o_proj.scales": "model.safetensors",
|
| 234 |
+
"language_model.model.layers.16.self_attn.o_proj.weight": "model.safetensors",
|
| 235 |
+
"language_model.model.layers.16.self_attn.q_proj.bias": "model.safetensors",
|
| 236 |
+
"language_model.model.layers.16.self_attn.q_proj.biases": "model.safetensors",
|
| 237 |
+
"language_model.model.layers.16.self_attn.q_proj.scales": "model.safetensors",
|
| 238 |
+
"language_model.model.layers.16.self_attn.q_proj.weight": "model.safetensors",
|
| 239 |
+
"language_model.model.layers.16.self_attn.v_proj.bias": "model.safetensors",
|
| 240 |
+
"language_model.model.layers.16.self_attn.v_proj.biases": "model.safetensors",
|
| 241 |
+
"language_model.model.layers.16.self_attn.v_proj.scales": "model.safetensors",
|
| 242 |
+
"language_model.model.layers.16.self_attn.v_proj.weight": "model.safetensors",
|
| 243 |
+
"language_model.model.layers.17.input_layernorm.weight": "model.safetensors",
|
| 244 |
+
"language_model.model.layers.17.mlp.down_proj.biases": "model.safetensors",
|
| 245 |
+
"language_model.model.layers.17.mlp.down_proj.scales": "model.safetensors",
|
| 246 |
+
"language_model.model.layers.17.mlp.down_proj.weight": "model.safetensors",
|
| 247 |
+
"language_model.model.layers.17.mlp.gate_proj.biases": "model.safetensors",
|
| 248 |
+
"language_model.model.layers.17.mlp.gate_proj.scales": "model.safetensors",
|
| 249 |
+
"language_model.model.layers.17.mlp.gate_proj.weight": "model.safetensors",
|
| 250 |
+
"language_model.model.layers.17.mlp.up_proj.biases": "model.safetensors",
|
| 251 |
+
"language_model.model.layers.17.mlp.up_proj.scales": "model.safetensors",
|
| 252 |
+
"language_model.model.layers.17.mlp.up_proj.weight": "model.safetensors",
|
| 253 |
+
"language_model.model.layers.17.post_attention_layernorm.weight": "model.safetensors",
|
| 254 |
+
"language_model.model.layers.17.self_attn.k_proj.bias": "model.safetensors",
|
| 255 |
+
"language_model.model.layers.17.self_attn.k_proj.biases": "model.safetensors",
|
| 256 |
+
"language_model.model.layers.17.self_attn.k_proj.scales": "model.safetensors",
|
| 257 |
+
"language_model.model.layers.17.self_attn.k_proj.weight": "model.safetensors",
|
| 258 |
+
"language_model.model.layers.17.self_attn.o_proj.biases": "model.safetensors",
|
| 259 |
+
"language_model.model.layers.17.self_attn.o_proj.scales": "model.safetensors",
|
| 260 |
+
"language_model.model.layers.17.self_attn.o_proj.weight": "model.safetensors",
|
| 261 |
+
"language_model.model.layers.17.self_attn.q_proj.bias": "model.safetensors",
|
| 262 |
+
"language_model.model.layers.17.self_attn.q_proj.biases": "model.safetensors",
|
| 263 |
+
"language_model.model.layers.17.self_attn.q_proj.scales": "model.safetensors",
|
| 264 |
+
"language_model.model.layers.17.self_attn.q_proj.weight": "model.safetensors",
|
| 265 |
+
"language_model.model.layers.17.self_attn.v_proj.bias": "model.safetensors",
|
| 266 |
+
"language_model.model.layers.17.self_attn.v_proj.biases": "model.safetensors",
|
| 267 |
+
"language_model.model.layers.17.self_attn.v_proj.scales": "model.safetensors",
|
| 268 |
+
"language_model.model.layers.17.self_attn.v_proj.weight": "model.safetensors",
|
| 269 |
+
"language_model.model.layers.18.input_layernorm.weight": "model.safetensors",
|
| 270 |
+
"language_model.model.layers.18.mlp.down_proj.biases": "model.safetensors",
|
| 271 |
+
"language_model.model.layers.18.mlp.down_proj.scales": "model.safetensors",
|
| 272 |
+
"language_model.model.layers.18.mlp.down_proj.weight": "model.safetensors",
|
| 273 |
+
"language_model.model.layers.18.mlp.gate_proj.biases": "model.safetensors",
|
| 274 |
+
"language_model.model.layers.18.mlp.gate_proj.scales": "model.safetensors",
|
| 275 |
+
"language_model.model.layers.18.mlp.gate_proj.weight": "model.safetensors",
|
| 276 |
+
"language_model.model.layers.18.mlp.up_proj.biases": "model.safetensors",
|
| 277 |
+
"language_model.model.layers.18.mlp.up_proj.scales": "model.safetensors",
|
| 278 |
+
"language_model.model.layers.18.mlp.up_proj.weight": "model.safetensors",
|
| 279 |
+
"language_model.model.layers.18.post_attention_layernorm.weight": "model.safetensors",
|
| 280 |
+
"language_model.model.layers.18.self_attn.k_proj.bias": "model.safetensors",
|
| 281 |
+
"language_model.model.layers.18.self_attn.k_proj.biases": "model.safetensors",
|
| 282 |
+
"language_model.model.layers.18.self_attn.k_proj.scales": "model.safetensors",
|
| 283 |
+
"language_model.model.layers.18.self_attn.k_proj.weight": "model.safetensors",
|
| 284 |
+
"language_model.model.layers.18.self_attn.o_proj.biases": "model.safetensors",
|
| 285 |
+
"language_model.model.layers.18.self_attn.o_proj.scales": "model.safetensors",
|
| 286 |
+
"language_model.model.layers.18.self_attn.o_proj.weight": "model.safetensors",
|
| 287 |
+
"language_model.model.layers.18.self_attn.q_proj.bias": "model.safetensors",
|
| 288 |
+
"language_model.model.layers.18.self_attn.q_proj.biases": "model.safetensors",
|
| 289 |
+
"language_model.model.layers.18.self_attn.q_proj.scales": "model.safetensors",
|
| 290 |
+
"language_model.model.layers.18.self_attn.q_proj.weight": "model.safetensors",
|
| 291 |
+
"language_model.model.layers.18.self_attn.v_proj.bias": "model.safetensors",
|
| 292 |
+
"language_model.model.layers.18.self_attn.v_proj.biases": "model.safetensors",
|
| 293 |
+
"language_model.model.layers.18.self_attn.v_proj.scales": "model.safetensors",
|
| 294 |
+
"language_model.model.layers.18.self_attn.v_proj.weight": "model.safetensors",
|
| 295 |
+
"language_model.model.layers.19.input_layernorm.weight": "model.safetensors",
|
| 296 |
+
"language_model.model.layers.19.mlp.down_proj.biases": "model.safetensors",
|
| 297 |
+
"language_model.model.layers.19.mlp.down_proj.scales": "model.safetensors",
|
| 298 |
+
"language_model.model.layers.19.mlp.down_proj.weight": "model.safetensors",
|
| 299 |
+
"language_model.model.layers.19.mlp.gate_proj.biases": "model.safetensors",
|
| 300 |
+
"language_model.model.layers.19.mlp.gate_proj.scales": "model.safetensors",
|
| 301 |
+
"language_model.model.layers.19.mlp.gate_proj.weight": "model.safetensors",
|
| 302 |
+
"language_model.model.layers.19.mlp.up_proj.biases": "model.safetensors",
|
| 303 |
+
"language_model.model.layers.19.mlp.up_proj.scales": "model.safetensors",
|
| 304 |
+
"language_model.model.layers.19.mlp.up_proj.weight": "model.safetensors",
|
| 305 |
+
"language_model.model.layers.19.post_attention_layernorm.weight": "model.safetensors",
|
| 306 |
+
"language_model.model.layers.19.self_attn.k_proj.bias": "model.safetensors",
|
| 307 |
+
"language_model.model.layers.19.self_attn.k_proj.biases": "model.safetensors",
|
| 308 |
+
"language_model.model.layers.19.self_attn.k_proj.scales": "model.safetensors",
|
| 309 |
+
"language_model.model.layers.19.self_attn.k_proj.weight": "model.safetensors",
|
| 310 |
+
"language_model.model.layers.19.self_attn.o_proj.biases": "model.safetensors",
|
| 311 |
+
"language_model.model.layers.19.self_attn.o_proj.scales": "model.safetensors",
|
| 312 |
+
"language_model.model.layers.19.self_attn.o_proj.weight": "model.safetensors",
|
| 313 |
+
"language_model.model.layers.19.self_attn.q_proj.bias": "model.safetensors",
|
| 314 |
+
"language_model.model.layers.19.self_attn.q_proj.biases": "model.safetensors",
|
| 315 |
+
"language_model.model.layers.19.self_attn.q_proj.scales": "model.safetensors",
|
| 316 |
+
"language_model.model.layers.19.self_attn.q_proj.weight": "model.safetensors",
|
| 317 |
+
"language_model.model.layers.19.self_attn.v_proj.bias": "model.safetensors",
|
| 318 |
+
"language_model.model.layers.19.self_attn.v_proj.biases": "model.safetensors",
|
| 319 |
+
"language_model.model.layers.19.self_attn.v_proj.scales": "model.safetensors",
|
| 320 |
+
"language_model.model.layers.19.self_attn.v_proj.weight": "model.safetensors",
|
| 321 |
+
"language_model.model.layers.2.input_layernorm.weight": "model.safetensors",
|
| 322 |
+
"language_model.model.layers.2.mlp.down_proj.biases": "model.safetensors",
|
| 323 |
+
"language_model.model.layers.2.mlp.down_proj.scales": "model.safetensors",
|
| 324 |
+
"language_model.model.layers.2.mlp.down_proj.weight": "model.safetensors",
|
| 325 |
+
"language_model.model.layers.2.mlp.gate_proj.biases": "model.safetensors",
|
| 326 |
+
"language_model.model.layers.2.mlp.gate_proj.scales": "model.safetensors",
|
| 327 |
+
"language_model.model.layers.2.mlp.gate_proj.weight": "model.safetensors",
|
| 328 |
+
"language_model.model.layers.2.mlp.up_proj.biases": "model.safetensors",
|
| 329 |
+
"language_model.model.layers.2.mlp.up_proj.scales": "model.safetensors",
|
| 330 |
+
"language_model.model.layers.2.mlp.up_proj.weight": "model.safetensors",
|
| 331 |
+
"language_model.model.layers.2.post_attention_layernorm.weight": "model.safetensors",
|
| 332 |
+
"language_model.model.layers.2.self_attn.k_proj.bias": "model.safetensors",
|
| 333 |
+
"language_model.model.layers.2.self_attn.k_proj.biases": "model.safetensors",
|
| 334 |
+
"language_model.model.layers.2.self_attn.k_proj.scales": "model.safetensors",
|
| 335 |
+
"language_model.model.layers.2.self_attn.k_proj.weight": "model.safetensors",
|
| 336 |
+
"language_model.model.layers.2.self_attn.o_proj.biases": "model.safetensors",
|
| 337 |
+
"language_model.model.layers.2.self_attn.o_proj.scales": "model.safetensors",
|
| 338 |
+
"language_model.model.layers.2.self_attn.o_proj.weight": "model.safetensors",
|
| 339 |
+
"language_model.model.layers.2.self_attn.q_proj.bias": "model.safetensors",
|
| 340 |
+
"language_model.model.layers.2.self_attn.q_proj.biases": "model.safetensors",
|
| 341 |
+
"language_model.model.layers.2.self_attn.q_proj.scales": "model.safetensors",
|
| 342 |
+
"language_model.model.layers.2.self_attn.q_proj.weight": "model.safetensors",
|
| 343 |
+
"language_model.model.layers.2.self_attn.v_proj.bias": "model.safetensors",
|
| 344 |
+
"language_model.model.layers.2.self_attn.v_proj.biases": "model.safetensors",
|
| 345 |
+
"language_model.model.layers.2.self_attn.v_proj.scales": "model.safetensors",
|
| 346 |
+
"language_model.model.layers.2.self_attn.v_proj.weight": "model.safetensors",
|
| 347 |
+
"language_model.model.layers.20.input_layernorm.weight": "model.safetensors",
|
| 348 |
+
"language_model.model.layers.20.mlp.down_proj.biases": "model.safetensors",
|
| 349 |
+
"language_model.model.layers.20.mlp.down_proj.scales": "model.safetensors",
|
| 350 |
+
"language_model.model.layers.20.mlp.down_proj.weight": "model.safetensors",
|
| 351 |
+
"language_model.model.layers.20.mlp.gate_proj.biases": "model.safetensors",
|
| 352 |
+
"language_model.model.layers.20.mlp.gate_proj.scales": "model.safetensors",
|
| 353 |
+
"language_model.model.layers.20.mlp.gate_proj.weight": "model.safetensors",
|
| 354 |
+
"language_model.model.layers.20.mlp.up_proj.biases": "model.safetensors",
|
| 355 |
+
"language_model.model.layers.20.mlp.up_proj.scales": "model.safetensors",
|
| 356 |
+
"language_model.model.layers.20.mlp.up_proj.weight": "model.safetensors",
|
| 357 |
+
"language_model.model.layers.20.post_attention_layernorm.weight": "model.safetensors",
|
| 358 |
+
"language_model.model.layers.20.self_attn.k_proj.bias": "model.safetensors",
|
| 359 |
+
"language_model.model.layers.20.self_attn.k_proj.biases": "model.safetensors",
|
| 360 |
+
"language_model.model.layers.20.self_attn.k_proj.scales": "model.safetensors",
|
| 361 |
+
"language_model.model.layers.20.self_attn.k_proj.weight": "model.safetensors",
|
| 362 |
+
"language_model.model.layers.20.self_attn.o_proj.biases": "model.safetensors",
|
| 363 |
+
"language_model.model.layers.20.self_attn.o_proj.scales": "model.safetensors",
|
| 364 |
+
"language_model.model.layers.20.self_attn.o_proj.weight": "model.safetensors",
|
| 365 |
+
"language_model.model.layers.20.self_attn.q_proj.bias": "model.safetensors",
|
| 366 |
+
"language_model.model.layers.20.self_attn.q_proj.biases": "model.safetensors",
|
| 367 |
+
"language_model.model.layers.20.self_attn.q_proj.scales": "model.safetensors",
|
| 368 |
+
"language_model.model.layers.20.self_attn.q_proj.weight": "model.safetensors",
|
| 369 |
+
"language_model.model.layers.20.self_attn.v_proj.bias": "model.safetensors",
|
| 370 |
+
"language_model.model.layers.20.self_attn.v_proj.biases": "model.safetensors",
|
| 371 |
+
"language_model.model.layers.20.self_attn.v_proj.scales": "model.safetensors",
|
| 372 |
+
"language_model.model.layers.20.self_attn.v_proj.weight": "model.safetensors",
|
| 373 |
+
"language_model.model.layers.21.input_layernorm.weight": "model.safetensors",
|
| 374 |
+
"language_model.model.layers.21.mlp.down_proj.biases": "model.safetensors",
|
| 375 |
+
"language_model.model.layers.21.mlp.down_proj.scales": "model.safetensors",
|
| 376 |
+
"language_model.model.layers.21.mlp.down_proj.weight": "model.safetensors",
|
| 377 |
+
"language_model.model.layers.21.mlp.gate_proj.biases": "model.safetensors",
|
| 378 |
+
"language_model.model.layers.21.mlp.gate_proj.scales": "model.safetensors",
|
| 379 |
+
"language_model.model.layers.21.mlp.gate_proj.weight": "model.safetensors",
|
| 380 |
+
"language_model.model.layers.21.mlp.up_proj.biases": "model.safetensors",
|
| 381 |
+
"language_model.model.layers.21.mlp.up_proj.scales": "model.safetensors",
|
| 382 |
+
"language_model.model.layers.21.mlp.up_proj.weight": "model.safetensors",
|
| 383 |
+
"language_model.model.layers.21.post_attention_layernorm.weight": "model.safetensors",
|
| 384 |
+
"language_model.model.layers.21.self_attn.k_proj.bias": "model.safetensors",
|
| 385 |
+
"language_model.model.layers.21.self_attn.k_proj.biases": "model.safetensors",
|
| 386 |
+
"language_model.model.layers.21.self_attn.k_proj.scales": "model.safetensors",
|
| 387 |
+
"language_model.model.layers.21.self_attn.k_proj.weight": "model.safetensors",
|
| 388 |
+
"language_model.model.layers.21.self_attn.o_proj.biases": "model.safetensors",
|
| 389 |
+
"language_model.model.layers.21.self_attn.o_proj.scales": "model.safetensors",
|
| 390 |
+
"language_model.model.layers.21.self_attn.o_proj.weight": "model.safetensors",
|
| 391 |
+
"language_model.model.layers.21.self_attn.q_proj.bias": "model.safetensors",
|
| 392 |
+
"language_model.model.layers.21.self_attn.q_proj.biases": "model.safetensors",
|
| 393 |
+
"language_model.model.layers.21.self_attn.q_proj.scales": "model.safetensors",
|
| 394 |
+
"language_model.model.layers.21.self_attn.q_proj.weight": "model.safetensors",
|
| 395 |
+
"language_model.model.layers.21.self_attn.v_proj.bias": "model.safetensors",
|
| 396 |
+
"language_model.model.layers.21.self_attn.v_proj.biases": "model.safetensors",
|
| 397 |
+
"language_model.model.layers.21.self_attn.v_proj.scales": "model.safetensors",
|
| 398 |
+
"language_model.model.layers.21.self_attn.v_proj.weight": "model.safetensors",
|
| 399 |
+
"language_model.model.layers.22.input_layernorm.weight": "model.safetensors",
|
| 400 |
+
"language_model.model.layers.22.mlp.down_proj.biases": "model.safetensors",
|
| 401 |
+
"language_model.model.layers.22.mlp.down_proj.scales": "model.safetensors",
|
| 402 |
+
"language_model.model.layers.22.mlp.down_proj.weight": "model.safetensors",
|
| 403 |
+
"language_model.model.layers.22.mlp.gate_proj.biases": "model.safetensors",
|
| 404 |
+
"language_model.model.layers.22.mlp.gate_proj.scales": "model.safetensors",
|
| 405 |
+
"language_model.model.layers.22.mlp.gate_proj.weight": "model.safetensors",
|
| 406 |
+
"language_model.model.layers.22.mlp.up_proj.biases": "model.safetensors",
|
| 407 |
+
"language_model.model.layers.22.mlp.up_proj.scales": "model.safetensors",
|
| 408 |
+
"language_model.model.layers.22.mlp.up_proj.weight": "model.safetensors",
|
| 409 |
+
"language_model.model.layers.22.post_attention_layernorm.weight": "model.safetensors",
|
| 410 |
+
"language_model.model.layers.22.self_attn.k_proj.bias": "model.safetensors",
|
| 411 |
+
"language_model.model.layers.22.self_attn.k_proj.biases": "model.safetensors",
|
| 412 |
+
"language_model.model.layers.22.self_attn.k_proj.scales": "model.safetensors",
|
| 413 |
+
"language_model.model.layers.22.self_attn.k_proj.weight": "model.safetensors",
|
| 414 |
+
"language_model.model.layers.22.self_attn.o_proj.biases": "model.safetensors",
|
| 415 |
+
"language_model.model.layers.22.self_attn.o_proj.scales": "model.safetensors",
|
| 416 |
+
"language_model.model.layers.22.self_attn.o_proj.weight": "model.safetensors",
|
| 417 |
+
"language_model.model.layers.22.self_attn.q_proj.bias": "model.safetensors",
|
| 418 |
+
"language_model.model.layers.22.self_attn.q_proj.biases": "model.safetensors",
|
| 419 |
+
"language_model.model.layers.22.self_attn.q_proj.scales": "model.safetensors",
|
| 420 |
+
"language_model.model.layers.22.self_attn.q_proj.weight": "model.safetensors",
|
| 421 |
+
"language_model.model.layers.22.self_attn.v_proj.bias": "model.safetensors",
|
| 422 |
+
"language_model.model.layers.22.self_attn.v_proj.biases": "model.safetensors",
|
| 423 |
+
"language_model.model.layers.22.self_attn.v_proj.scales": "model.safetensors",
|
| 424 |
+
"language_model.model.layers.22.self_attn.v_proj.weight": "model.safetensors",
|
| 425 |
+
"language_model.model.layers.23.input_layernorm.weight": "model.safetensors",
|
| 426 |
+
"language_model.model.layers.23.mlp.down_proj.biases": "model.safetensors",
|
| 427 |
+
"language_model.model.layers.23.mlp.down_proj.scales": "model.safetensors",
|
| 428 |
+
"language_model.model.layers.23.mlp.down_proj.weight": "model.safetensors",
|
| 429 |
+
"language_model.model.layers.23.mlp.gate_proj.biases": "model.safetensors",
|
| 430 |
+
"language_model.model.layers.23.mlp.gate_proj.scales": "model.safetensors",
|
| 431 |
+
"language_model.model.layers.23.mlp.gate_proj.weight": "model.safetensors",
|
| 432 |
+
"language_model.model.layers.23.mlp.up_proj.biases": "model.safetensors",
|
| 433 |
+
"language_model.model.layers.23.mlp.up_proj.scales": "model.safetensors",
|
| 434 |
+
"language_model.model.layers.23.mlp.up_proj.weight": "model.safetensors",
|
| 435 |
+
"language_model.model.layers.23.post_attention_layernorm.weight": "model.safetensors",
|
| 436 |
+
"language_model.model.layers.23.self_attn.k_proj.bias": "model.safetensors",
|
| 437 |
+
"language_model.model.layers.23.self_attn.k_proj.biases": "model.safetensors",
|
| 438 |
+
"language_model.model.layers.23.self_attn.k_proj.scales": "model.safetensors",
|
| 439 |
+
"language_model.model.layers.23.self_attn.k_proj.weight": "model.safetensors",
|
| 440 |
+
"language_model.model.layers.23.self_attn.o_proj.biases": "model.safetensors",
|
| 441 |
+
"language_model.model.layers.23.self_attn.o_proj.scales": "model.safetensors",
|
| 442 |
+
"language_model.model.layers.23.self_attn.o_proj.weight": "model.safetensors",
|
| 443 |
+
"language_model.model.layers.23.self_attn.q_proj.bias": "model.safetensors",
|
| 444 |
+
"language_model.model.layers.23.self_attn.q_proj.biases": "model.safetensors",
|
| 445 |
+
"language_model.model.layers.23.self_attn.q_proj.scales": "model.safetensors",
|
| 446 |
+
"language_model.model.layers.23.self_attn.q_proj.weight": "model.safetensors",
|
| 447 |
+
"language_model.model.layers.23.self_attn.v_proj.bias": "model.safetensors",
|
| 448 |
+
"language_model.model.layers.23.self_attn.v_proj.biases": "model.safetensors",
|
| 449 |
+
"language_model.model.layers.23.self_attn.v_proj.scales": "model.safetensors",
|
| 450 |
+
"language_model.model.layers.23.self_attn.v_proj.weight": "model.safetensors",
|
| 451 |
+
"language_model.model.layers.24.input_layernorm.weight": "model.safetensors",
|
| 452 |
+
"language_model.model.layers.24.mlp.down_proj.biases": "model.safetensors",
|
| 453 |
+
"language_model.model.layers.24.mlp.down_proj.scales": "model.safetensors",
|
| 454 |
+
"language_model.model.layers.24.mlp.down_proj.weight": "model.safetensors",
|
| 455 |
+
"language_model.model.layers.24.mlp.gate_proj.biases": "model.safetensors",
|
| 456 |
+
"language_model.model.layers.24.mlp.gate_proj.scales": "model.safetensors",
|
| 457 |
+
"language_model.model.layers.24.mlp.gate_proj.weight": "model.safetensors",
|
| 458 |
+
"language_model.model.layers.24.mlp.up_proj.biases": "model.safetensors",
|
| 459 |
+
"language_model.model.layers.24.mlp.up_proj.scales": "model.safetensors",
|
| 460 |
+
"language_model.model.layers.24.mlp.up_proj.weight": "model.safetensors",
|
| 461 |
+
"language_model.model.layers.24.post_attention_layernorm.weight": "model.safetensors",
|
| 462 |
+
"language_model.model.layers.24.self_attn.k_proj.bias": "model.safetensors",
|
| 463 |
+
"language_model.model.layers.24.self_attn.k_proj.biases": "model.safetensors",
|
| 464 |
+
"language_model.model.layers.24.self_attn.k_proj.scales": "model.safetensors",
|
| 465 |
+
"language_model.model.layers.24.self_attn.k_proj.weight": "model.safetensors",
|
| 466 |
+
"language_model.model.layers.24.self_attn.o_proj.biases": "model.safetensors",
|
| 467 |
+
"language_model.model.layers.24.self_attn.o_proj.scales": "model.safetensors",
|
| 468 |
+
"language_model.model.layers.24.self_attn.o_proj.weight": "model.safetensors",
|
| 469 |
+
"language_model.model.layers.24.self_attn.q_proj.bias": "model.safetensors",
|
| 470 |
+
"language_model.model.layers.24.self_attn.q_proj.biases": "model.safetensors",
|
| 471 |
+
"language_model.model.layers.24.self_attn.q_proj.scales": "model.safetensors",
|
| 472 |
+
"language_model.model.layers.24.self_attn.q_proj.weight": "model.safetensors",
|
| 473 |
+
"language_model.model.layers.24.self_attn.v_proj.bias": "model.safetensors",
|
| 474 |
+
"language_model.model.layers.24.self_attn.v_proj.biases": "model.safetensors",
|
| 475 |
+
"language_model.model.layers.24.self_attn.v_proj.scales": "model.safetensors",
|
| 476 |
+
"language_model.model.layers.24.self_attn.v_proj.weight": "model.safetensors",
|
| 477 |
+
"language_model.model.layers.25.input_layernorm.weight": "model.safetensors",
|
| 478 |
+
"language_model.model.layers.25.mlp.down_proj.biases": "model.safetensors",
|
| 479 |
+
"language_model.model.layers.25.mlp.down_proj.scales": "model.safetensors",
|
| 480 |
+
"language_model.model.layers.25.mlp.down_proj.weight": "model.safetensors",
|
| 481 |
+
"language_model.model.layers.25.mlp.gate_proj.biases": "model.safetensors",
|
| 482 |
+
"language_model.model.layers.25.mlp.gate_proj.scales": "model.safetensors",
|
| 483 |
+
"language_model.model.layers.25.mlp.gate_proj.weight": "model.safetensors",
|
| 484 |
+
"language_model.model.layers.25.mlp.up_proj.biases": "model.safetensors",
|
| 485 |
+
"language_model.model.layers.25.mlp.up_proj.scales": "model.safetensors",
|
| 486 |
+
"language_model.model.layers.25.mlp.up_proj.weight": "model.safetensors",
|
| 487 |
+
"language_model.model.layers.25.post_attention_layernorm.weight": "model.safetensors",
|
| 488 |
+
"language_model.model.layers.25.self_attn.k_proj.bias": "model.safetensors",
|
| 489 |
+
"language_model.model.layers.25.self_attn.k_proj.biases": "model.safetensors",
|
| 490 |
+
"language_model.model.layers.25.self_attn.k_proj.scales": "model.safetensors",
|
| 491 |
+
"language_model.model.layers.25.self_attn.k_proj.weight": "model.safetensors",
|
| 492 |
+
"language_model.model.layers.25.self_attn.o_proj.biases": "model.safetensors",
|
| 493 |
+
"language_model.model.layers.25.self_attn.o_proj.scales": "model.safetensors",
|
| 494 |
+
"language_model.model.layers.25.self_attn.o_proj.weight": "model.safetensors",
|
| 495 |
+
"language_model.model.layers.25.self_attn.q_proj.bias": "model.safetensors",
|
| 496 |
+
"language_model.model.layers.25.self_attn.q_proj.biases": "model.safetensors",
|
| 497 |
+
"language_model.model.layers.25.self_attn.q_proj.scales": "model.safetensors",
|
| 498 |
+
"language_model.model.layers.25.self_attn.q_proj.weight": "model.safetensors",
|
| 499 |
+
"language_model.model.layers.25.self_attn.v_proj.bias": "model.safetensors",
|
| 500 |
+
"language_model.model.layers.25.self_attn.v_proj.biases": "model.safetensors",
|
| 501 |
+
"language_model.model.layers.25.self_attn.v_proj.scales": "model.safetensors",
|
| 502 |
+
"language_model.model.layers.25.self_attn.v_proj.weight": "model.safetensors",
|
| 503 |
+
"language_model.model.layers.26.input_layernorm.weight": "model.safetensors",
|
| 504 |
+
"language_model.model.layers.26.mlp.down_proj.biases": "model.safetensors",
|
| 505 |
+
"language_model.model.layers.26.mlp.down_proj.scales": "model.safetensors",
|
| 506 |
+
"language_model.model.layers.26.mlp.down_proj.weight": "model.safetensors",
|
| 507 |
+
"language_model.model.layers.26.mlp.gate_proj.biases": "model.safetensors",
|
| 508 |
+
"language_model.model.layers.26.mlp.gate_proj.scales": "model.safetensors",
|
| 509 |
+
"language_model.model.layers.26.mlp.gate_proj.weight": "model.safetensors",
|
| 510 |
+
"language_model.model.layers.26.mlp.up_proj.biases": "model.safetensors",
|
| 511 |
+
"language_model.model.layers.26.mlp.up_proj.scales": "model.safetensors",
|
| 512 |
+
"language_model.model.layers.26.mlp.up_proj.weight": "model.safetensors",
|
| 513 |
+
"language_model.model.layers.26.post_attention_layernorm.weight": "model.safetensors",
|
| 514 |
+
"language_model.model.layers.26.self_attn.k_proj.bias": "model.safetensors",
|
| 515 |
+
"language_model.model.layers.26.self_attn.k_proj.biases": "model.safetensors",
|
| 516 |
+
"language_model.model.layers.26.self_attn.k_proj.scales": "model.safetensors",
|
| 517 |
+
"language_model.model.layers.26.self_attn.k_proj.weight": "model.safetensors",
|
| 518 |
+
"language_model.model.layers.26.self_attn.o_proj.biases": "model.safetensors",
|
| 519 |
+
"language_model.model.layers.26.self_attn.o_proj.scales": "model.safetensors",
|
| 520 |
+
"language_model.model.layers.26.self_attn.o_proj.weight": "model.safetensors",
|
| 521 |
+
"language_model.model.layers.26.self_attn.q_proj.bias": "model.safetensors",
|
| 522 |
+
"language_model.model.layers.26.self_attn.q_proj.biases": "model.safetensors",
|
| 523 |
+
"language_model.model.layers.26.self_attn.q_proj.scales": "model.safetensors",
|
| 524 |
+
"language_model.model.layers.26.self_attn.q_proj.weight": "model.safetensors",
|
| 525 |
+
"language_model.model.layers.26.self_attn.v_proj.bias": "model.safetensors",
|
| 526 |
+
"language_model.model.layers.26.self_attn.v_proj.biases": "model.safetensors",
|
| 527 |
+
"language_model.model.layers.26.self_attn.v_proj.scales": "model.safetensors",
|
| 528 |
+
"language_model.model.layers.26.self_attn.v_proj.weight": "model.safetensors",
|
| 529 |
+
"language_model.model.layers.27.input_layernorm.weight": "model.safetensors",
|
| 530 |
+
"language_model.model.layers.27.mlp.down_proj.biases": "model.safetensors",
|
| 531 |
+
"language_model.model.layers.27.mlp.down_proj.scales": "model.safetensors",
|
| 532 |
+
"language_model.model.layers.27.mlp.down_proj.weight": "model.safetensors",
|
| 533 |
+
"language_model.model.layers.27.mlp.gate_proj.biases": "model.safetensors",
|
| 534 |
+
"language_model.model.layers.27.mlp.gate_proj.scales": "model.safetensors",
|
| 535 |
+
"language_model.model.layers.27.mlp.gate_proj.weight": "model.safetensors",
|
| 536 |
+
"language_model.model.layers.27.mlp.up_proj.biases": "model.safetensors",
|
| 537 |
+
"language_model.model.layers.27.mlp.up_proj.scales": "model.safetensors",
|
| 538 |
+
"language_model.model.layers.27.mlp.up_proj.weight": "model.safetensors",
|
| 539 |
+
"language_model.model.layers.27.post_attention_layernorm.weight": "model.safetensors",
|
| 540 |
+
"language_model.model.layers.27.self_attn.k_proj.bias": "model.safetensors",
|
| 541 |
+
"language_model.model.layers.27.self_attn.k_proj.biases": "model.safetensors",
|
| 542 |
+
"language_model.model.layers.27.self_attn.k_proj.scales": "model.safetensors",
|
| 543 |
+
"language_model.model.layers.27.self_attn.k_proj.weight": "model.safetensors",
|
| 544 |
+
"language_model.model.layers.27.self_attn.o_proj.biases": "model.safetensors",
|
| 545 |
+
"language_model.model.layers.27.self_attn.o_proj.scales": "model.safetensors",
|
| 546 |
+
"language_model.model.layers.27.self_attn.o_proj.weight": "model.safetensors",
|
| 547 |
+
"language_model.model.layers.27.self_attn.q_proj.bias": "model.safetensors",
|
| 548 |
+
"language_model.model.layers.27.self_attn.q_proj.biases": "model.safetensors",
|
| 549 |
+
"language_model.model.layers.27.self_attn.q_proj.scales": "model.safetensors",
|
| 550 |
+
"language_model.model.layers.27.self_attn.q_proj.weight": "model.safetensors",
|
| 551 |
+
"language_model.model.layers.27.self_attn.v_proj.bias": "model.safetensors",
|
| 552 |
+
"language_model.model.layers.27.self_attn.v_proj.biases": "model.safetensors",
|
| 553 |
+
"language_model.model.layers.27.self_attn.v_proj.scales": "model.safetensors",
|
| 554 |
+
"language_model.model.layers.27.self_attn.v_proj.weight": "model.safetensors",
|
| 555 |
+
"language_model.model.layers.3.input_layernorm.weight": "model.safetensors",
|
| 556 |
+
"language_model.model.layers.3.mlp.down_proj.biases": "model.safetensors",
|
| 557 |
+
"language_model.model.layers.3.mlp.down_proj.scales": "model.safetensors",
|
| 558 |
+
"language_model.model.layers.3.mlp.down_proj.weight": "model.safetensors",
|
| 559 |
+
"language_model.model.layers.3.mlp.gate_proj.biases": "model.safetensors",
|
| 560 |
+
"language_model.model.layers.3.mlp.gate_proj.scales": "model.safetensors",
|
| 561 |
+
"language_model.model.layers.3.mlp.gate_proj.weight": "model.safetensors",
|
| 562 |
+
"language_model.model.layers.3.mlp.up_proj.biases": "model.safetensors",
|
| 563 |
+
"language_model.model.layers.3.mlp.up_proj.scales": "model.safetensors",
|
| 564 |
+
"language_model.model.layers.3.mlp.up_proj.weight": "model.safetensors",
|
| 565 |
+
"language_model.model.layers.3.post_attention_layernorm.weight": "model.safetensors",
|
| 566 |
+
"language_model.model.layers.3.self_attn.k_proj.bias": "model.safetensors",
|
| 567 |
+
"language_model.model.layers.3.self_attn.k_proj.biases": "model.safetensors",
|
| 568 |
+
"language_model.model.layers.3.self_attn.k_proj.scales": "model.safetensors",
|
| 569 |
+
"language_model.model.layers.3.self_attn.k_proj.weight": "model.safetensors",
|
| 570 |
+
"language_model.model.layers.3.self_attn.o_proj.biases": "model.safetensors",
|
| 571 |
+
"language_model.model.layers.3.self_attn.o_proj.scales": "model.safetensors",
|
| 572 |
+
"language_model.model.layers.3.self_attn.o_proj.weight": "model.safetensors",
|
| 573 |
+
"language_model.model.layers.3.self_attn.q_proj.bias": "model.safetensors",
|
| 574 |
+
"language_model.model.layers.3.self_attn.q_proj.biases": "model.safetensors",
|
| 575 |
+
"language_model.model.layers.3.self_attn.q_proj.scales": "model.safetensors",
|
| 576 |
+
"language_model.model.layers.3.self_attn.q_proj.weight": "model.safetensors",
|
| 577 |
+
"language_model.model.layers.3.self_attn.v_proj.bias": "model.safetensors",
|
| 578 |
+
"language_model.model.layers.3.self_attn.v_proj.biases": "model.safetensors",
|
| 579 |
+
"language_model.model.layers.3.self_attn.v_proj.scales": "model.safetensors",
|
| 580 |
+
"language_model.model.layers.3.self_attn.v_proj.weight": "model.safetensors",
|
| 581 |
+
"language_model.model.layers.4.input_layernorm.weight": "model.safetensors",
|
| 582 |
+
"language_model.model.layers.4.mlp.down_proj.biases": "model.safetensors",
|
| 583 |
+
"language_model.model.layers.4.mlp.down_proj.scales": "model.safetensors",
|
| 584 |
+
"language_model.model.layers.4.mlp.down_proj.weight": "model.safetensors",
|
| 585 |
+
"language_model.model.layers.4.mlp.gate_proj.biases": "model.safetensors",
|
| 586 |
+
"language_model.model.layers.4.mlp.gate_proj.scales": "model.safetensors",
|
| 587 |
+
"language_model.model.layers.4.mlp.gate_proj.weight": "model.safetensors",
|
| 588 |
+
"language_model.model.layers.4.mlp.up_proj.biases": "model.safetensors",
|
| 589 |
+
"language_model.model.layers.4.mlp.up_proj.scales": "model.safetensors",
|
| 590 |
+
"language_model.model.layers.4.mlp.up_proj.weight": "model.safetensors",
|
| 591 |
+
"language_model.model.layers.4.post_attention_layernorm.weight": "model.safetensors",
|
| 592 |
+
"language_model.model.layers.4.self_attn.k_proj.bias": "model.safetensors",
|
| 593 |
+
"language_model.model.layers.4.self_attn.k_proj.biases": "model.safetensors",
|
| 594 |
+
"language_model.model.layers.4.self_attn.k_proj.scales": "model.safetensors",
|
| 595 |
+
"language_model.model.layers.4.self_attn.k_proj.weight": "model.safetensors",
|
| 596 |
+
"language_model.model.layers.4.self_attn.o_proj.biases": "model.safetensors",
|
| 597 |
+
"language_model.model.layers.4.self_attn.o_proj.scales": "model.safetensors",
|
| 598 |
+
"language_model.model.layers.4.self_attn.o_proj.weight": "model.safetensors",
|
| 599 |
+
"language_model.model.layers.4.self_attn.q_proj.bias": "model.safetensors",
|
| 600 |
+
"language_model.model.layers.4.self_attn.q_proj.biases": "model.safetensors",
|
| 601 |
+
"language_model.model.layers.4.self_attn.q_proj.scales": "model.safetensors",
|
| 602 |
+
"language_model.model.layers.4.self_attn.q_proj.weight": "model.safetensors",
|
| 603 |
+
"language_model.model.layers.4.self_attn.v_proj.bias": "model.safetensors",
|
| 604 |
+
"language_model.model.layers.4.self_attn.v_proj.biases": "model.safetensors",
|
| 605 |
+
"language_model.model.layers.4.self_attn.v_proj.scales": "model.safetensors",
|
| 606 |
+
"language_model.model.layers.4.self_attn.v_proj.weight": "model.safetensors",
|
| 607 |
+
"language_model.model.layers.5.input_layernorm.weight": "model.safetensors",
|
| 608 |
+
"language_model.model.layers.5.mlp.down_proj.biases": "model.safetensors",
|
| 609 |
+
"language_model.model.layers.5.mlp.down_proj.scales": "model.safetensors",
|
| 610 |
+
"language_model.model.layers.5.mlp.down_proj.weight": "model.safetensors",
|
| 611 |
+
"language_model.model.layers.5.mlp.gate_proj.biases": "model.safetensors",
|
| 612 |
+
"language_model.model.layers.5.mlp.gate_proj.scales": "model.safetensors",
|
| 613 |
+
"language_model.model.layers.5.mlp.gate_proj.weight": "model.safetensors",
|
| 614 |
+
"language_model.model.layers.5.mlp.up_proj.biases": "model.safetensors",
|
| 615 |
+
"language_model.model.layers.5.mlp.up_proj.scales": "model.safetensors",
|
| 616 |
+
"language_model.model.layers.5.mlp.up_proj.weight": "model.safetensors",
|
| 617 |
+
"language_model.model.layers.5.post_attention_layernorm.weight": "model.safetensors",
|
| 618 |
+
"language_model.model.layers.5.self_attn.k_proj.bias": "model.safetensors",
|
| 619 |
+
"language_model.model.layers.5.self_attn.k_proj.biases": "model.safetensors",
|
| 620 |
+
"language_model.model.layers.5.self_attn.k_proj.scales": "model.safetensors",
|
| 621 |
+
"language_model.model.layers.5.self_attn.k_proj.weight": "model.safetensors",
|
| 622 |
+
"language_model.model.layers.5.self_attn.o_proj.biases": "model.safetensors",
|
| 623 |
+
"language_model.model.layers.5.self_attn.o_proj.scales": "model.safetensors",
|
| 624 |
+
"language_model.model.layers.5.self_attn.o_proj.weight": "model.safetensors",
|
| 625 |
+
"language_model.model.layers.5.self_attn.q_proj.bias": "model.safetensors",
|
| 626 |
+
"language_model.model.layers.5.self_attn.q_proj.biases": "model.safetensors",
|
| 627 |
+
"language_model.model.layers.5.self_attn.q_proj.scales": "model.safetensors",
|
| 628 |
+
"language_model.model.layers.5.self_attn.q_proj.weight": "model.safetensors",
|
| 629 |
+
"language_model.model.layers.5.self_attn.v_proj.bias": "model.safetensors",
|
| 630 |
+
"language_model.model.layers.5.self_attn.v_proj.biases": "model.safetensors",
|
| 631 |
+
"language_model.model.layers.5.self_attn.v_proj.scales": "model.safetensors",
|
| 632 |
+
"language_model.model.layers.5.self_attn.v_proj.weight": "model.safetensors",
|
| 633 |
+
"language_model.model.layers.6.input_layernorm.weight": "model.safetensors",
|
| 634 |
+
"language_model.model.layers.6.mlp.down_proj.biases": "model.safetensors",
|
| 635 |
+
"language_model.model.layers.6.mlp.down_proj.scales": "model.safetensors",
|
| 636 |
+
"language_model.model.layers.6.mlp.down_proj.weight": "model.safetensors",
|
| 637 |
+
"language_model.model.layers.6.mlp.gate_proj.biases": "model.safetensors",
|
| 638 |
+
"language_model.model.layers.6.mlp.gate_proj.scales": "model.safetensors",
|
| 639 |
+
"language_model.model.layers.6.mlp.gate_proj.weight": "model.safetensors",
|
| 640 |
+
"language_model.model.layers.6.mlp.up_proj.biases": "model.safetensors",
|
| 641 |
+
"language_model.model.layers.6.mlp.up_proj.scales": "model.safetensors",
|
| 642 |
+
"language_model.model.layers.6.mlp.up_proj.weight": "model.safetensors",
|
| 643 |
+
"language_model.model.layers.6.post_attention_layernorm.weight": "model.safetensors",
|
| 644 |
+
"language_model.model.layers.6.self_attn.k_proj.bias": "model.safetensors",
|
| 645 |
+
"language_model.model.layers.6.self_attn.k_proj.biases": "model.safetensors",
|
| 646 |
+
"language_model.model.layers.6.self_attn.k_proj.scales": "model.safetensors",
|
| 647 |
+
"language_model.model.layers.6.self_attn.k_proj.weight": "model.safetensors",
|
| 648 |
+
"language_model.model.layers.6.self_attn.o_proj.biases": "model.safetensors",
|
| 649 |
+
"language_model.model.layers.6.self_attn.o_proj.scales": "model.safetensors",
|
| 650 |
+
"language_model.model.layers.6.self_attn.o_proj.weight": "model.safetensors",
|
| 651 |
+
"language_model.model.layers.6.self_attn.q_proj.bias": "model.safetensors",
|
| 652 |
+
"language_model.model.layers.6.self_attn.q_proj.biases": "model.safetensors",
|
| 653 |
+
"language_model.model.layers.6.self_attn.q_proj.scales": "model.safetensors",
|
| 654 |
+
"language_model.model.layers.6.self_attn.q_proj.weight": "model.safetensors",
|
| 655 |
+
"language_model.model.layers.6.self_attn.v_proj.bias": "model.safetensors",
|
| 656 |
+
"language_model.model.layers.6.self_attn.v_proj.biases": "model.safetensors",
|
| 657 |
+
"language_model.model.layers.6.self_attn.v_proj.scales": "model.safetensors",
|
| 658 |
+
"language_model.model.layers.6.self_attn.v_proj.weight": "model.safetensors",
|
| 659 |
+
"language_model.model.layers.7.input_layernorm.weight": "model.safetensors",
|
| 660 |
+
"language_model.model.layers.7.mlp.down_proj.biases": "model.safetensors",
|
| 661 |
+
"language_model.model.layers.7.mlp.down_proj.scales": "model.safetensors",
|
| 662 |
+
"language_model.model.layers.7.mlp.down_proj.weight": "model.safetensors",
|
| 663 |
+
"language_model.model.layers.7.mlp.gate_proj.biases": "model.safetensors",
|
| 664 |
+
"language_model.model.layers.7.mlp.gate_proj.scales": "model.safetensors",
|
| 665 |
+
"language_model.model.layers.7.mlp.gate_proj.weight": "model.safetensors",
|
| 666 |
+
"language_model.model.layers.7.mlp.up_proj.biases": "model.safetensors",
|
| 667 |
+
"language_model.model.layers.7.mlp.up_proj.scales": "model.safetensors",
|
| 668 |
+
"language_model.model.layers.7.mlp.up_proj.weight": "model.safetensors",
|
| 669 |
+
"language_model.model.layers.7.post_attention_layernorm.weight": "model.safetensors",
|
| 670 |
+
"language_model.model.layers.7.self_attn.k_proj.bias": "model.safetensors",
|
| 671 |
+
"language_model.model.layers.7.self_attn.k_proj.biases": "model.safetensors",
|
| 672 |
+
"language_model.model.layers.7.self_attn.k_proj.scales": "model.safetensors",
|
| 673 |
+
"language_model.model.layers.7.self_attn.k_proj.weight": "model.safetensors",
|
| 674 |
+
"language_model.model.layers.7.self_attn.o_proj.biases": "model.safetensors",
|
| 675 |
+
"language_model.model.layers.7.self_attn.o_proj.scales": "model.safetensors",
|
| 676 |
+
"language_model.model.layers.7.self_attn.o_proj.weight": "model.safetensors",
|
| 677 |
+
"language_model.model.layers.7.self_attn.q_proj.bias": "model.safetensors",
|
| 678 |
+
"language_model.model.layers.7.self_attn.q_proj.biases": "model.safetensors",
|
| 679 |
+
"language_model.model.layers.7.self_attn.q_proj.scales": "model.safetensors",
|
| 680 |
+
"language_model.model.layers.7.self_attn.q_proj.weight": "model.safetensors",
|
| 681 |
+
"language_model.model.layers.7.self_attn.v_proj.bias": "model.safetensors",
|
| 682 |
+
"language_model.model.layers.7.self_attn.v_proj.biases": "model.safetensors",
|
| 683 |
+
"language_model.model.layers.7.self_attn.v_proj.scales": "model.safetensors",
|
| 684 |
+
"language_model.model.layers.7.self_attn.v_proj.weight": "model.safetensors",
|
| 685 |
+
"language_model.model.layers.8.input_layernorm.weight": "model.safetensors",
|
| 686 |
+
"language_model.model.layers.8.mlp.down_proj.biases": "model.safetensors",
|
| 687 |
+
"language_model.model.layers.8.mlp.down_proj.scales": "model.safetensors",
|
| 688 |
+
"language_model.model.layers.8.mlp.down_proj.weight": "model.safetensors",
|
| 689 |
+
"language_model.model.layers.8.mlp.gate_proj.biases": "model.safetensors",
|
| 690 |
+
"language_model.model.layers.8.mlp.gate_proj.scales": "model.safetensors",
|
| 691 |
+
"language_model.model.layers.8.mlp.gate_proj.weight": "model.safetensors",
|
| 692 |
+
"language_model.model.layers.8.mlp.up_proj.biases": "model.safetensors",
|
| 693 |
+
"language_model.model.layers.8.mlp.up_proj.scales": "model.safetensors",
|
| 694 |
+
"language_model.model.layers.8.mlp.up_proj.weight": "model.safetensors",
|
| 695 |
+
"language_model.model.layers.8.post_attention_layernorm.weight": "model.safetensors",
|
| 696 |
+
"language_model.model.layers.8.self_attn.k_proj.bias": "model.safetensors",
|
| 697 |
+
"language_model.model.layers.8.self_attn.k_proj.biases": "model.safetensors",
|
| 698 |
+
"language_model.model.layers.8.self_attn.k_proj.scales": "model.safetensors",
|
| 699 |
+
"language_model.model.layers.8.self_attn.k_proj.weight": "model.safetensors",
|
| 700 |
+
"language_model.model.layers.8.self_attn.o_proj.biases": "model.safetensors",
|
| 701 |
+
"language_model.model.layers.8.self_attn.o_proj.scales": "model.safetensors",
|
| 702 |
+
"language_model.model.layers.8.self_attn.o_proj.weight": "model.safetensors",
|
| 703 |
+
"language_model.model.layers.8.self_attn.q_proj.bias": "model.safetensors",
|
| 704 |
+
"language_model.model.layers.8.self_attn.q_proj.biases": "model.safetensors",
|
| 705 |
+
"language_model.model.layers.8.self_attn.q_proj.scales": "model.safetensors",
|
| 706 |
+
"language_model.model.layers.8.self_attn.q_proj.weight": "model.safetensors",
|
| 707 |
+
"language_model.model.layers.8.self_attn.v_proj.bias": "model.safetensors",
|
| 708 |
+
"language_model.model.layers.8.self_attn.v_proj.biases": "model.safetensors",
|
| 709 |
+
"language_model.model.layers.8.self_attn.v_proj.scales": "model.safetensors",
|
| 710 |
+
"language_model.model.layers.8.self_attn.v_proj.weight": "model.safetensors",
|
| 711 |
+
"language_model.model.layers.9.input_layernorm.weight": "model.safetensors",
|
| 712 |
+
"language_model.model.layers.9.mlp.down_proj.biases": "model.safetensors",
|
| 713 |
+
"language_model.model.layers.9.mlp.down_proj.scales": "model.safetensors",
|
| 714 |
+
"language_model.model.layers.9.mlp.down_proj.weight": "model.safetensors",
|
| 715 |
+
"language_model.model.layers.9.mlp.gate_proj.biases": "model.safetensors",
|
| 716 |
+
"language_model.model.layers.9.mlp.gate_proj.scales": "model.safetensors",
|
| 717 |
+
"language_model.model.layers.9.mlp.gate_proj.weight": "model.safetensors",
|
| 718 |
+
"language_model.model.layers.9.mlp.up_proj.biases": "model.safetensors",
|
| 719 |
+
"language_model.model.layers.9.mlp.up_proj.scales": "model.safetensors",
|
| 720 |
+
"language_model.model.layers.9.mlp.up_proj.weight": "model.safetensors",
|
| 721 |
+
"language_model.model.layers.9.post_attention_layernorm.weight": "model.safetensors",
|
| 722 |
+
"language_model.model.layers.9.self_attn.k_proj.bias": "model.safetensors",
|
| 723 |
+
"language_model.model.layers.9.self_attn.k_proj.biases": "model.safetensors",
|
| 724 |
+
"language_model.model.layers.9.self_attn.k_proj.scales": "model.safetensors",
|
| 725 |
+
"language_model.model.layers.9.self_attn.k_proj.weight": "model.safetensors",
|
| 726 |
+
"language_model.model.layers.9.self_attn.o_proj.biases": "model.safetensors",
|
| 727 |
+
"language_model.model.layers.9.self_attn.o_proj.scales": "model.safetensors",
|
| 728 |
+
"language_model.model.layers.9.self_attn.o_proj.weight": "model.safetensors",
|
| 729 |
+
"language_model.model.layers.9.self_attn.q_proj.bias": "model.safetensors",
|
| 730 |
+
"language_model.model.layers.9.self_attn.q_proj.biases": "model.safetensors",
|
| 731 |
+
"language_model.model.layers.9.self_attn.q_proj.scales": "model.safetensors",
|
| 732 |
+
"language_model.model.layers.9.self_attn.q_proj.weight": "model.safetensors",
|
| 733 |
+
"language_model.model.layers.9.self_attn.v_proj.bias": "model.safetensors",
|
| 734 |
+
"language_model.model.layers.9.self_attn.v_proj.biases": "model.safetensors",
|
| 735 |
+
"language_model.model.layers.9.self_attn.v_proj.scales": "model.safetensors",
|
| 736 |
+
"language_model.model.layers.9.self_attn.v_proj.weight": "model.safetensors",
|
| 737 |
+
"language_model.model.lm_head.biases": "model.safetensors",
|
| 738 |
+
"language_model.model.lm_head.scales": "model.safetensors",
|
| 739 |
+
"language_model.model.lm_head.weight": "model.safetensors",
|
| 740 |
+
"language_model.model.norm.weight": "model.safetensors",
|
| 741 |
+
"vision_tower.blocks.0.attn.proj.weight": "model.safetensors",
|
| 742 |
+
"vision_tower.blocks.0.attn.qkv.weight": "model.safetensors",
|
| 743 |
+
"vision_tower.blocks.0.mlp.fc1.weight": "model.safetensors",
|
| 744 |
+
"vision_tower.blocks.0.mlp.fc2.weight": "model.safetensors",
|
| 745 |
+
"vision_tower.blocks.0.mlp.fc3.weight": "model.safetensors",
|
| 746 |
+
"vision_tower.blocks.0.norm1.weight": "model.safetensors",
|
| 747 |
+
"vision_tower.blocks.0.norm2.weight": "model.safetensors",
|
| 748 |
+
"vision_tower.blocks.1.attn.proj.weight": "model.safetensors",
|
| 749 |
+
"vision_tower.blocks.1.attn.qkv.weight": "model.safetensors",
|
| 750 |
+
"vision_tower.blocks.1.mlp.fc1.weight": "model.safetensors",
|
| 751 |
+
"vision_tower.blocks.1.mlp.fc2.weight": "model.safetensors",
|
| 752 |
+
"vision_tower.blocks.1.mlp.fc3.weight": "model.safetensors",
|
| 753 |
+
"vision_tower.blocks.1.norm1.weight": "model.safetensors",
|
| 754 |
+
"vision_tower.blocks.1.norm2.weight": "model.safetensors",
|
| 755 |
+
"vision_tower.blocks.10.attn.proj.weight": "model.safetensors",
|
| 756 |
+
"vision_tower.blocks.10.attn.qkv.weight": "model.safetensors",
|
| 757 |
+
"vision_tower.blocks.10.mlp.fc1.weight": "model.safetensors",
|
| 758 |
+
"vision_tower.blocks.10.mlp.fc2.weight": "model.safetensors",
|
| 759 |
+
"vision_tower.blocks.10.mlp.fc3.weight": "model.safetensors",
|
| 760 |
+
"vision_tower.blocks.10.norm1.weight": "model.safetensors",
|
| 761 |
+
"vision_tower.blocks.10.norm2.weight": "model.safetensors",
|
| 762 |
+
"vision_tower.blocks.11.attn.proj.weight": "model.safetensors",
|
| 763 |
+
"vision_tower.blocks.11.attn.qkv.weight": "model.safetensors",
|
| 764 |
+
"vision_tower.blocks.11.mlp.fc1.weight": "model.safetensors",
|
| 765 |
+
"vision_tower.blocks.11.mlp.fc2.weight": "model.safetensors",
|
| 766 |
+
"vision_tower.blocks.11.mlp.fc3.weight": "model.safetensors",
|
| 767 |
+
"vision_tower.blocks.11.norm1.weight": "model.safetensors",
|
| 768 |
+
"vision_tower.blocks.11.norm2.weight": "model.safetensors",
|
| 769 |
+
"vision_tower.blocks.12.attn.proj.weight": "model.safetensors",
|
| 770 |
+
"vision_tower.blocks.12.attn.qkv.weight": "model.safetensors",
|
| 771 |
+
"vision_tower.blocks.12.mlp.fc1.weight": "model.safetensors",
|
| 772 |
+
"vision_tower.blocks.12.mlp.fc2.weight": "model.safetensors",
|
| 773 |
+
"vision_tower.blocks.12.mlp.fc3.weight": "model.safetensors",
|
| 774 |
+
"vision_tower.blocks.12.norm1.weight": "model.safetensors",
|
| 775 |
+
"vision_tower.blocks.12.norm2.weight": "model.safetensors",
|
| 776 |
+
"vision_tower.blocks.13.attn.proj.weight": "model.safetensors",
|
| 777 |
+
"vision_tower.blocks.13.attn.qkv.weight": "model.safetensors",
|
| 778 |
+
"vision_tower.blocks.13.mlp.fc1.weight": "model.safetensors",
|
| 779 |
+
"vision_tower.blocks.13.mlp.fc2.weight": "model.safetensors",
|
| 780 |
+
"vision_tower.blocks.13.mlp.fc3.weight": "model.safetensors",
|
| 781 |
+
"vision_tower.blocks.13.norm1.weight": "model.safetensors",
|
| 782 |
+
"vision_tower.blocks.13.norm2.weight": "model.safetensors",
|
| 783 |
+
"vision_tower.blocks.14.attn.proj.weight": "model.safetensors",
|
| 784 |
+
"vision_tower.blocks.14.attn.qkv.weight": "model.safetensors",
|
| 785 |
+
"vision_tower.blocks.14.mlp.fc1.weight": "model.safetensors",
|
| 786 |
+
"vision_tower.blocks.14.mlp.fc2.weight": "model.safetensors",
|
| 787 |
+
"vision_tower.blocks.14.mlp.fc3.weight": "model.safetensors",
|
| 788 |
+
"vision_tower.blocks.14.norm1.weight": "model.safetensors",
|
| 789 |
+
"vision_tower.blocks.14.norm2.weight": "model.safetensors",
|
| 790 |
+
"vision_tower.blocks.15.attn.proj.weight": "model.safetensors",
|
| 791 |
+
"vision_tower.blocks.15.attn.qkv.weight": "model.safetensors",
|
| 792 |
+
"vision_tower.blocks.15.mlp.fc1.weight": "model.safetensors",
|
| 793 |
+
"vision_tower.blocks.15.mlp.fc2.weight": "model.safetensors",
|
| 794 |
+
"vision_tower.blocks.15.mlp.fc3.weight": "model.safetensors",
|
| 795 |
+
"vision_tower.blocks.15.norm1.weight": "model.safetensors",
|
| 796 |
+
"vision_tower.blocks.15.norm2.weight": "model.safetensors",
|
| 797 |
+
"vision_tower.blocks.16.attn.proj.weight": "model.safetensors",
|
| 798 |
+
"vision_tower.blocks.16.attn.qkv.weight": "model.safetensors",
|
| 799 |
+
"vision_tower.blocks.16.mlp.fc1.weight": "model.safetensors",
|
| 800 |
+
"vision_tower.blocks.16.mlp.fc2.weight": "model.safetensors",
|
| 801 |
+
"vision_tower.blocks.16.mlp.fc3.weight": "model.safetensors",
|
| 802 |
+
"vision_tower.blocks.16.norm1.weight": "model.safetensors",
|
| 803 |
+
"vision_tower.blocks.16.norm2.weight": "model.safetensors",
|
| 804 |
+
"vision_tower.blocks.17.attn.proj.weight": "model.safetensors",
|
| 805 |
+
"vision_tower.blocks.17.attn.qkv.weight": "model.safetensors",
|
| 806 |
+
"vision_tower.blocks.17.mlp.fc1.weight": "model.safetensors",
|
| 807 |
+
"vision_tower.blocks.17.mlp.fc2.weight": "model.safetensors",
|
| 808 |
+
"vision_tower.blocks.17.mlp.fc3.weight": "model.safetensors",
|
| 809 |
+
"vision_tower.blocks.17.norm1.weight": "model.safetensors",
|
| 810 |
+
"vision_tower.blocks.17.norm2.weight": "model.safetensors",
|
| 811 |
+
"vision_tower.blocks.18.attn.proj.weight": "model.safetensors",
|
| 812 |
+
"vision_tower.blocks.18.attn.qkv.weight": "model.safetensors",
|
| 813 |
+
"vision_tower.blocks.18.mlp.fc1.weight": "model.safetensors",
|
| 814 |
+
"vision_tower.blocks.18.mlp.fc2.weight": "model.safetensors",
|
| 815 |
+
"vision_tower.blocks.18.mlp.fc3.weight": "model.safetensors",
|
| 816 |
+
"vision_tower.blocks.18.norm1.weight": "model.safetensors",
|
| 817 |
+
"vision_tower.blocks.18.norm2.weight": "model.safetensors",
|
| 818 |
+
"vision_tower.blocks.19.attn.proj.weight": "model.safetensors",
|
| 819 |
+
"vision_tower.blocks.19.attn.qkv.weight": "model.safetensors",
|
| 820 |
+
"vision_tower.blocks.19.mlp.fc1.weight": "model.safetensors",
|
| 821 |
+
"vision_tower.blocks.19.mlp.fc2.weight": "model.safetensors",
|
| 822 |
+
"vision_tower.blocks.19.mlp.fc3.weight": "model.safetensors",
|
| 823 |
+
"vision_tower.blocks.19.norm1.weight": "model.safetensors",
|
| 824 |
+
"vision_tower.blocks.19.norm2.weight": "model.safetensors",
|
| 825 |
+
"vision_tower.blocks.2.attn.proj.weight": "model.safetensors",
|
| 826 |
+
"vision_tower.blocks.2.attn.qkv.weight": "model.safetensors",
|
| 827 |
+
"vision_tower.blocks.2.mlp.fc1.weight": "model.safetensors",
|
| 828 |
+
"vision_tower.blocks.2.mlp.fc2.weight": "model.safetensors",
|
| 829 |
+
"vision_tower.blocks.2.mlp.fc3.weight": "model.safetensors",
|
| 830 |
+
"vision_tower.blocks.2.norm1.weight": "model.safetensors",
|
| 831 |
+
"vision_tower.blocks.2.norm2.weight": "model.safetensors",
|
| 832 |
+
"vision_tower.blocks.20.attn.proj.weight": "model.safetensors",
|
| 833 |
+
"vision_tower.blocks.20.attn.qkv.weight": "model.safetensors",
|
| 834 |
+
"vision_tower.blocks.20.mlp.fc1.weight": "model.safetensors",
|
| 835 |
+
"vision_tower.blocks.20.mlp.fc2.weight": "model.safetensors",
|
| 836 |
+
"vision_tower.blocks.20.mlp.fc3.weight": "model.safetensors",
|
| 837 |
+
"vision_tower.blocks.20.norm1.weight": "model.safetensors",
|
| 838 |
+
"vision_tower.blocks.20.norm2.weight": "model.safetensors",
|
| 839 |
+
"vision_tower.blocks.21.attn.proj.weight": "model.safetensors",
|
| 840 |
+
"vision_tower.blocks.21.attn.qkv.weight": "model.safetensors",
|
| 841 |
+
"vision_tower.blocks.21.mlp.fc1.weight": "model.safetensors",
|
| 842 |
+
"vision_tower.blocks.21.mlp.fc2.weight": "model.safetensors",
|
| 843 |
+
"vision_tower.blocks.21.mlp.fc3.weight": "model.safetensors",
|
| 844 |
+
"vision_tower.blocks.21.norm1.weight": "model.safetensors",
|
| 845 |
+
"vision_tower.blocks.21.norm2.weight": "model.safetensors",
|
| 846 |
+
"vision_tower.blocks.22.attn.proj.weight": "model.safetensors",
|
| 847 |
+
"vision_tower.blocks.22.attn.qkv.weight": "model.safetensors",
|
| 848 |
+
"vision_tower.blocks.22.mlp.fc1.weight": "model.safetensors",
|
| 849 |
+
"vision_tower.blocks.22.mlp.fc2.weight": "model.safetensors",
|
| 850 |
+
"vision_tower.blocks.22.mlp.fc3.weight": "model.safetensors",
|
| 851 |
+
"vision_tower.blocks.22.norm1.weight": "model.safetensors",
|
| 852 |
+
"vision_tower.blocks.22.norm2.weight": "model.safetensors",
|
| 853 |
+
"vision_tower.blocks.23.attn.proj.weight": "model.safetensors",
|
| 854 |
+
"vision_tower.blocks.23.attn.qkv.weight": "model.safetensors",
|
| 855 |
+
"vision_tower.blocks.23.mlp.fc1.weight": "model.safetensors",
|
| 856 |
+
"vision_tower.blocks.23.mlp.fc2.weight": "model.safetensors",
|
| 857 |
+
"vision_tower.blocks.23.mlp.fc3.weight": "model.safetensors",
|
| 858 |
+
"vision_tower.blocks.23.norm1.weight": "model.safetensors",
|
| 859 |
+
"vision_tower.blocks.23.norm2.weight": "model.safetensors",
|
| 860 |
+
"vision_tower.blocks.24.attn.proj.weight": "model.safetensors",
|
| 861 |
+
"vision_tower.blocks.24.attn.qkv.weight": "model.safetensors",
|
| 862 |
+
"vision_tower.blocks.24.mlp.fc1.weight": "model.safetensors",
|
| 863 |
+
"vision_tower.blocks.24.mlp.fc2.weight": "model.safetensors",
|
| 864 |
+
"vision_tower.blocks.24.mlp.fc3.weight": "model.safetensors",
|
| 865 |
+
"vision_tower.blocks.24.norm1.weight": "model.safetensors",
|
| 866 |
+
"vision_tower.blocks.24.norm2.weight": "model.safetensors",
|
| 867 |
+
"vision_tower.blocks.25.attn.proj.weight": "model.safetensors",
|
| 868 |
+
"vision_tower.blocks.25.attn.qkv.weight": "model.safetensors",
|
| 869 |
+
"vision_tower.blocks.25.mlp.fc1.weight": "model.safetensors",
|
| 870 |
+
"vision_tower.blocks.25.mlp.fc2.weight": "model.safetensors",
|
| 871 |
+
"vision_tower.blocks.25.mlp.fc3.weight": "model.safetensors",
|
| 872 |
+
"vision_tower.blocks.25.norm1.weight": "model.safetensors",
|
| 873 |
+
"vision_tower.blocks.25.norm2.weight": "model.safetensors",
|
| 874 |
+
"vision_tower.blocks.26.attn.proj.weight": "model.safetensors",
|
| 875 |
+
"vision_tower.blocks.26.attn.qkv.weight": "model.safetensors",
|
| 876 |
+
"vision_tower.blocks.26.mlp.fc1.weight": "model.safetensors",
|
| 877 |
+
"vision_tower.blocks.26.mlp.fc2.weight": "model.safetensors",
|
| 878 |
+
"vision_tower.blocks.26.mlp.fc3.weight": "model.safetensors",
|
| 879 |
+
"vision_tower.blocks.26.norm1.weight": "model.safetensors",
|
| 880 |
+
"vision_tower.blocks.26.norm2.weight": "model.safetensors",
|
| 881 |
+
"vision_tower.blocks.27.attn.proj.weight": "model.safetensors",
|
| 882 |
+
"vision_tower.blocks.27.attn.qkv.weight": "model.safetensors",
|
| 883 |
+
"vision_tower.blocks.27.mlp.fc1.weight": "model.safetensors",
|
| 884 |
+
"vision_tower.blocks.27.mlp.fc2.weight": "model.safetensors",
|
| 885 |
+
"vision_tower.blocks.27.mlp.fc3.weight": "model.safetensors",
|
| 886 |
+
"vision_tower.blocks.27.norm1.weight": "model.safetensors",
|
| 887 |
+
"vision_tower.blocks.27.norm2.weight": "model.safetensors",
|
| 888 |
+
"vision_tower.blocks.28.attn.proj.weight": "model.safetensors",
|
| 889 |
+
"vision_tower.blocks.28.attn.qkv.weight": "model.safetensors",
|
| 890 |
+
"vision_tower.blocks.28.mlp.fc1.weight": "model.safetensors",
|
| 891 |
+
"vision_tower.blocks.28.mlp.fc2.weight": "model.safetensors",
|
| 892 |
+
"vision_tower.blocks.28.mlp.fc3.weight": "model.safetensors",
|
| 893 |
+
"vision_tower.blocks.28.norm1.weight": "model.safetensors",
|
| 894 |
+
"vision_tower.blocks.28.norm2.weight": "model.safetensors",
|
| 895 |
+
"vision_tower.blocks.29.attn.proj.weight": "model.safetensors",
|
| 896 |
+
"vision_tower.blocks.29.attn.qkv.weight": "model.safetensors",
|
| 897 |
+
"vision_tower.blocks.29.mlp.fc1.weight": "model.safetensors",
|
| 898 |
+
"vision_tower.blocks.29.mlp.fc2.weight": "model.safetensors",
|
| 899 |
+
"vision_tower.blocks.29.mlp.fc3.weight": "model.safetensors",
|
| 900 |
+
"vision_tower.blocks.29.norm1.weight": "model.safetensors",
|
| 901 |
+
"vision_tower.blocks.29.norm2.weight": "model.safetensors",
|
| 902 |
+
"vision_tower.blocks.3.attn.proj.weight": "model.safetensors",
|
| 903 |
+
"vision_tower.blocks.3.attn.qkv.weight": "model.safetensors",
|
| 904 |
+
"vision_tower.blocks.3.mlp.fc1.weight": "model.safetensors",
|
| 905 |
+
"vision_tower.blocks.3.mlp.fc2.weight": "model.safetensors",
|
| 906 |
+
"vision_tower.blocks.3.mlp.fc3.weight": "model.safetensors",
|
| 907 |
+
"vision_tower.blocks.3.norm1.weight": "model.safetensors",
|
| 908 |
+
"vision_tower.blocks.3.norm2.weight": "model.safetensors",
|
| 909 |
+
"vision_tower.blocks.30.attn.proj.weight": "model.safetensors",
|
| 910 |
+
"vision_tower.blocks.30.attn.qkv.weight": "model.safetensors",
|
| 911 |
+
"vision_tower.blocks.30.mlp.fc1.weight": "model.safetensors",
|
| 912 |
+
"vision_tower.blocks.30.mlp.fc2.weight": "model.safetensors",
|
| 913 |
+
"vision_tower.blocks.30.mlp.fc3.weight": "model.safetensors",
|
| 914 |
+
"vision_tower.blocks.30.norm1.weight": "model.safetensors",
|
| 915 |
+
"vision_tower.blocks.30.norm2.weight": "model.safetensors",
|
| 916 |
+
"vision_tower.blocks.31.attn.proj.weight": "model.safetensors",
|
| 917 |
+
"vision_tower.blocks.31.attn.qkv.weight": "model.safetensors",
|
| 918 |
+
"vision_tower.blocks.31.mlp.fc1.weight": "model.safetensors",
|
| 919 |
+
"vision_tower.blocks.31.mlp.fc2.weight": "model.safetensors",
|
| 920 |
+
"vision_tower.blocks.31.mlp.fc3.weight": "model.safetensors",
|
| 921 |
+
"vision_tower.blocks.31.norm1.weight": "model.safetensors",
|
| 922 |
+
"vision_tower.blocks.31.norm2.weight": "model.safetensors",
|
| 923 |
+
"vision_tower.blocks.32.attn.proj.weight": "model.safetensors",
|
| 924 |
+
"vision_tower.blocks.32.attn.qkv.weight": "model.safetensors",
|
| 925 |
+
"vision_tower.blocks.32.mlp.fc1.weight": "model.safetensors",
|
| 926 |
+
"vision_tower.blocks.32.mlp.fc2.weight": "model.safetensors",
|
| 927 |
+
"vision_tower.blocks.32.mlp.fc3.weight": "model.safetensors",
|
| 928 |
+
"vision_tower.blocks.32.norm1.weight": "model.safetensors",
|
| 929 |
+
"vision_tower.blocks.32.norm2.weight": "model.safetensors",
|
| 930 |
+
"vision_tower.blocks.33.attn.proj.weight": "model.safetensors",
|
| 931 |
+
"vision_tower.blocks.33.attn.qkv.weight": "model.safetensors",
|
| 932 |
+
"vision_tower.blocks.33.mlp.fc1.weight": "model.safetensors",
|
| 933 |
+
"vision_tower.blocks.33.mlp.fc2.weight": "model.safetensors",
|
| 934 |
+
"vision_tower.blocks.33.mlp.fc3.weight": "model.safetensors",
|
| 935 |
+
"vision_tower.blocks.33.norm1.weight": "model.safetensors",
|
| 936 |
+
"vision_tower.blocks.33.norm2.weight": "model.safetensors",
|
| 937 |
+
"vision_tower.blocks.34.attn.proj.weight": "model.safetensors",
|
| 938 |
+
"vision_tower.blocks.34.attn.qkv.weight": "model.safetensors",
|
| 939 |
+
"vision_tower.blocks.34.mlp.fc1.weight": "model.safetensors",
|
| 940 |
+
"vision_tower.blocks.34.mlp.fc2.weight": "model.safetensors",
|
| 941 |
+
"vision_tower.blocks.34.mlp.fc3.weight": "model.safetensors",
|
| 942 |
+
"vision_tower.blocks.34.norm1.weight": "model.safetensors",
|
| 943 |
+
"vision_tower.blocks.34.norm2.weight": "model.safetensors",
|
| 944 |
+
"vision_tower.blocks.35.attn.proj.weight": "model.safetensors",
|
| 945 |
+
"vision_tower.blocks.35.attn.qkv.weight": "model.safetensors",
|
| 946 |
+
"vision_tower.blocks.35.mlp.fc1.weight": "model.safetensors",
|
| 947 |
+
"vision_tower.blocks.35.mlp.fc2.weight": "model.safetensors",
|
| 948 |
+
"vision_tower.blocks.35.mlp.fc3.weight": "model.safetensors",
|
| 949 |
+
"vision_tower.blocks.35.norm1.weight": "model.safetensors",
|
| 950 |
+
"vision_tower.blocks.35.norm2.weight": "model.safetensors",
|
| 951 |
+
"vision_tower.blocks.36.attn.proj.weight": "model.safetensors",
|
| 952 |
+
"vision_tower.blocks.36.attn.qkv.weight": "model.safetensors",
|
| 953 |
+
"vision_tower.blocks.36.mlp.fc1.weight": "model.safetensors",
|
| 954 |
+
"vision_tower.blocks.36.mlp.fc2.weight": "model.safetensors",
|
| 955 |
+
"vision_tower.blocks.36.mlp.fc3.weight": "model.safetensors",
|
| 956 |
+
"vision_tower.blocks.36.norm1.weight": "model.safetensors",
|
| 957 |
+
"vision_tower.blocks.36.norm2.weight": "model.safetensors",
|
| 958 |
+
"vision_tower.blocks.37.attn.proj.weight": "model.safetensors",
|
| 959 |
+
"vision_tower.blocks.37.attn.qkv.weight": "model.safetensors",
|
| 960 |
+
"vision_tower.blocks.37.mlp.fc1.weight": "model.safetensors",
|
| 961 |
+
"vision_tower.blocks.37.mlp.fc2.weight": "model.safetensors",
|
| 962 |
+
"vision_tower.blocks.37.mlp.fc3.weight": "model.safetensors",
|
| 963 |
+
"vision_tower.blocks.37.norm1.weight": "model.safetensors",
|
| 964 |
+
"vision_tower.blocks.37.norm2.weight": "model.safetensors",
|
| 965 |
+
"vision_tower.blocks.38.attn.proj.weight": "model.safetensors",
|
| 966 |
+
"vision_tower.blocks.38.attn.qkv.weight": "model.safetensors",
|
| 967 |
+
"vision_tower.blocks.38.mlp.fc1.weight": "model.safetensors",
|
| 968 |
+
"vision_tower.blocks.38.mlp.fc2.weight": "model.safetensors",
|
| 969 |
+
"vision_tower.blocks.38.mlp.fc3.weight": "model.safetensors",
|
| 970 |
+
"vision_tower.blocks.38.norm1.weight": "model.safetensors",
|
| 971 |
+
"vision_tower.blocks.38.norm2.weight": "model.safetensors",
|
| 972 |
+
"vision_tower.blocks.39.attn.proj.weight": "model.safetensors",
|
| 973 |
+
"vision_tower.blocks.39.attn.qkv.weight": "model.safetensors",
|
| 974 |
+
"vision_tower.blocks.39.mlp.fc1.weight": "model.safetensors",
|
| 975 |
+
"vision_tower.blocks.39.mlp.fc2.weight": "model.safetensors",
|
| 976 |
+
"vision_tower.blocks.39.mlp.fc3.weight": "model.safetensors",
|
| 977 |
+
"vision_tower.blocks.39.norm1.weight": "model.safetensors",
|
| 978 |
+
"vision_tower.blocks.39.norm2.weight": "model.safetensors",
|
| 979 |
+
"vision_tower.blocks.4.attn.proj.weight": "model.safetensors",
|
| 980 |
+
"vision_tower.blocks.4.attn.qkv.weight": "model.safetensors",
|
| 981 |
+
"vision_tower.blocks.4.mlp.fc1.weight": "model.safetensors",
|
| 982 |
+
"vision_tower.blocks.4.mlp.fc2.weight": "model.safetensors",
|
| 983 |
+
"vision_tower.blocks.4.mlp.fc3.weight": "model.safetensors",
|
| 984 |
+
"vision_tower.blocks.4.norm1.weight": "model.safetensors",
|
| 985 |
+
"vision_tower.blocks.4.norm2.weight": "model.safetensors",
|
| 986 |
+
"vision_tower.blocks.40.attn.proj.weight": "model.safetensors",
|
| 987 |
+
"vision_tower.blocks.40.attn.qkv.weight": "model.safetensors",
|
| 988 |
+
"vision_tower.blocks.40.mlp.fc1.weight": "model.safetensors",
|
| 989 |
+
"vision_tower.blocks.40.mlp.fc2.weight": "model.safetensors",
|
| 990 |
+
"vision_tower.blocks.40.mlp.fc3.weight": "model.safetensors",
|
| 991 |
+
"vision_tower.blocks.40.norm1.weight": "model.safetensors",
|
| 992 |
+
"vision_tower.blocks.40.norm2.weight": "model.safetensors",
|
| 993 |
+
"vision_tower.blocks.41.attn.proj.weight": "model.safetensors",
|
| 994 |
+
"vision_tower.blocks.41.attn.qkv.weight": "model.safetensors",
|
| 995 |
+
"vision_tower.blocks.41.mlp.fc1.weight": "model.safetensors",
|
| 996 |
+
"vision_tower.blocks.41.mlp.fc2.weight": "model.safetensors",
|
| 997 |
+
"vision_tower.blocks.41.mlp.fc3.weight": "model.safetensors",
|
| 998 |
+
"vision_tower.blocks.41.norm1.weight": "model.safetensors",
|
| 999 |
+
"vision_tower.blocks.41.norm2.weight": "model.safetensors",
|
| 1000 |
+
"vision_tower.blocks.5.attn.proj.weight": "model.safetensors",
|
| 1001 |
+
"vision_tower.blocks.5.attn.qkv.weight": "model.safetensors",
|
| 1002 |
+
"vision_tower.blocks.5.mlp.fc1.weight": "model.safetensors",
|
| 1003 |
+
"vision_tower.blocks.5.mlp.fc2.weight": "model.safetensors",
|
| 1004 |
+
"vision_tower.blocks.5.mlp.fc3.weight": "model.safetensors",
|
| 1005 |
+
"vision_tower.blocks.5.norm1.weight": "model.safetensors",
|
| 1006 |
+
"vision_tower.blocks.5.norm2.weight": "model.safetensors",
|
| 1007 |
+
"vision_tower.blocks.6.attn.proj.weight": "model.safetensors",
|
| 1008 |
+
"vision_tower.blocks.6.attn.qkv.weight": "model.safetensors",
|
| 1009 |
+
"vision_tower.blocks.6.mlp.fc1.weight": "model.safetensors",
|
| 1010 |
+
"vision_tower.blocks.6.mlp.fc2.weight": "model.safetensors",
|
| 1011 |
+
"vision_tower.blocks.6.mlp.fc3.weight": "model.safetensors",
|
| 1012 |
+
"vision_tower.blocks.6.norm1.weight": "model.safetensors",
|
| 1013 |
+
"vision_tower.blocks.6.norm2.weight": "model.safetensors",
|
| 1014 |
+
"vision_tower.blocks.7.attn.proj.weight": "model.safetensors",
|
| 1015 |
+
"vision_tower.blocks.7.attn.qkv.weight": "model.safetensors",
|
| 1016 |
+
"vision_tower.blocks.7.mlp.fc1.weight": "model.safetensors",
|
| 1017 |
+
"vision_tower.blocks.7.mlp.fc2.weight": "model.safetensors",
|
| 1018 |
+
"vision_tower.blocks.7.mlp.fc3.weight": "model.safetensors",
|
| 1019 |
+
"vision_tower.blocks.7.norm1.weight": "model.safetensors",
|
| 1020 |
+
"vision_tower.blocks.7.norm2.weight": "model.safetensors",
|
| 1021 |
+
"vision_tower.blocks.8.attn.proj.weight": "model.safetensors",
|
| 1022 |
+
"vision_tower.blocks.8.attn.qkv.weight": "model.safetensors",
|
| 1023 |
+
"vision_tower.blocks.8.mlp.fc1.weight": "model.safetensors",
|
| 1024 |
+
"vision_tower.blocks.8.mlp.fc2.weight": "model.safetensors",
|
| 1025 |
+
"vision_tower.blocks.8.mlp.fc3.weight": "model.safetensors",
|
| 1026 |
+
"vision_tower.blocks.8.norm1.weight": "model.safetensors",
|
| 1027 |
+
"vision_tower.blocks.8.norm2.weight": "model.safetensors",
|
| 1028 |
+
"vision_tower.blocks.9.attn.proj.weight": "model.safetensors",
|
| 1029 |
+
"vision_tower.blocks.9.attn.qkv.weight": "model.safetensors",
|
| 1030 |
+
"vision_tower.blocks.9.mlp.fc1.weight": "model.safetensors",
|
| 1031 |
+
"vision_tower.blocks.9.mlp.fc2.weight": "model.safetensors",
|
| 1032 |
+
"vision_tower.blocks.9.mlp.fc3.weight": "model.safetensors",
|
| 1033 |
+
"vision_tower.blocks.9.norm1.weight": "model.safetensors",
|
| 1034 |
+
"vision_tower.blocks.9.norm2.weight": "model.safetensors",
|
| 1035 |
+
"vision_tower.merger.ln_q.bias": "model.safetensors",
|
| 1036 |
+
"vision_tower.merger.ln_q.weight": "model.safetensors",
|
| 1037 |
+
"vision_tower.merger.mlp.0.bias": "model.safetensors",
|
| 1038 |
+
"vision_tower.merger.mlp.0.weight": "model.safetensors",
|
| 1039 |
+
"vision_tower.merger.mlp.2.bias": "model.safetensors",
|
| 1040 |
+
"vision_tower.merger.mlp.2.weight": "model.safetensors",
|
| 1041 |
+
"vision_tower.patch_embed.patchifier.norm.weight": "model.safetensors",
|
| 1042 |
+
"vision_tower.patch_embed.patchifier.proj.bias": "model.safetensors",
|
| 1043 |
+
"vision_tower.patch_embed.patchifier.proj.weight": "model.safetensors",
|
| 1044 |
+
"vision_tower.post_trunk_norm.weight": "model.safetensors"
|
| 1045 |
+
}
|
| 1046 |
+
}
|
modeling_dots_ocr.py
ADDED
|
@@ -0,0 +1,131 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from typing import List, Optional, Tuple, Union
|
| 2 |
+
|
| 3 |
+
import torch
|
| 4 |
+
from transformers.modeling_outputs import CausalLMOutputWithPast
|
| 5 |
+
from transformers.models.qwen2 import Qwen2ForCausalLM
|
| 6 |
+
|
| 7 |
+
from .configuration_dots import DotsVisionConfig, DotsOCRConfig
|
| 8 |
+
from .modeling_dots_vision import DotsVisionTransformer
|
| 9 |
+
|
| 10 |
+
|
| 11 |
+
DOTS_VLM_MAX_IMAGES = 200
|
| 12 |
+
|
| 13 |
+
|
| 14 |
+
class DotsOCRForCausalLM(Qwen2ForCausalLM):
|
| 15 |
+
config_class = DotsOCRConfig
|
| 16 |
+
|
| 17 |
+
def __init__(self, config: DotsOCRConfig):
|
| 18 |
+
super().__init__(config)
|
| 19 |
+
|
| 20 |
+
if isinstance(self.config.vision_config, dict):
|
| 21 |
+
vision_config = DotsVisionConfig(**self.config.vision_config)
|
| 22 |
+
self.config.vision_config = vision_config
|
| 23 |
+
else:
|
| 24 |
+
vision_config = self.config.vision_config
|
| 25 |
+
|
| 26 |
+
self.vision_tower = DotsVisionTransformer(vision_config)
|
| 27 |
+
|
| 28 |
+
def prepare_inputs_embeds(
|
| 29 |
+
self,
|
| 30 |
+
input_ids: torch.LongTensor,
|
| 31 |
+
pixel_values: Optional[torch.FloatTensor] = None,
|
| 32 |
+
grid_thw: Optional[torch.FloatTensor] = None,
|
| 33 |
+
img_mask: Optional[torch.BoolTensor] = None,
|
| 34 |
+
) -> torch.Tensor:
|
| 35 |
+
inputs_embeds = self.get_input_embeddings()(input_ids)
|
| 36 |
+
|
| 37 |
+
if pixel_values is not None:
|
| 38 |
+
assert img_mask is not None
|
| 39 |
+
if grid_thw.shape[0] > DOTS_VLM_MAX_IMAGES:
|
| 40 |
+
print(
|
| 41 |
+
f"Num image exceeded: {grid_thw.shape[0]} > {DOTS_VLM_MAX_IMAGES}, which may cause FSDP hang"
|
| 42 |
+
)
|
| 43 |
+
|
| 44 |
+
vision_embeddings = self.vision_tower(pixel_values, grid_thw)
|
| 45 |
+
|
| 46 |
+
true_indices = torch.nonzero(img_mask).squeeze()
|
| 47 |
+
if len(true_indices) > vision_embeddings.size(0):
|
| 48 |
+
print(
|
| 49 |
+
f"img_mask sum > VE and will be truncated, mask.sum()={len(true_indices)} {vision_embeddings.size(0)=}"
|
| 50 |
+
)
|
| 51 |
+
true_indices = true_indices[: vision_embeddings.size(0)]
|
| 52 |
+
new_img_mask = torch.zeros_like(img_mask, device=img_mask.device)
|
| 53 |
+
new_img_mask[true_indices[:, 0], true_indices[:, 1]] = True
|
| 54 |
+
else:
|
| 55 |
+
new_img_mask = img_mask
|
| 56 |
+
|
| 57 |
+
assert (
|
| 58 |
+
vision_embeddings.size(0) == new_img_mask.sum()
|
| 59 |
+
), f"{vision_embeddings.size(0)=}, {new_img_mask.sum()=}"
|
| 60 |
+
|
| 61 |
+
inputs_embeds = inputs_embeds.masked_scatter(
|
| 62 |
+
new_img_mask.to(inputs_embeds.device).unsqueeze(-1).expand_as(inputs_embeds),
|
| 63 |
+
vision_embeddings.to(inputs_embeds.device).type(inputs_embeds.dtype),
|
| 64 |
+
)
|
| 65 |
+
|
| 66 |
+
return inputs_embeds
|
| 67 |
+
|
| 68 |
+
def forward(
|
| 69 |
+
self,
|
| 70 |
+
input_ids: torch.LongTensor,
|
| 71 |
+
pixel_values: Optional[torch.FloatTensor] = None,
|
| 72 |
+
image_grid_thw: Optional[torch.FloatTensor] = None,
|
| 73 |
+
inputs_embeds: Optional[torch.Tensor] = None,
|
| 74 |
+
attention_mask: Optional[torch.Tensor] = None,
|
| 75 |
+
position_ids: Optional[torch.LongTensor] = None,
|
| 76 |
+
past_key_values: Optional[List[torch.FloatTensor]] = None,
|
| 77 |
+
labels: Optional[torch.LongTensor] = None,
|
| 78 |
+
output_attentions: Optional[bool] = None,
|
| 79 |
+
output_hidden_states: Optional[bool] = None,
|
| 80 |
+
return_dict: Optional[bool] = None,
|
| 81 |
+
use_cache: Optional[bool] = None,
|
| 82 |
+
logits_to_keep: int = 0,
|
| 83 |
+
**loss_kwargs,
|
| 84 |
+
) -> Union[Tuple, CausalLMOutputWithPast]:
|
| 85 |
+
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
| 86 |
+
assert len(input_ids) >= 1, f"empty input_ids {input_ids.shape=} will cause gradnorm nan"
|
| 87 |
+
if inputs_embeds is None:
|
| 88 |
+
img_mask = input_ids == self.config.image_token_id
|
| 89 |
+
inputs_embeds = self.prepare_inputs_embeds(input_ids, pixel_values, image_grid_thw, img_mask)
|
| 90 |
+
|
| 91 |
+
outputs = super().forward(
|
| 92 |
+
inputs_embeds=inputs_embeds,
|
| 93 |
+
attention_mask=attention_mask,
|
| 94 |
+
position_ids=position_ids,
|
| 95 |
+
past_key_values=past_key_values,
|
| 96 |
+
labels=labels,
|
| 97 |
+
use_cache=use_cache if use_cache is not None else self.config.use_cache,
|
| 98 |
+
output_attentions=output_attentions,
|
| 99 |
+
output_hidden_states=output_hidden_states,
|
| 100 |
+
# return_dict=return_dict,
|
| 101 |
+
logits_to_keep=logits_to_keep,
|
| 102 |
+
**loss_kwargs,
|
| 103 |
+
)
|
| 104 |
+
|
| 105 |
+
return outputs
|
| 106 |
+
|
| 107 |
+
def prepare_inputs_for_generation(
|
| 108 |
+
self,
|
| 109 |
+
input_ids,
|
| 110 |
+
past_key_values=None,
|
| 111 |
+
inputs_embeds=None,
|
| 112 |
+
pixel_values=None,
|
| 113 |
+
attention_mask=None,
|
| 114 |
+
cache_position=None,
|
| 115 |
+
num_logits_to_keep=None,
|
| 116 |
+
**kwargs,
|
| 117 |
+
):
|
| 118 |
+
model_inputs = super().prepare_inputs_for_generation(
|
| 119 |
+
input_ids,
|
| 120 |
+
past_key_values=past_key_values,
|
| 121 |
+
inputs_embeds=inputs_embeds,
|
| 122 |
+
attention_mask=attention_mask,
|
| 123 |
+
cache_position=cache_position,
|
| 124 |
+
num_logits_to_keep=num_logits_to_keep,
|
| 125 |
+
**kwargs,
|
| 126 |
+
)
|
| 127 |
+
|
| 128 |
+
if cache_position[0] == 0:
|
| 129 |
+
model_inputs["pixel_values"] = pixel_values
|
| 130 |
+
|
| 131 |
+
return model_inputs
|
modeling_dots_vision.py
ADDED
|
@@ -0,0 +1,404 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import math
|
| 2 |
+
|
| 3 |
+
import torch
|
| 4 |
+
import torch.nn as nn
|
| 5 |
+
import torch.nn.functional as F
|
| 6 |
+
import torch.utils.checkpoint
|
| 7 |
+
from flash_attn import flash_attn_varlen_func
|
| 8 |
+
from torch.nn import LayerNorm
|
| 9 |
+
from transformers.modeling_utils import PreTrainedModel
|
| 10 |
+
from .configuration_dots import DotsVisionConfig
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
def rotate_half(x):
|
| 14 |
+
"""Rotates half the hidden dims of the input."""
|
| 15 |
+
x1 = x[..., : x.shape[-1] // 2]
|
| 16 |
+
x2 = x[..., x.shape[-1] // 2 :]
|
| 17 |
+
return torch.cat((-x2, x1), dim=-1)
|
| 18 |
+
|
| 19 |
+
|
| 20 |
+
def apply_rotary_pos_emb_vision(tensor: torch.Tensor, freqs: torch.Tensor) -> torch.Tensor:
|
| 21 |
+
orig_dtype = tensor.dtype
|
| 22 |
+
tensor = tensor.float()
|
| 23 |
+
|
| 24 |
+
cos = freqs.cos()
|
| 25 |
+
sin = freqs.sin()
|
| 26 |
+
|
| 27 |
+
cos = cos.unsqueeze(1).repeat(1, 1, 2).unsqueeze(0).float()
|
| 28 |
+
sin = sin.unsqueeze(1).repeat(1, 1, 2).unsqueeze(0).float()
|
| 29 |
+
|
| 30 |
+
output = (tensor * cos) + (rotate_half(tensor) * sin)
|
| 31 |
+
|
| 32 |
+
output = output.to(orig_dtype)
|
| 33 |
+
|
| 34 |
+
return output
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
class VisionRotaryEmbedding(nn.Module):
|
| 38 |
+
def __init__(self, dim: int, theta: float = 10000.0) -> None:
|
| 39 |
+
super().__init__()
|
| 40 |
+
inv_freq = 1.0 / (theta ** (torch.arange(0, dim, 2, dtype=torch.float) / dim))
|
| 41 |
+
self.register_buffer("inv_freq", inv_freq, persistent=False)
|
| 42 |
+
|
| 43 |
+
def forward(self, seqlen: int) -> torch.Tensor:
|
| 44 |
+
seq = torch.arange(seqlen, device=self.inv_freq.device, dtype=self.inv_freq.dtype)
|
| 45 |
+
freqs = torch.outer(seq, self.inv_freq)
|
| 46 |
+
return freqs
|
| 47 |
+
|
| 48 |
+
|
| 49 |
+
class PatchMerger(nn.Module):
|
| 50 |
+
def __init__(
|
| 51 |
+
self,
|
| 52 |
+
dim: int,
|
| 53 |
+
context_dim: int,
|
| 54 |
+
spatial_merge_size: int = 2,
|
| 55 |
+
pre_norm="layernorm",
|
| 56 |
+
init_merger_std=None,
|
| 57 |
+
) -> None:
|
| 58 |
+
super().__init__()
|
| 59 |
+
self.hidden_size = context_dim * (spatial_merge_size ** 2)
|
| 60 |
+
self.pre_norm = pre_norm
|
| 61 |
+
if self.pre_norm == "layernorm":
|
| 62 |
+
self.ln_q = LayerNorm(context_dim, eps=1e-6)
|
| 63 |
+
elif self.pre_norm == "rmsnorm":
|
| 64 |
+
self.ln_q = RMSNorm(context_dim, eps=1e-6)
|
| 65 |
+
else:
|
| 66 |
+
print("no norm in patch merger")
|
| 67 |
+
|
| 68 |
+
self.mlp = nn.Sequential(
|
| 69 |
+
nn.Linear(self.hidden_size, self.hidden_size),
|
| 70 |
+
nn.GELU(),
|
| 71 |
+
nn.Linear(self.hidden_size, dim),
|
| 72 |
+
)
|
| 73 |
+
|
| 74 |
+
if init_merger_std is not None:
|
| 75 |
+
nn.init.normal_(self.mlp[0].weight, mean=0.0, std=init_merger_std)
|
| 76 |
+
nn.init.zeros_(self.mlp[0].bias)
|
| 77 |
+
nn.init.normal_(self.mlp[2].weight, mean=0.0, std=init_merger_std)
|
| 78 |
+
nn.init.zeros_(self.mlp[2].bias)
|
| 79 |
+
|
| 80 |
+
def forward(self, x: torch.Tensor) -> torch.Tensor:
|
| 81 |
+
if self.pre_norm:
|
| 82 |
+
x = self.mlp(self.ln_q(x).view(-1, self.hidden_size))
|
| 83 |
+
else:
|
| 84 |
+
x = self.mlp(x.view(-1, self.hidden_size))
|
| 85 |
+
return x
|
| 86 |
+
|
| 87 |
+
|
| 88 |
+
class VisionAttention(nn.Module):
|
| 89 |
+
def __init__(self, config, dim: int, num_heads: int = 16, bias=True) -> None:
|
| 90 |
+
super().__init__()
|
| 91 |
+
self.num_heads = num_heads
|
| 92 |
+
self.head_dim = dim // num_heads
|
| 93 |
+
self.qkv = nn.Linear(dim, dim * 3, bias=bias)
|
| 94 |
+
self.proj = nn.Linear(dim, dim, bias=bias)
|
| 95 |
+
|
| 96 |
+
def forward(
|
| 97 |
+
self,
|
| 98 |
+
hidden_states: torch.Tensor,
|
| 99 |
+
cu_seqlens: torch.Tensor,
|
| 100 |
+
rotary_pos_emb: torch.Tensor = None,
|
| 101 |
+
) -> torch.Tensor:
|
| 102 |
+
seq_length = hidden_states.shape[0]
|
| 103 |
+
|
| 104 |
+
q, k, v = self.qkv(hidden_states).reshape(seq_length, 3, self.num_heads, -1).permute(1, 0, 2, 3).unbind(0)
|
| 105 |
+
q = apply_rotary_pos_emb_vision(q.unsqueeze(0), rotary_pos_emb).squeeze(0)
|
| 106 |
+
k = apply_rotary_pos_emb_vision(k.unsqueeze(0), rotary_pos_emb).squeeze(0)
|
| 107 |
+
|
| 108 |
+
attention_mask = torch.full(
|
| 109 |
+
[1, seq_length, seq_length], torch.finfo(q.dtype).min, device=q.device, dtype=q.dtype
|
| 110 |
+
)
|
| 111 |
+
for i in range(1, len(cu_seqlens)):
|
| 112 |
+
attention_mask[..., cu_seqlens[i - 1] : cu_seqlens[i], cu_seqlens[i - 1] : cu_seqlens[i]] = 0
|
| 113 |
+
|
| 114 |
+
q = q.transpose(0, 1)
|
| 115 |
+
k = k.transpose(0, 1)
|
| 116 |
+
v = v.transpose(0, 1)
|
| 117 |
+
attn_weights = torch.matmul(q, k.transpose(1, 2)) / math.sqrt(self.head_dim)
|
| 118 |
+
attn_weights = attn_weights + attention_mask
|
| 119 |
+
attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(q.dtype)
|
| 120 |
+
attn_output = torch.matmul(attn_weights, v)
|
| 121 |
+
attn_output = attn_output.transpose(0, 1)
|
| 122 |
+
attn_output = attn_output.reshape(seq_length, -1)
|
| 123 |
+
attn_output = self.proj(attn_output)
|
| 124 |
+
return attn_output
|
| 125 |
+
|
| 126 |
+
|
| 127 |
+
class VisionFlashAttention2(nn.Module):
|
| 128 |
+
def __init__(self, config, dim: int, num_heads: int = 16, bias=True) -> None:
|
| 129 |
+
super().__init__()
|
| 130 |
+
self.num_heads = num_heads
|
| 131 |
+
self.qkv = nn.Linear(dim, dim * 3, bias=bias)
|
| 132 |
+
self.proj = nn.Linear(dim, dim, bias=bias)
|
| 133 |
+
self.config = config
|
| 134 |
+
self.is_causal = config.is_causal
|
| 135 |
+
|
| 136 |
+
def forward(
|
| 137 |
+
self,
|
| 138 |
+
hidden_states: torch.Tensor,
|
| 139 |
+
cu_seqlens: torch.Tensor,
|
| 140 |
+
rotary_pos_emb: torch.Tensor = None,
|
| 141 |
+
) -> torch.Tensor:
|
| 142 |
+
seq_length = hidden_states.shape[0]
|
| 143 |
+
q, k, v = (
|
| 144 |
+
self.qkv(hidden_states).reshape(seq_length, 3, self.num_heads, -1).permute(1, 0, 2, 3).unbind(0)
|
| 145 |
+
) # 'shd'
|
| 146 |
+
q = apply_rotary_pos_emb_vision(q.unsqueeze(0), rotary_pos_emb).squeeze(0)
|
| 147 |
+
k = apply_rotary_pos_emb_vision(k.unsqueeze(0), rotary_pos_emb).squeeze(0)
|
| 148 |
+
max_seqlen = (cu_seqlens[1:] - cu_seqlens[:-1]).max().item()
|
| 149 |
+
attn_output = flash_attn_varlen_func(
|
| 150 |
+
q, k, v, cu_seqlens, cu_seqlens, max_seqlen, max_seqlen, causal=self.is_causal
|
| 151 |
+
).reshape(seq_length, -1)
|
| 152 |
+
attn_output = self.proj(attn_output)
|
| 153 |
+
|
| 154 |
+
return attn_output
|
| 155 |
+
|
| 156 |
+
|
| 157 |
+
class VisionSdpaAttention(nn.Module):
|
| 158 |
+
def __init__(self, config, dim: int, num_heads: int = 16, bias=True) -> None:
|
| 159 |
+
super().__init__()
|
| 160 |
+
self.num_heads = num_heads
|
| 161 |
+
self.qkv = nn.Linear(dim, dim * 3, bias=bias)
|
| 162 |
+
self.proj = nn.Linear(dim, dim, bias=bias)
|
| 163 |
+
self.config = config
|
| 164 |
+
|
| 165 |
+
def forward(
|
| 166 |
+
self,
|
| 167 |
+
hidden_states: torch.Tensor,
|
| 168 |
+
cu_seqlens: torch.Tensor,
|
| 169 |
+
rotary_pos_emb: torch.Tensor = None,
|
| 170 |
+
) -> torch.Tensor:
|
| 171 |
+
seq_length = hidden_states.shape[0]
|
| 172 |
+
q, k, v = self.qkv(hidden_states).reshape(seq_length, 3, self.num_heads, -1).permute(1, 0, 2, 3).unbind(0)
|
| 173 |
+
|
| 174 |
+
q = apply_rotary_pos_emb_vision(q.unsqueeze(0), rotary_pos_emb).squeeze(0)
|
| 175 |
+
k = apply_rotary_pos_emb_vision(k.unsqueeze(0), rotary_pos_emb).squeeze(0)
|
| 176 |
+
|
| 177 |
+
attention_mask = torch.zeros([1, seq_length, seq_length], device=q.device, dtype=torch.bool)
|
| 178 |
+
for i in range(1, len(cu_seqlens)):
|
| 179 |
+
attention_mask[..., cu_seqlens[i - 1] : cu_seqlens[i], cu_seqlens[i - 1] : cu_seqlens[i]] = True
|
| 180 |
+
|
| 181 |
+
q = q.transpose(0, 1)
|
| 182 |
+
k = k.transpose(0, 1)
|
| 183 |
+
v = v.transpose(0, 1)
|
| 184 |
+
|
| 185 |
+
attn_output = F.scaled_dot_product_attention(q, k, v, attention_mask, dropout_p=0.0)
|
| 186 |
+
attn_output = attn_output.transpose(0, 1)
|
| 187 |
+
attn_output = attn_output.reshape(seq_length, -1)
|
| 188 |
+
|
| 189 |
+
attn_output = self.proj(attn_output)
|
| 190 |
+
return attn_output
|
| 191 |
+
|
| 192 |
+
|
| 193 |
+
DOTS_VISION_ATTENTION_CLASSES = {
|
| 194 |
+
"eager": VisionAttention,
|
| 195 |
+
"flash_attention_2": VisionFlashAttention2,
|
| 196 |
+
"sdpa": VisionSdpaAttention,
|
| 197 |
+
}
|
| 198 |
+
|
| 199 |
+
|
| 200 |
+
class RMSNorm(nn.Module):
|
| 201 |
+
def __init__(self, dim: int, eps: float = 1e-6):
|
| 202 |
+
super().__init__()
|
| 203 |
+
self.weight = nn.Parameter(torch.ones(dim))
|
| 204 |
+
self.eps = eps
|
| 205 |
+
|
| 206 |
+
def forward(self, x: torch.Tensor) -> torch.Tensor:
|
| 207 |
+
output = self._norm(x.float()).type_as(x)
|
| 208 |
+
return output * self.weight
|
| 209 |
+
|
| 210 |
+
def extra_repr(self) -> str:
|
| 211 |
+
return f"{tuple(self.weight.shape)}, eps={self.eps}"
|
| 212 |
+
|
| 213 |
+
def _norm(self, x: torch.Tensor) -> torch.Tensor:
|
| 214 |
+
return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps)
|
| 215 |
+
|
| 216 |
+
|
| 217 |
+
class DotsSwiGLUFFN(nn.Module):
|
| 218 |
+
def __init__(self, config):
|
| 219 |
+
super().__init__()
|
| 220 |
+
hidden_features = config.intermediate_size
|
| 221 |
+
in_features = config.embed_dim
|
| 222 |
+
bias = config.use_bias
|
| 223 |
+
|
| 224 |
+
self.fc1 = nn.Linear(in_features, hidden_features, bias=bias)
|
| 225 |
+
self.fc2 = nn.Linear(hidden_features, in_features, bias=bias)
|
| 226 |
+
self.fc3 = nn.Linear(in_features, hidden_features, bias=bias)
|
| 227 |
+
|
| 228 |
+
def forward(self, x: torch.Tensor) -> torch.Tensor:
|
| 229 |
+
x = F.silu(self.fc1(x)) * self.fc3(x)
|
| 230 |
+
x = self.fc2(x)
|
| 231 |
+
return x
|
| 232 |
+
|
| 233 |
+
|
| 234 |
+
|
| 235 |
+
class DotsPatchEmbed(nn.Module):
|
| 236 |
+
def __init__(self, config):
|
| 237 |
+
super().__init__()
|
| 238 |
+
self.num_channels = config.num_channels
|
| 239 |
+
self.patch_size = config.patch_size
|
| 240 |
+
self.temporal_patch_size = config.temporal_patch_size
|
| 241 |
+
self.embed_dim = config.embed_dim
|
| 242 |
+
self.config = config
|
| 243 |
+
self.proj = nn.Conv2d(
|
| 244 |
+
config.num_channels,
|
| 245 |
+
config.embed_dim,
|
| 246 |
+
kernel_size=(config.patch_size, config.patch_size),
|
| 247 |
+
stride=(config.patch_size, config.patch_size),
|
| 248 |
+
)
|
| 249 |
+
self.norm = RMSNorm(config.embed_dim, eps=config.rms_norm_eps)
|
| 250 |
+
|
| 251 |
+
def forward(self, x: torch.Tensor, grid_thw=None) -> torch.Tensor:
|
| 252 |
+
x = x.view(-1, self.num_channels, self.temporal_patch_size, self.patch_size, self.patch_size)[:, :, 0]
|
| 253 |
+
x = self.proj(x).view(-1, self.embed_dim)
|
| 254 |
+
x = self.norm(x)
|
| 255 |
+
return x
|
| 256 |
+
|
| 257 |
+
|
| 258 |
+
class DotsViTPreprocessor(nn.Module):
|
| 259 |
+
def __init__(self, config):
|
| 260 |
+
super().__init__()
|
| 261 |
+
self.patch_h = config.patch_size
|
| 262 |
+
self.patch_w = config.patch_size
|
| 263 |
+
self.embed_dim = config.embed_dim
|
| 264 |
+
self.config = config
|
| 265 |
+
self.patchifier = DotsPatchEmbed(config)
|
| 266 |
+
|
| 267 |
+
def forward(self, x: torch.Tensor, grid_thw=None) -> torch.Tensor:
|
| 268 |
+
tokens = self.patchifier(x, grid_thw)
|
| 269 |
+
return tokens
|
| 270 |
+
|
| 271 |
+
|
| 272 |
+
class DotsVisionBlock(nn.Module):
|
| 273 |
+
def __init__(self, config, attn_implementation: str = "flash_attention_2"):
|
| 274 |
+
super().__init__()
|
| 275 |
+
self.attn = DOTS_VISION_ATTENTION_CLASSES[attn_implementation](
|
| 276 |
+
config, config.embed_dim, num_heads=config.num_attention_heads, bias=config.use_bias
|
| 277 |
+
)
|
| 278 |
+
self.norm1 = RMSNorm(config.embed_dim, eps=config.rms_norm_eps)
|
| 279 |
+
self.mlp = DotsSwiGLUFFN(config)
|
| 280 |
+
self.norm2 = RMSNorm(config.embed_dim, eps=config.rms_norm_eps)
|
| 281 |
+
|
| 282 |
+
def forward(self, hidden_states, cu_seqlens, rotary_pos_emb) -> torch.Tensor:
|
| 283 |
+
hidden_states = hidden_states + self.attn(
|
| 284 |
+
self.norm1(hidden_states), cu_seqlens=cu_seqlens, rotary_pos_emb=rotary_pos_emb
|
| 285 |
+
)
|
| 286 |
+
hidden_states = hidden_states + self.mlp(self.norm2(hidden_states))
|
| 287 |
+
return hidden_states
|
| 288 |
+
|
| 289 |
+
|
| 290 |
+
class DotsVisionTransformer(PreTrainedModel):
|
| 291 |
+
def __init__(self, config: DotsVisionConfig) -> None:
|
| 292 |
+
super().__init__(config)
|
| 293 |
+
self.config = config
|
| 294 |
+
self.spatial_merge_size = config.spatial_merge_size
|
| 295 |
+
|
| 296 |
+
self.patch_embed = DotsViTPreprocessor(config)
|
| 297 |
+
self._init_weights(self.patch_embed.patchifier.proj)
|
| 298 |
+
|
| 299 |
+
head_dim = config.embed_dim // config.num_attention_heads
|
| 300 |
+
|
| 301 |
+
self.rotary_pos_emb = VisionRotaryEmbedding(head_dim // 2)
|
| 302 |
+
|
| 303 |
+
_num_hidden_layers = config.num_hidden_layers
|
| 304 |
+
self.blocks = nn.ModuleList(
|
| 305 |
+
[DotsVisionBlock(config, config.attn_implementation) for _ in range(_num_hidden_layers)]
|
| 306 |
+
)
|
| 307 |
+
|
| 308 |
+
if self.config.post_norm:
|
| 309 |
+
self.post_trunk_norm = RMSNorm(config.embed_dim, eps=config.rms_norm_eps)
|
| 310 |
+
|
| 311 |
+
self.merger = PatchMerger(
|
| 312 |
+
dim=config.hidden_size,
|
| 313 |
+
context_dim=config.embed_dim,
|
| 314 |
+
spatial_merge_size=config.spatial_merge_size,
|
| 315 |
+
init_merger_std=self.config.init_merger_std,
|
| 316 |
+
)
|
| 317 |
+
|
| 318 |
+
self.gradient_checkpointing = False
|
| 319 |
+
self._gradient_checkpointing_func = torch.utils.checkpoint.checkpoint
|
| 320 |
+
|
| 321 |
+
def _init_weights(self, module):
|
| 322 |
+
std = self.config.initializer_range
|
| 323 |
+
if isinstance(module, (nn.Linear, nn.Conv3d)):
|
| 324 |
+
module.weight.data.normal_(mean=0.0, std=std)
|
| 325 |
+
if module.bias is not None:
|
| 326 |
+
module.bias.data.zero_()
|
| 327 |
+
elif isinstance(module, nn.Embedding):
|
| 328 |
+
module.weight.data.normal_(mean=0.0, std=std)
|
| 329 |
+
if module.padding_idx is not None:
|
| 330 |
+
module.weight.data[module.padding_idx].zero_()
|
| 331 |
+
|
| 332 |
+
@property
|
| 333 |
+
def dtype(self) -> torch.dtype:
|
| 334 |
+
return self.blocks[0].mlp.fc2.weight.dtype
|
| 335 |
+
|
| 336 |
+
@property
|
| 337 |
+
def device(self) -> torch.device:
|
| 338 |
+
return self.blocks[0].mlp.fc2.weight.device
|
| 339 |
+
|
| 340 |
+
def get_pos_ids_by_grid(self, grid_thw):
|
| 341 |
+
pos_ids = []
|
| 342 |
+
for t, h, w in grid_thw:
|
| 343 |
+
hpos_ids = torch.arange(h).unsqueeze(1).expand(-1, w)
|
| 344 |
+
hpos_ids = hpos_ids.reshape(
|
| 345 |
+
h // self.spatial_merge_size,
|
| 346 |
+
self.spatial_merge_size,
|
| 347 |
+
w // self.spatial_merge_size,
|
| 348 |
+
self.spatial_merge_size,
|
| 349 |
+
)
|
| 350 |
+
hpos_ids = hpos_ids.permute(0, 2, 1, 3)
|
| 351 |
+
hpos_ids = hpos_ids.flatten()
|
| 352 |
+
|
| 353 |
+
wpos_ids = torch.arange(w).unsqueeze(0).expand(h, -1)
|
| 354 |
+
wpos_ids = wpos_ids.reshape(
|
| 355 |
+
h // self.spatial_merge_size,
|
| 356 |
+
self.spatial_merge_size,
|
| 357 |
+
w // self.spatial_merge_size,
|
| 358 |
+
self.spatial_merge_size,
|
| 359 |
+
)
|
| 360 |
+
wpos_ids = wpos_ids.permute(0, 2, 1, 3)
|
| 361 |
+
wpos_ids = wpos_ids.flatten()
|
| 362 |
+
pos_ids.append(
|
| 363 |
+
torch.stack([hpos_ids, wpos_ids], dim=-1).repeat(t, 1)
|
| 364 |
+
)
|
| 365 |
+
|
| 366 |
+
return pos_ids
|
| 367 |
+
|
| 368 |
+
def rot_pos_emb(self, grid_thw):
|
| 369 |
+
pos_ids = self.get_pos_ids_by_grid(grid_thw)
|
| 370 |
+
pos_ids = torch.cat(pos_ids, dim=0)
|
| 371 |
+
max_grid_size = grid_thw[:, 1:].max()
|
| 372 |
+
rotary_pos_emb_full = self.rotary_pos_emb(max_grid_size)
|
| 373 |
+
rotary_pos_emb = rotary_pos_emb_full[pos_ids].flatten(1)
|
| 374 |
+
return rotary_pos_emb
|
| 375 |
+
|
| 376 |
+
def forward(self, hidden_states: torch.Tensor, grid_thw: torch.Tensor, bf16=True) -> torch.Tensor:
|
| 377 |
+
if bf16:
|
| 378 |
+
hidden_states = hidden_states.bfloat16()
|
| 379 |
+
hidden_states = self.patch_embed(hidden_states, grid_thw)
|
| 380 |
+
|
| 381 |
+
rotary_pos_emb = self.rot_pos_emb(grid_thw)
|
| 382 |
+
|
| 383 |
+
cu_seqlens = torch.repeat_interleave(grid_thw[:, 1] * grid_thw[:, 2], grid_thw[:, 0]).cumsum(
|
| 384 |
+
dim=0,
|
| 385 |
+
dtype=grid_thw.dtype if torch.jit.is_tracing() else torch.int32,
|
| 386 |
+
)
|
| 387 |
+
cu_seqlens = F.pad(cu_seqlens, (1, 0), value=0)
|
| 388 |
+
|
| 389 |
+
for blk in self.blocks:
|
| 390 |
+
if self.gradient_checkpointing and self.training:
|
| 391 |
+
hidden_states = self._gradient_checkpointing_func(
|
| 392 |
+
blk.__call__,
|
| 393 |
+
hidden_states,
|
| 394 |
+
cu_seqlens,
|
| 395 |
+
rotary_pos_emb,
|
| 396 |
+
)
|
| 397 |
+
else:
|
| 398 |
+
hidden_states = blk(hidden_states, cu_seqlens=cu_seqlens, rotary_pos_emb=rotary_pos_emb)
|
| 399 |
+
|
| 400 |
+
if self.config.post_norm:
|
| 401 |
+
hidden_states = self.post_trunk_norm(hidden_states)
|
| 402 |
+
|
| 403 |
+
hidden_states = self.merger(hidden_states)
|
| 404 |
+
return hidden_states
|
preprocessor_config.json
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"auto_map": {
|
| 3 |
+
"AutoProcessor": "configuration_dots.DotsVLProcessor"
|
| 4 |
+
},
|
| 5 |
+
"min_pixels": 3136,
|
| 6 |
+
"max_pixels": 11289600,
|
| 7 |
+
"patch_size": 14,
|
| 8 |
+
"temporal_patch_size": 1,
|
| 9 |
+
"merge_size": 2,
|
| 10 |
+
"image_mean": [
|
| 11 |
+
0.48145466,
|
| 12 |
+
0.4578275,
|
| 13 |
+
0.40821073
|
| 14 |
+
],
|
| 15 |
+
"image_std": [
|
| 16 |
+
0.26862954,
|
| 17 |
+
0.26130258,
|
| 18 |
+
0.27577711
|
| 19 |
+
],
|
| 20 |
+
"image_processor_type": "Qwen2VLImageProcessor",
|
| 21 |
+
"processor_class": "DotsVLProcessor"
|
| 22 |
+
}
|
processor_config.json
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"image_processor": {
|
| 3 |
+
"auto_map": {
|
| 4 |
+
"AutoProcessor": "configuration_dots.DotsVLProcessor"
|
| 5 |
+
},
|
| 6 |
+
"do_convert_rgb": true,
|
| 7 |
+
"do_normalize": true,
|
| 8 |
+
"do_rescale": true,
|
| 9 |
+
"do_resize": true,
|
| 10 |
+
"image_mean": [
|
| 11 |
+
0.48145466,
|
| 12 |
+
0.4578275,
|
| 13 |
+
0.40821073
|
| 14 |
+
],
|
| 15 |
+
"image_processor_type": "Qwen2VLImageProcessor",
|
| 16 |
+
"image_std": [
|
| 17 |
+
0.26862954,
|
| 18 |
+
0.26130258,
|
| 19 |
+
0.27577711
|
| 20 |
+
],
|
| 21 |
+
"max_pixels": 11289600,
|
| 22 |
+
"merge_size": 2,
|
| 23 |
+
"min_pixels": 3136,
|
| 24 |
+
"patch_size": 14,
|
| 25 |
+
"resample": 3,
|
| 26 |
+
"rescale_factor": 0.00392156862745098,
|
| 27 |
+
"size": {
|
| 28 |
+
"longest_edge": 11289600,
|
| 29 |
+
"shortest_edge": 3136
|
| 30 |
+
},
|
| 31 |
+
"temporal_patch_size": 1
|
| 32 |
+
},
|
| 33 |
+
"processor_class": "DotsVLProcessor"
|
| 34 |
+
}
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"additional_special_tokens": [
|
| 3 |
+
"<|im_start|>",
|
| 4 |
+
"<|im_end|>",
|
| 5 |
+
"<|object_ref_start|>",
|
| 6 |
+
"<|object_ref_end|>",
|
| 7 |
+
"<|box_start|>",
|
| 8 |
+
"<|box_end|>",
|
| 9 |
+
"<|quad_start|>",
|
| 10 |
+
"<|quad_end|>",
|
| 11 |
+
"<|vision_start|>",
|
| 12 |
+
"<|vision_end|>",
|
| 13 |
+
"<|vision_pad|>",
|
| 14 |
+
"<|image_pad|>",
|
| 15 |
+
"<|video_pad|>"
|
| 16 |
+
],
|
| 17 |
+
"eos_token": {
|
| 18 |
+
"content": "<|endoftext|>",
|
| 19 |
+
"lstrip": false,
|
| 20 |
+
"normalized": false,
|
| 21 |
+
"rstrip": false,
|
| 22 |
+
"single_word": false
|
| 23 |
+
},
|
| 24 |
+
"pad_token": "[PAD]"
|
| 25 |
+
}
|
tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:904d81ff0cfa066dbc0b6a21e10ded6ebb7c2d8df14100d851f90bb7878bd5de
|
| 3 |
+
size 11426251
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_prefix_space": false,
|
| 3 |
+
"backend": "tokenizers",
|
| 4 |
+
"bos_token": null,
|
| 5 |
+
"clean_up_tokenization_spaces": false,
|
| 6 |
+
"eos_token": "<|endoftext|>",
|
| 7 |
+
"errors": "replace",
|
| 8 |
+
"extra_special_tokens": [
|
| 9 |
+
"<|im_start|>",
|
| 10 |
+
"<|im_end|>",
|
| 11 |
+
"<|object_ref_start|>",
|
| 12 |
+
"<|object_ref_end|>",
|
| 13 |
+
"<|box_start|>",
|
| 14 |
+
"<|box_end|>",
|
| 15 |
+
"<|quad_start|>",
|
| 16 |
+
"<|quad_end|>",
|
| 17 |
+
"<|vision_start|>",
|
| 18 |
+
"<|vision_end|>",
|
| 19 |
+
"<|vision_pad|>",
|
| 20 |
+
"<|image_pad|>",
|
| 21 |
+
"<|video_pad|>"
|
| 22 |
+
],
|
| 23 |
+
"is_local": true,
|
| 24 |
+
"model_max_length": 131072,
|
| 25 |
+
"pad_token": "[PAD]",
|
| 26 |
+
"processor_class": "DotsVLProcessor",
|
| 27 |
+
"split_special_tokens": false,
|
| 28 |
+
"tokenizer_class": "TokenizersBackend",
|
| 29 |
+
"unk_token": null
|
| 30 |
+
}
|
vocab.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|