Hugging Face's logo

llava-hf
/
llava-onevision-qwen2-7b-ov-hf

Image-Text-to-Text
Safetensors
English
Chinese
llava_onevision
vision
conversational
Model card Files Community
8
RaushanTurganbay HF staff commited on
Commit
bd59bb6
·
verified ·
1 Parent(s): 319e304

Upload LlavaOnevisionForConditionalGeneration

Browse files
Files changed (3) hide show
  1. config.json +4 -9
  2. model-00001-of-00004.safetensors +1 -1
  3. model-00004-of-00004.safetensors +1 -1
config.json CHANGED
@@ -1,7 +1,7 @@
1
  {
2
- "_name_or_path": "/raid/raushan/ov-7b",
3
  "architectures": [
4
- "LlavaNextForConditionalGeneration"
5
  ],
6
  "ignore_index": -100,
7
  "image_grid_pinpoints": [
@@ -151,7 +151,7 @@
151
  ]
152
  ],
153
  "image_token_index": 151646,
154
- "model_type": "llava_next",
155
  "projector_hidden_act": "gelu",
156
  "text_config": {
157
  "_name_or_path": "Qwen/Qwen2-7B-Instruct",
@@ -162,29 +162,24 @@
162
  "eos_token_id": 151645,
163
  "hidden_size": 3584,
164
  "intermediate_size": 18944,
165
- "max_position_embeddings": 32768,
166
- "max_window_layers": 28,
167
  "model_type": "qwen2",
168
  "num_attention_heads": 28,
169
  "num_hidden_layers": 28,
170
  "num_key_value_heads": 4,
171
  "rope_theta": 1000000.0,
172
- "sliding_window": null,
173
  "torch_dtype": "bfloat16",
174
- "use_sliding_window": false,
175
  "vocab_size": 152128
176
  },
177
  "tie_word_embeddings": false,
178
  "torch_dtype": "float16",
179
  "transformers_version": "4.45.0.dev0",
180
  "use_image_newline_parameter": true,
 
181
  "vision_aspect_ratio": "anyres_max_9",
182
  "vision_config": {
183
- "hidden_act": "gelu_pytorch_tanh",
184
  "hidden_size": 1152,
185
  "image_size": 384,
186
  "intermediate_size": 4304,
187
- "layer_norm_eps": 1e-06,
188
  "model_type": "siglip_vision_model",
189
  "num_attention_heads": 16,
190
  "num_hidden_layers": 26,
 
1
  {
2
+ "_name_or_path": "/raid/raushan/si-7b",
3
  "architectures": [
4
+ "LlavaOnevisionForConditionalGeneration"
5
  ],
6
  "ignore_index": -100,
7
  "image_grid_pinpoints": [
 
151
  ]
152
  ],
153
  "image_token_index": 151646,
154
+ "model_type": "llava_onevision",
155
  "projector_hidden_act": "gelu",
156
  "text_config": {
157
  "_name_or_path": "Qwen/Qwen2-7B-Instruct",
 
162
  "eos_token_id": 151645,
163
  "hidden_size": 3584,
164
  "intermediate_size": 18944,
 
 
165
  "model_type": "qwen2",
166
  "num_attention_heads": 28,
167
  "num_hidden_layers": 28,
168
  "num_key_value_heads": 4,
169
  "rope_theta": 1000000.0,
 
170
  "torch_dtype": "bfloat16",
 
171
  "vocab_size": 152128
172
  },
173
  "tie_word_embeddings": false,
174
  "torch_dtype": "float16",
175
  "transformers_version": "4.45.0.dev0",
176
  "use_image_newline_parameter": true,
177
+ "video_token_index": 151647,
178
  "vision_aspect_ratio": "anyres_max_9",
179
  "vision_config": {
 
180
  "hidden_size": 1152,
181
  "image_size": 384,
182
  "intermediate_size": 4304,
 
183
  "model_type": "siglip_vision_model",
184
  "num_attention_heads": 16,
185
  "num_hidden_layers": 26,
model-00001-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:cf7993771c8a5d9e67031d4ee564de16af03149c94a41fb022b7f837cbff2929
3
  size 4911200360
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad4cb27d2a3f6e987147619cf2508a5bbb1b6b4c2d5533501fcc4a287a90e5af
3
  size 4911200360
model-00004-of-00004.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:648b3aad9870352447111467fbae2393c5482c492f9cbc7f99c7d21b0a3ff9f2
3
  size 1226266240
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2d1e40e47e9b524efcbbcab68e97fe4e7aadfe9a51bc14b54cc2331c48ebff6
3
  size 1226266240