nielsr HF Staff commited on
Commit
855f2bf
·
verified ·
1 Parent(s): af492a0

Add library name and paper link to model card

Browse files

This PR adds the `library_name` to the metadata section of the model card, clarifying that this model works with the Hugging Face `transformers` library. It also adds a link to the paper on Hugging Face.

Files changed (1) hide show
  1. README.md +11 -5
README.md CHANGED
@@ -1,11 +1,13 @@
1
  ---
2
  license: llama2
3
  pipeline_tag: video-text-to-text
 
4
  ---
 
5
  # Chat-UniVi: Unified Visual Representation Empowers Large Language Models with Image and Video Understanding
6
 
7
  **Paper or resources for more information:**
8
- [[Paper](https://huggingface.co/papers/2311.08046)] [[Code](https://github.com/PKU-YuanGroup/Chat-UniVi)]
9
 
10
  ## License
11
  Llama 2 is licensed under the LLAMA 2 Community License,
@@ -133,9 +135,11 @@ if __name__ == '__main__':
133
 
134
  cur_prompt = qs
135
  if model.config.mm_use_im_start_end:
136
- qs = DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_TOKEN * slice_len + DEFAULT_IM_END_TOKEN + '\n' + qs
 
137
  else:
138
- qs = DEFAULT_IMAGE_TOKEN * slice_len + '\n' + qs
 
139
 
140
  conv = conv_templates[conv_mode].copy()
141
  conv.append_message(conv.roles[0], qs)
@@ -224,9 +228,11 @@ if __name__ == '__main__':
224
  if image_path is not None:
225
  cur_prompt = qs
226
  if model.config.mm_use_im_start_end:
227
- qs = DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_TOKEN + DEFAULT_IM_END_TOKEN + '\n' + qs
 
228
  else:
229
- qs = DEFAULT_IMAGE_TOKEN + '\n' + qs
 
230
 
231
  conv = conv_templates[conv_mode].copy()
232
  conv.append_message(conv.roles[0], qs)
 
1
  ---
2
  license: llama2
3
  pipeline_tag: video-text-to-text
4
+ library_name: transformers
5
  ---
6
+
7
  # Chat-UniVi: Unified Visual Representation Empowers Large Language Models with Image and Video Understanding
8
 
9
  **Paper or resources for more information:**
10
+ [[Paper](https://huggingface.co/papers/2503.04504)] [[Code](https://github.com/PKU-YuanGroup/Chat-UniVi)]
11
 
12
  ## License
13
  Llama 2 is licensed under the LLAMA 2 Community License,
 
135
 
136
  cur_prompt = qs
137
  if model.config.mm_use_im_start_end:
138
+ qs = DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_TOKEN * slice_len + DEFAULT_IM_END_TOKEN + '
139
+ ' + qs
140
  else:
141
+ qs = DEFAULT_IMAGE_TOKEN * slice_len + '
142
+ ' + qs
143
 
144
  conv = conv_templates[conv_mode].copy()
145
  conv.append_message(conv.roles[0], qs)
 
228
  if image_path is not None:
229
  cur_prompt = qs
230
  if model.config.mm_use_im_start_end:
231
+ qs = DEFAULT_IM_START_TOKEN + DEFAULT_IMAGE_TOKEN + DEFAULT_IM_END_TOKEN + '
232
+ ' + qs
233
  else:
234
+ qs = DEFAULT_IMAGE_TOKEN + '
235
+ ' + qs
236
 
237
  conv = conv_templates[conv_mode].copy()
238
  conv.append_message(conv.roles[0], qs)