wenhuchen commited on
Commit
a4a785c
1 Parent(s): de9f67d
app.py CHANGED
@@ -101,4 +101,4 @@ if __name__ == "__main__":
101
 
102
  elo_results_file, leaderboard_table_file = load_elo_results(elo_results_dir)
103
  demo = build_combine_demo(models, elo_results_file, leaderboard_table_file)
104
- demo.queue(max_size=20).launch(server_port=server_port, root_path=ROOT_PATH)
 
101
 
102
  elo_results_file, leaderboard_table_file = load_elo_results(elo_results_dir)
103
  demo = build_combine_demo(models, elo_results_file, leaderboard_table_file)
104
+ demo.queue(max_size=20).launch(server_port=server_port, root_path=ROOT_PATH)
arena_elo/elo_rating/elo_analysis.py CHANGED
@@ -171,20 +171,8 @@ def visualize_pairwise_win_fraction(battles, model_order):
171
  xaxis_side="top",
172
  title_y=0.07,
173
  title_x=0.5,
174
- # xaxis=dict(
175
- # tickfont=dict(size=16),
176
- # title=dict(font=dict(size=16)),
177
- # ),
178
- # yaxis=dict(
179
- # tickfont=dict(size=16),
180
- # title=dict(font=dict(size=16)),
181
- # ),
182
  )
183
  fig.update_traces(
184
- # textfont=dict(size=16),
185
- # colorbar=dict(
186
- # title=dict(font=dict(size=16))
187
- # ),
188
  hovertemplate="Model A: %{y}<br>Model B: %{x}<br>Fraction of A Wins: %{z}<extra></extra>"
189
  )
190
 
@@ -208,20 +196,8 @@ def visualize_battle_count(battles, model_order):
208
  xaxis_side="top",
209
  title_y=0.07,
210
  title_x=0.5,
211
- # xaxis=dict(
212
- # tickfont=dict(size=16),
213
- # title=dict(font=dict(size=16)),
214
- # ),
215
- # yaxis=dict(
216
- # tickfont=dict(size=16),
217
- # title=dict(font=dict(size=16)),
218
- # ),
219
  )
220
  fig.update_traces(
221
- # textfont=dict(size=16),
222
- # colorbar=dict(
223
- # title=dict(font=dict(size=16))
224
- # ),
225
  hovertemplate="Model A: %{y}<br>Model B: %{x}<br>Count: %{z}<extra></extra>"
226
  )
227
  return fig
@@ -239,14 +215,6 @@ def visualize_average_win_rate(battles, limit_show_number):
239
  )
240
  fig.update_layout(
241
  yaxis_title="Average Win Rate", xaxis_title="Model", showlegend=False,
242
- # xaxis=dict(
243
- # tickfont=dict(size=16),
244
- # title=dict(font=dict(size=16)),
245
- # ),
246
- # yaxis=dict(
247
- # tickfont=dict(size=16),
248
- # title=dict(font=dict(size=16)),
249
- # ),
250
  )
251
  fig.update_traces(textfont_size=16)
252
  return fig
@@ -278,16 +246,7 @@ def visualize_bootstrap_elo_rating(df, df_final, limit_show_number):
278
  height=500,
279
  width=700,
280
  )
281
- fig.update_layout(xaxis_title="Model", yaxis_title="Rating",
282
- # xaxis=dict(
283
- # tickfont=dict(size=16),
284
- # title=dict(font=dict(size=16)),
285
- # ),
286
- # yaxis=dict(
287
- # tickfont=dict(size=16),
288
- # title=dict(font=dict(size=16)),
289
- # ),
290
- )
291
  fig.update_traces(textfont_size=16)
292
  return fig
293
 
@@ -406,9 +365,6 @@ if __name__ == "__main__":
406
  battles_with_enough_battles = battles[
407
  ~battles["model_a"].isin(to_remove_models) & ~battles["model_b"].isin(to_remove_models)
408
  ]
409
- # battles_with_enough_battles = [
410
- # battle for battle in battles if battle["model_a"] not in to_remove_models and battle["model_b"] not in to_remove_models
411
- # ]
412
  print(f"Remove models with less than {args.min_num_battles_per_model} battles: {to_remove_models}")
413
  print(f"Number of battles: {len(battles)} -> {len(battles_with_enough_battles)}")
414
  battles = battles_with_enough_battles
 
171
  xaxis_side="top",
172
  title_y=0.07,
173
  title_x=0.5,
 
 
 
 
 
 
 
 
174
  )
175
  fig.update_traces(
 
 
 
 
176
  hovertemplate="Model A: %{y}<br>Model B: %{x}<br>Fraction of A Wins: %{z}<extra></extra>"
177
  )
178
 
 
196
  xaxis_side="top",
197
  title_y=0.07,
198
  title_x=0.5,
 
 
 
 
 
 
 
 
199
  )
200
  fig.update_traces(
 
 
 
 
201
  hovertemplate="Model A: %{y}<br>Model B: %{x}<br>Count: %{z}<extra></extra>"
202
  )
203
  return fig
 
215
  )
216
  fig.update_layout(
217
  yaxis_title="Average Win Rate", xaxis_title="Model", showlegend=False,
 
 
 
 
 
 
 
 
218
  )
219
  fig.update_traces(textfont_size=16)
220
  return fig
 
246
  height=500,
247
  width=700,
248
  )
249
+ fig.update_layout(xaxis_title="Model", yaxis_title="Rating")
 
 
 
 
 
 
 
 
 
250
  fig.update_traces(textfont_size=16)
251
  return fig
252
 
 
365
  battles_with_enough_battles = battles[
366
  ~battles["model_a"].isin(to_remove_models) & ~battles["model_b"].isin(to_remove_models)
367
  ]
 
 
 
368
  print(f"Remove models with less than {args.min_num_battles_per_model} battles: {to_remove_models}")
369
  print(f"Number of battles: {len(battles)} -> {len(battles_with_enough_battles)}")
370
  battles = battles_with_enough_battles
arena_elo/elo_rating/generate_leaderboard.py CHANGED
@@ -43,10 +43,6 @@ def main(
43
  model_info[model]["Arena Elo rating (full)"] = full_leaderboard_data.loc[model, "rating"]
44
  else:
45
  model_info[model]["Arena Elo rating (full)"] = 0
46
- # if model in anony_leaderboard_data.index:
47
- # model_info[model]["Arena Elo rating"] = anony_leaderboard_data.loc[model, "rating"]
48
- # else:
49
- # model_info[model]["Arena Elo rating"] = 0
50
 
51
  final_model_info = {}
52
  for model in model_info:
 
43
  model_info[model]["Arena Elo rating (full)"] = full_leaderboard_data.loc[model, "rating"]
44
  else:
45
  model_info[model]["Arena Elo rating (full)"] = 0
 
 
 
 
46
 
47
  final_model_info = {}
48
  for model in model_info:
arena_elo/results/latest/video_generation_leaderboard.csv CHANGED
@@ -1,11 +1,11 @@
1
  key,Model,Arena Elo rating (anony),Arena Elo rating (full),License,Organization,Link
2
- StableVideoDiffusion,StableVideoDiffusion,1143.2191298083312,1147.5193110801638,stable-video-diffusion-nc-community,Stability AI,https://fal.ai/models/fal-ai/fast-svd/text-to-video/api
3
  T2VTurbo,T2VTurbo,1092.4551860072625,1093.43771637836,cc-by-nc-4.0,"University of California, Santa Barbara",https://huggingface.co/jiachenli-ucsb/T2V-Turbo-VC2
4
  VideoCrafter2,VideoCrafter2,1076.7797371219958,1079.3282160297892,Apache 2.0,Tencent AI Lab,https://ailab-cvc.github.io/videocrafter2/
5
  AnimateDiff,AnimateDiff,1071.9888858847867,1073.0390578256142,creativeml-openrail-m,"The Chinese University of Hong Kong, Shanghai AI Lab, Stanford University",https://fal.ai/models/fast-animatediff-t2v
6
- CogVideoX,CogVideoX,1046.417835182279,1048.070311411077,N/A,N/A,N/A
7
  LaVie,LaVie,999.6270938177271,1001.6816790561326,Apache 2.0,Shanghai AI Lab,https://github.com/Vchitect/LaVie
8
- OpenSora12,OpenSora12,924.0203683794037,903.7447677401199,N/A,N/A,N/A
9
  OpenSora,OpenSora,916.0498767986071,917.9961732110512,Apache 2.0,HPC-AI Tech,https://github.com/hpcaitech/Open-Sora
10
  ModelScope,ModelScope,865.3398542588171,868.9309810684535,cc-by-nc-4.0,Alibaba Group,https://arxiv.org/abs/2308.06571
11
  AnimateDiffTurbo,AnimateDiffTurbo,864.1020327407906,866.2517861992403,creativeml-openrail-m,"The Chinese University of Hong Kong, Shanghai AI Lab, Stanford University",https://fal.ai/models/fast-animatediff-t2v-turbo
 
1
  key,Model,Arena Elo rating (anony),Arena Elo rating (full),License,Organization,Link
2
+ StableVideoDiffusion,StableVideoDiffusion,1143.2191298083312,1147.5193110801638,SVD-nc-community,Stability AI,https://fal.ai/models/fal-ai/fast-svd/text-to-video/api
3
  T2VTurbo,T2VTurbo,1092.4551860072625,1093.43771637836,cc-by-nc-4.0,"University of California, Santa Barbara",https://huggingface.co/jiachenli-ucsb/T2V-Turbo-VC2
4
  VideoCrafter2,VideoCrafter2,1076.7797371219958,1079.3282160297892,Apache 2.0,Tencent AI Lab,https://ailab-cvc.github.io/videocrafter2/
5
  AnimateDiff,AnimateDiff,1071.9888858847867,1073.0390578256142,creativeml-openrail-m,"The Chinese University of Hong Kong, Shanghai AI Lab, Stanford University",https://fal.ai/models/fast-animatediff-t2v
6
+ CogVideoX,CogVideoX,1046.417835182279,1048.070311411077,CogVideoX License,Tsinghua University,https://huggingface.co/THUDM/CogVideoX-2b
7
  LaVie,LaVie,999.6270938177271,1001.6816790561326,Apache 2.0,Shanghai AI Lab,https://github.com/Vchitect/LaVie
8
+ OpenSora12,OpenSora12,924.0203683794037,903.7447677401199,Apache 2.0,HPC-AI Tech,https://github.com/hpcaitech/Open-Sora
9
  OpenSora,OpenSora,916.0498767986071,917.9961732110512,Apache 2.0,HPC-AI Tech,https://github.com/hpcaitech/Open-Sora
10
  ModelScope,ModelScope,865.3398542588171,868.9309810684535,cc-by-nc-4.0,Alibaba Group,https://arxiv.org/abs/2308.06571
11
  AnimateDiffTurbo,AnimateDiffTurbo,864.1020327407906,866.2517861992403,creativeml-openrail-m,"The Chinese University of Hong Kong, Shanghai AI Lab, Stanford University",https://fal.ai/models/fast-animatediff-t2v-turbo
arena_elo/video_generation_model_info.json CHANGED
@@ -24,11 +24,21 @@
24
  "License": "cc-by-nc-4.0",
25
  "Organization": "Alibaba Group"
26
  },
 
 
 
 
 
27
  "OpenSora": {
28
  "Link": "https://github.com/hpcaitech/Open-Sora",
29
  "License": "Apache 2.0",
30
  "Organization": "HPC-AI Tech"
31
  },
 
 
 
 
 
32
  "StableVideoDiffusion": {
33
  "Link": "https://fal.ai/models/fal-ai/fast-svd/text-to-video/api",
34
  "License": "SVD-nc-community",
 
24
  "License": "cc-by-nc-4.0",
25
  "Organization": "Alibaba Group"
26
  },
27
+ "OpenSora12": {
28
+ "Link": "https://github.com/hpcaitech/Open-Sora",
29
+ "License": "Apache 2.0",
30
+ "Organization": "HPC-AI Tech"
31
+ },
32
  "OpenSora": {
33
  "Link": "https://github.com/hpcaitech/Open-Sora",
34
  "License": "Apache 2.0",
35
  "Organization": "HPC-AI Tech"
36
  },
37
+ "CogVideoX": {
38
+ "Link": "https://huggingface.co/THUDM/CogVideoX-2b",
39
+ "License": "CogVideoX LICENSE",
40
+ "Organization": "THUDM"
41
+ },
42
  "StableVideoDiffusion": {
43
  "Link": "https://fal.ai/models/fal-ai/fast-svd/text-to-video/api",
44
  "License": "SVD-nc-community",