Merge branch 'main' of https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena

README.md

@@ -7,7 +7,7 @@ sdk: gradio
 sdk_version: 4.41.0
 python_version: 3.12
 app_file: app.py
-pinned: false
+pinned: true
 license: mit
 tags:
 - arena

app.py

@@ -97,7 +97,7 @@ if __name__ == "__main__":
     root_path = ROOT_PATH
     elo_results_dir = ELO_RESULTS_DIR
     models = ModelManager(enable_nsfw=False, do_pre_download=True, do_debug_packages=True)
-    # models = ModelManager(enable_nsfw=False, pre_download=False, debug_packages=False)
+    # models = ModelManager(enable_nsfw=False, do_pre_download=False, do_debug_packages=False)
     
     elo_results_file, leaderboard_table_file = load_elo_results(elo_results_dir)
     demo = build_combine_demo(models, elo_results_file, leaderboard_table_file)

arena_elo/video_generation_model_info.json

@@ -31,7 +31,7 @@
     },
     "StableVideoDiffusion": {
         "Link": "https://fal.ai/models/fal-ai/fast-svd/text-to-video/api",
-        "License": "stable-video-diffusion-nc-community",
+        "License": "SVD-nc-community",
         "Organization": "Stability AI"
     },
     "T2VTurbo": {

model/model_registry.py

@@ -258,15 +258,6 @@ register_model_info(
     "AnimateDiff Turbo is a lightning version of AnimateDiff.",
 )
 
-"""
-register_model_info(
-    ["videogenhub_LaVie_generation"],
-    "LaVie",
-    "https://github.com/Vchitect/LaVie",
-    "LaVie is a video generation model with cascaded latent diffusion models.",
-)
-
-
 register_model_info(
     ["videogenhub_VideoCrafter2_generation"],
     "VideoCrafter2",
@@ -274,6 +265,13 @@ register_model_info(
     "VideoCrafter2 is a T2V model that disentangling motion from appearance.",
 )
 
+"""
+register_model_info(
+    ["videogenhub_LaVie_generation"],
+    "LaVie",
+    "https://github.com/Vchitect/LaVie",
+    "LaVie is a video generation model with cascaded latent diffusion models.",
+)
 register_model_info(
     ["videogenhub_ModelScope_generation"],
     "ModelScope",
@@ -303,7 +301,7 @@ register_model_info(
 )
     
 register_model_info(
-    ["fal_T2VTurbo_generation"],
+    ["fal_T2VTurbo_text2video"],
     "T2V-Turbo",
     "https://github.com/Ji4chenLi/t2v-turbo",
     "Video Consistency Model with Mixed Reward Feedback.",

model/models/__init__.py

@@ -18,7 +18,7 @@ IMAGE_EDITION_MODELS = ['imagenhub_CycleDiffusion_edition', 'imagenhub_Pix2PixZe
 VIDEO_GENERATION_MODELS = ['fal_AnimateDiff_text2video',
                            'fal_AnimateDiffTurbo_text2video',
                            #'videogenhub_LaVie_generation', 
-                           #'videogenhub_VideoCrafter2_generation',
+                           'videogenhub_VideoCrafter2_generation',
                            #'videogenhub_ModelScope_generation', 
                            'videogenhub_CogVideoX_generation', 'videogenhub_OpenSora12_generation',
                            #'videogenhub_OpenSora_generation', 

model/models/fal_api_models.py

@@ -54,24 +54,6 @@ class FalModel():
             return result
         elif self.model_type == "image2image":
             raise NotImplementedError("image2image model is not implemented yet")
-            # assert "image" in kwargs or "image_url" in kwargs, "image or image_url is required for image2image model"
-            # if "image" in kwargs:
-            #     image_url = None
-            #     pass
-            # handler = fal_client.submit(
-            #     f"fal-ai/{self.model_name}",
-            #     arguments={
-            #         "image_url": image_url
-            #     },
-            # )
-            #
-            # for event in handler.iter_events():
-            #     if isinstance(event, fal_client.InProgress):
-            #         print('Request in progress')
-            #         print(event.logs)
-            #
-            # result = handler.get()
-            # return result
         elif self.model_type == "text2video":
             assert "prompt" in kwargs, "prompt is required for text2video model"
             if self.model_name == 'AnimateDiff':

requirements.txt

@@ -68,4 +68,4 @@ tensorboard
 timm
 wandb
 pandarallel
-kaleido
+kaleido

serve/leaderboard.py

@@ -22,20 +22,6 @@ basic_component_values = [None] * 6
 leader_component_values = [None] * 5
 
 
-# def make_leaderboard_md(elo_results):
-#     leaderboard_md = f"""
-# # 🏆 Chatbot Arena Leaderboard
-# | [Blog](https://lmsys.org/blog/2023-05-03-arena/) | [GitHub](https://github.com/lm-sys/FastChat) | [Paper](https://arxiv.org/abs/2306.05685) | [Dataset](https://github.com/lm-sys/FastChat/blob/main/docs/dataset_release.md) | [Twitter](https://twitter.com/lmsysorg) | [Discord](https://discord.gg/HSWAKCrnFx) |
-
-# This leaderboard is based on the following three benchmarks.
-# - [Chatbot Arena](https://lmsys.org/blog/2023-05-03-arena/) - a crowdsourced, randomized battle platform. We use 100K+ user votes to compute Elo ratings.
-# - [MT-Bench](https://arxiv.org/abs/2306.05685) - a set of challenging multi-turn questions. We use GPT-4 to grade the model responses.
-# - [MMLU](https://arxiv.org/abs/2009.03300) (5-shot) - a test to measure a model's multitask accuracy on 57 tasks.
-
-# 💻 Code: The Arena Elo ratings are computed by this [notebook]({notebook_url}). The MT-bench scores (single-answer grading on a scale of 10) are computed by [fastchat.llm_judge](https://github.com/lm-sys/FastChat/tree/main/fastchat/llm_judge). The MMLU scores are mostly computed by [InstructEval](https://github.com/declare-lab/instruct-eval). Higher values are better for all benchmarks. Empty cells mean not available. Last updated: November, 2023.
-# """
-#     return leaderboard_md
-
 def make_leaderboard_md(elo_results):
     leaderboard_md = f"""
 # 🏆 GenAI-Arena Leaderboard
@@ -324,31 +310,6 @@ def build_leaderboard_tab(elo_results_file, leaderboard_table_file, show_plot=Tr
 
     leader_component_values[:] = [md, p1, p2, p3, p4]
 
-    """
-    with gr.Row():
-        with gr.Column():
-            gr.Markdown(
-                "#### Figure 1: Fraction of Model A Wins for All Non-tied A vs. B Battles"
-            )
-            plot_1 = gr.Plot(p1, show_label=False)
-        with gr.Column():
-            gr.Markdown(
-                "#### Figure 2: Battle Count for Each Combination of Models (without Ties)"
-            )
-            plot_2 = gr.Plot(p2, show_label=False)
-    with gr.Row():
-        with gr.Column():
-            gr.Markdown(
-                "#### Figure 3: Bootstrap of Elo Estimates (1000 Rounds of Random Sampling)"
-            )
-            plot_3 = gr.Plot(p3, show_label=False)
-        with gr.Column():
-            gr.Markdown(
-                "#### Figure 4: Average Win Rate Against All Other Models (Assuming Uniform Sampling and No Ties)"
-            )
-            plot_4 = gr.Plot(p4, show_label=False)
-    """
-
     from .utils import acknowledgment_md
 
     gr.Markdown(acknowledgment_md)