add requirements.txt

app.py

@@ -321,11 +321,11 @@ with gr.Blocks(title='VideoLLaMA 2 🔥🚀🔥', theme=theme, css=block_css) as
                 examples=[
                     [
                         f"{cur_dir}/examples/00000368.mp4",
-                        "Where is the loudest instrument?",
+                        "Please describe the video with audio information:",
                     ],
                     [
                         f"{cur_dir}/examples/00003491.mp4",
-                        "Is the instrument on the left louder than the instrument on the right?",
+                        "Where is the loudest instrument?",
                     ],
                 ],
                 inputs=[video, textbox],

requirements.txt

@@ -0,0 +1,41 @@
+# basic dependencies
+torch==2.2.0
+torchaudio==2.2.0
+torchvision==0.17.0
+transformers==4.42.3
+tokenizers==0.19.1
+deepspeed==0.13.1
+accelerate==0.26.1
+peft==0.4.0
+timm==1.0.3
+numpy==1.24.4
+# data processing
+decord==0.6.0
+imageio==2.34.0
+imageio-ffmpeg==0.4.9
+moviepy==1.0.3
+scenedetect==0.6.3
+opencv-python==4.6.0.66
+pysubs2
+librosa
+pytorchvideo
+# misc
+scikit-learn==1.2.2
+huggingface_hub==0.23.4
+sentencepiece==0.1.99
+shortuuid
+einops==0.6.1
+einops-exts==0.0.4
+bitsandbytes==0.43.0
+pydantic>=2.0
+markdown2[all]
+gradio==3.50.0
+gradio_client==0.6.1
+httpx==0.24.1
+openai==1.33.0
+requests
+uvicorn
+fastapi
+tensorboard
+wandb
+tabulate