Add Initial Support for Running HunyuanVideo on Google Colab (Help Needed with OOM Errors)

README.md

@@ -283,6 +283,36 @@ python3 gradio_server.py --flow-reverse
 # SERVER_NAME=0.0.0.0 SERVER_PORT=8081 python3 gradio_server.py --flow-reverse
 ```
 
+
+### Running on Google Colab
+To run the Gradio Server on Google Colab, follow the steps described in the Download Pretrained Models section first, and then use the following commands:
+
+```bash
+# clone
+!git clone https://github.com/tencent/HunyuanVideo.git
+%cd HunyuanVideo
+
+# setup
+!python -m pip install "huggingface_hub[cli]"
+!huggingface-cli download tencent/HunyuanVideo --local-dir ./ckpts
+!huggingface-cli download xtuner/llava-llama-3-8b-v1_1-transformers --local-dir ./ckpts/llava-llama-3-8b-v1_1-transformers
+!python hyvideo/utils/preprocess_text_encoder_tokenizer_utils.py --input_dir ckpts/llava-llama-3-8b-v1_1-transformers --output_dir ckpts/text_encoder
+!huggingface-cli download openai/clip-vit-large-patch14 --local-dir ./ckpts/text_encoder_2
+!pip install -r requirements.txt
+!python -m pip install ninja
+# Download and install pre-built flash-attention binaries
+# For details, see: https://github.com/kun432/flash-attention-prebuild-wheels
+!wget https://github.com/kun432/flash-attention-prebuild-wheels/releases/download/v0.0.0-test/flash_attn-2.6.3+cu121torch2.5-cp310-cp310-linux_x86_64.whl
+!pip install --no-dependencies --upgrade flash_attn-2.6.3+cu121torch2.5-cp310-cp310-linux_x86_64.whl
+!pip install gradio
+!pip install loguru
+
+# start
+!python3 gradio_server.py --flow-reverse --share
+```
+
+This will provide you with a shareable link to the Gradio interface for generating videos from text prompts in the Colab environment.
+
 ### More Configurations
 
 We list some more useful configurations for easy usage:

gradio_server.py

@@ -5,6 +5,7 @@
 from datetime import datetime
 import gradio as gr
 import random
+import sys
 
 from hyvideo.utils.file_utils import save_videos_grid
 from hyvideo.config import parse_args
@@ -135,7 +136,16 @@ def create_demo(model_path, save_path):
     os.environ["GRADIO_ANALYTICS_ENABLED"] = "False"
     server_name = os.getenv("SERVER_NAME", "0.0.0.0")
     server_port = int(os.getenv("SERVER_PORT", "8081"))
+
+    # Check if --share is specified in command line arguments without changing hyvideo/config.py
+    if "--share" in sys.argv:
+        sys.argv.remove("--share")
+        share_flag = True
+    else:
+        share_flag = False
+
     args = parse_args()
     print(args)
+
     demo = create_demo(args.model_base, args.save_path)
-    demo.launch(server_name=server_name, server_port=server_port)
+    demo.launch(server_name=server_name, server_port=server_port, share=share_flag)