add modal serving file

pyproject.toml

@@ -0,0 +1,52 @@
+[project]
+name = "wan2gp"
+version = "0.1.0"
+description = "Add your description here"
+requires-python = ">=3.11"
+dependencies = [
+    "accelerate>=1.1.1",
+    "av>=15.0.0",
+    "dashscope>=1.24.0",
+    "decord>=0.6.0",
+    "diffusers>=0.31.0",
+    "easydict>=1.13",
+    "einops>=0.8.1",
+    "ffmpeg-python>=0.2.0",
+    "ftfy>=6.3.1",
+    "gradio==5.23.0",
+    "hydra-core>=1.3.2",
+    "imageio>=2.37.0",
+    "imageio-ffmpeg>=0.6.0",
+    "librosa==0.11.0",
+    "loguru>=0.7.3",
+    "matplotlib>=3.10.3",
+    "misaki>=0.9.4",
+    "mmgp==3.5.1",
+    "modal>=1.1.0",
+    "moviepy==1.0.3",
+    "mutagen>=1.47.0",
+    "numpy>=1.23.5,<2",
+    "omegaconf>=2.3.0",
+    "onnxruntime-gpu>=1.22.0",
+    "open-clip-torch>=2.29.0",
+    "opencv-python>=4.9.0.80",
+    "peft==0.15.0",
+    "pyannote-audio>=3.3.2",
+    "pydantic==2.10.6",
+    "pygame>=2.1.0",
+    "pyloudnorm>=0.1.1",
+    "rembg[gpu]==2.0.65",
+    "segment-anything>=1.0",
+    "sentencepiece>=0.2.0",
+    "sounddevice>=0.4.0",
+    "soundfile>=0.13.1",
+    "tensordict>=0.6.1",
+    "timm>=1.0.19",
+    "tokenizers>=0.20.3",
+    "torch>=2.4.0",
+    "torchdiffeq>=0.2.5",
+    "torchvision>=0.19.0",
+    "tqdm>=4.67.1",
+    "transformers==4.51.3",
+]
+

serve.py

@@ -0,0 +1,67 @@
+import modal
+
+image = (
+    modal.Image.debian_slim()
+    .apt_install("libgl1", "libglib2.0-0", "libsm6", "libxrender1", "libxext6", "ffmpeg")
+    .pip_install("torch", "torchvision", "torchaudio", index_url="https://download.pytorch.org/whl/cu128")
+    .pip_install_from_pyproject("./pyproject.toml")
+    .pip_install("sageattention")
+    .add_local_dir(local_path="../Wan2GP", remote_path="/root/Wan2GP", ignore=["*.venv"])
+)
+
+app = modal.App(name="Wan2GP", image=image)
+
+vol = modal.Volume.from_name("wan2gp-ckpts", create_if_missing=True, environment_name="dev")
+outputs = modal.Volume.from_name("wan2gp-outputs", create_if_missing=True, environment_name="dev")
+
+with image.imports():
+    import subprocess
+    import os
+
+
+    # @app.function(max_containers=1, gpu="L40S",
+    #               scaledown_window=900,
+    #               timeout=3600,
+    #               volumes={"/root/Wan2GP/ckpts": vol,
+    #                        "/root/Wan2GP/outputs": outputs})
+    # @modal.concurrent(max_inputs=50)
+    # @modal.web_server(8000, startup_timeout=2400)
+    # def ui():
+    #     os.chdir("/root/Wan2GP")
+    #     subprocess.Popen(
+    #         "python /root/Wan2GP/wgp.py --listen --server-port 8000 --compile --attention sage --profile 1",
+    #         shell=True)
+
+    @app.function(max_containers=1, gpu="L40S",
+                  scaledown_window=900,
+                  timeout=60 * 60 * 18,
+                  volumes={"/root/Wan2GP/ckpts": vol,
+                           "/root/Wan2GP/outputs": outputs})
+    @modal.concurrent(max_inputs=1000)
+    @modal.asgi_app()
+    def serve():
+        import asyncio
+        import atexit
+        from fastapi import FastAPI
+        from gradio import mount_gradio_app
+        os.chdir("/root/Wan2GP")
+        import sys
+        sys.path.append('/root/Wan2GP')
+        sys.argv = ["--listen", "--server-port", "8000", "--compile", "--attention", "sage", "--profile", "1"]
+        from wgp import autosave_queue, create_ui, save_path
+
+        app = FastAPI()
+
+        atexit.register(autosave_queue)
+        # download_ffmpeg()
+        # threading.Thread(target=runner, daemon=True).start()
+        os.environ["GRADIO_ANALYTICS_ENABLED"] = "False"
+        server_port = 8000
+        if os.name == "nt":
+            asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())
+        server_name = "0.0.0.0"
+        demo = create_ui()
+        demo.queue(max_size=5)
+        return mount_gradio_app(app=app, blocks=demo, path="/",
+                                server_name=server_name, server_port=server_port,
+                                allowed_paths=[save_path])