Use the latest modelscope

Dockerfile

@@ -43,6 +43,13 @@ RUN pyenv install ${PYTHON_VERSION} && \
 COPY --chown=1000 ./requirements.txt /tmp/requirements.txt
 RUN pip install --no-cache-dir --upgrade -r /tmp/requirements.txt
 
+RUN git clone https://github.com/modelscope/modelscope && \
+    cd modelscope && \
+    pip install -r requirements.txt && \
+    pip install . && \
+    cd .. && \
+    rm -rf modelscope
+
 COPY --chown=1000 . ${HOME}/app
 ENV PYTHONPATH=${HOME}/app \
     PYTHONUNBUFFERED=1 \

README.md

@@ -1,10 +1,11 @@
 ---
 title: ModelScope-Vid2Vid-XL
-emoji: 📉
+emoji: 🔥
 colorFrom: indigo
 colorTo: purple
 sdk: docker
 pinned: false
+suggested_hardware: a10g-large
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -4,6 +4,7 @@ import os
 import pathlib
 import tempfile
 
+import cv2
 import gradio as gr
 import torch
 from huggingface_hub import snapshot_download
@@ -11,6 +12,8 @@ from modelscope.outputs import OutputKeys
 from modelscope.pipelines import pipeline
 
 DESCRIPTION = "# ModelScope-Vid2Vid-XL"
+if not torch.cuda.is_available():
+    DESCRIPTION += "\n<p>Running on CPU 🥶 This demo does not work on CPU.</p>"
 
 if torch.cuda.is_available():
     model_cache_dir = os.getenv("MODEL_CACHE_DIR", "./models")
@@ -21,7 +24,20 @@ else:
     pipe = None
 
 
+def check_input_video(video_path: str) -> None:
+    cap = cv2.VideoCapture(video_path)
+    n_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+    width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+    height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+    cap.release()
+    if n_frames != 32 or width != 448 or height != 256:
+        raise gr.Error(
+            f"Input video must be 32 frames of size 448x256. Your video is {n_frames} frames of size {width}x{height}."
+        )
+
+
 def video_to_video(video_path: str, text: str) -> str:
+    check_input_video(video_path)
     p_input = {"video_path": video_path, "text": text}
     output_file = tempfile.NamedTemporaryFile(suffix=".mp4", delete=False)
     pipe(p_input, output_video=output_file.name)[OutputKeys.OUTPUT_VIDEO]
@@ -30,14 +46,38 @@ def video_to_video(video_path: str, text: str) -> str:
 
 with gr.Blocks(css="style.css") as demo:
     gr.Markdown(DESCRIPTION)
-    input_video = gr.Video(label="Input video", type="filepath")
-    text_description = gr.Textbox(label="Text description")
-    run_button = gr.Button()
-    output_video = gr.Video(label="Output video")
+    gr.DuplicateButton(
+        value="Duplicate Space for private use",
+        elem_id="duplicate-button",
+        visible=os.getenv("SHOW_DUPLICATE_BUTTON") == "1",
+    )
+    with gr.Group():
+        input_video = gr.Video(label="Input video", type="filepath")
+        text_description = gr.Textbox(label="Text description")
+        run_button = gr.Button()
+        output_video = gr.Video(label="Output video")
+    text_description.submit(
+        fn=check_input_video,
+        inputs=input_video,
+        queue=False,
+        api_name=None,
+    ).success(
+        fn=video_to_video,
+        inputs=[input_video, text_description],
+        outputs=output_video,
+        api_name=None,
+    )
     run_button.click(
+        fn=check_input_video,
+        inputs=input_video,
+        queue=False,
+        api_name=None,
+    ).success(
         fn=video_to_video,
         inputs=[input_video, text_description],
         outputs=output_video,
         api_name="run",
     )
-demo.queue(max_size=20).launch()
+
+if __name__ == "__main__":
+    demo.queue(max_size=10).launch()

requirements.txt

@@ -1,7 +1,7 @@
 easydict==1.10
 einops==0.6.1
 fairscale==0.4.13
-gradio==3.41.1
+gradio==3.41.2
 huggingface_hub==0.16.4
 imageio==2.31.1
 modelscope==1.8.4