gradio

Dockerfile

@@ -3,8 +3,8 @@ FROM pytorch/pytorch:2.2.2-cuda12.1-cudnn8-runtime
 SHELL ["/bin/bash", "-c"]
 
 # Environment variables for Hugging Face cache
-ENV HF_HOME=/home/user/app/hf_cache
-ENV TRANSFORMERS_CACHE=/home/user/app/hf_cache
+ENV HF_HOME=/app/hf_cache
+ENV TRANSFORMERS_CACHE=/app/hf_cache
 ENV HF_TOKEN=${HF_TOKEN}
 ENV PATH=/opt/conda/bin:$PATH
 # Install system dependencies
@@ -19,16 +19,16 @@ WORKDIR /app
 COPY . /app
 
 # Fix permissions for all subdirectories
-RUN mkdir -p /home/user/app/pretrained /home/user/app/hf_cache /.cache/gdown && \
+RUN mkdir -p /app/pretrained /app/hf_cache /.cache/gdown && \
     chmod -R 777 /app && \
     chmod -R 777 /.cache && \
     chmod -R 777 /root
 
 # Create conda environment and install dependencies
-COPY requirements.txt /home/user/app/requirements.txt
+COPY requirements.txt /app/requirements.txt
 RUN conda create -n epic python=3.10 -y && \
     conda run -n epic pip install --upgrade pip && \
-    conda run -n epic pip install -r /home/user/app/requirements.txt
+    conda run -n epic pip install -r /app/requirements.txt
 
 RUN chmod -R 777 /app /workspace
 

download/download.py

@@ -3,22 +3,22 @@ from huggingface_hub import snapshot_download
 def download_model():
     snapshot_download(
         repo_id="tencent/DepthCrafter",
-        local_dir="/home/user/app/pretrained/DepthCrafter",
+        local_dir="/app/pretrained/DepthCrafter",
         local_dir_use_symlinks=False,
     )
     snapshot_download(
         repo_id="stabilityai/stable-video-diffusion-img2vid",
-        local_dir="/home/user/app/pretrained/stable-video-diffusion-img2vid",
+        local_dir="/app/pretrained/stable-video-diffusion-img2vid",
         local_dir_use_symlinks=False,
     )
     snapshot_download(
         repo_id= "Qwen/Qwen2.5-VL-7B-Instruct",
-        local_dir="/home/user/app/pretrained/Qwen2.5-VL-7B-Instruct",
+        local_dir="/app/pretrained/Qwen2.5-VL-7B-Instruct",
         local_dir_use_symlinks=False,
     )
     snapshot_download(
         repo_id="THUDM/CogVideoX1.5-5B-SAT",
-        local_dir="/home/user/app/pretrained/CogVideoX-5b-I2V",
+        local_dir="/app/pretrained/CogVideoX-5b-I2V",
         local_dir_use_symlinks=False,
     )
 

download/download_models.sh

@@ -1,3 +1,3 @@
-mkdir -p /home/user/app/pretrained/RAFT
-gdown 1MqDajR89k-xLV0HIrmJ0k-n8ZpG6_suM -O /home/user/app/pretrained/RAFT/raft-things.pth
-python /home/user/app/download/download.py
+mkdir -p /app/pretrained/RAFT
+gdown 1MqDajR89k-xLV0HIrmJ0k-n8ZpG6_suM -O /app/pretrained/RAFT/raft-things.pth
+python /app/download/download.py

gradio_app.py

@@ -8,12 +8,12 @@ import os
 # -----------------------------
 # Setup paths and env
 # -----------------------------
-HF_HOME = "/home/user/app/hf_cache"
+HF_HOME = "/app/hf_cache"
 os.environ["HF_HOME"] = HF_HOME
 os.environ["TRANSFORMERS_CACHE"] = HF_HOME
 os.makedirs(HF_HOME, exist_ok=True)
 
-PRETRAINED_DIR = "/home/user/app/pretrained"
+PRETRAINED_DIR = "/app/pretrained"
 os.makedirs(PRETRAINED_DIR, exist_ok=True)
 
 # -----------------------------
@@ -43,8 +43,8 @@ def get_anchor_video(video_path, fps, num_frames, target_pose, mode,
                        seed_input, height, width, aspect_ratio_inputs,
                        init_dx, init_dy, init_dz):
 
-    temp_input_path = "/home/user/app/temp_input.mp4"
-    output_dir = "/home/user/app/output_anchor"
+    temp_input_path = "/app/temp_input.mp4"
+    output_dir = "/app/output_anchor"
     video_output_path = f"{output_dir}/masked_videos/output.mp4"
 
     if video_path:
@@ -59,7 +59,7 @@ def get_anchor_video(video_path, fps, num_frames, target_pose, mode,
     h_s, w_s = sample_size.strip().split(",")
     
     command = [
-        "python", "/home/user/app/inference/v2v_data/inference.py",
+        "python", "/app/inference/v2v_data/inference.py",
         "--video_path", temp_input_path,
         "--stride", "1",
         "--out_dir", output_dir,
@@ -111,13 +111,13 @@ def inference(
     seed, height, width, downscale_coef, vae_channels,
     controlnet_input_channels, controlnet_transformer_num_layers
 ):
-    model_path = "/home/user/app/pretrained/CogVideoX-5b-I2V"
-    ckpt_path = "/home/user/app/out/EPiC_pretrained/checkpoint-500.pt"
-    video_root_dir = "/home/user/app/output_anchor"
-    out_dir = "/home/user/app/output"
+    model_path = "/app/pretrained/CogVideoX-5b-I2V"
+    ckpt_path = "/app/out/EPiC_pretrained/checkpoint-500.pt"
+    video_root_dir = "/app/output_anchor"
+    out_dir = "/app/output"
 
     command = [
-        "python", "/home/user/app/inference/cli_demo_camera_i2v_pcd.py",
+        "python", "/app/inference/cli_demo_camera_i2v_pcd.py",
         "--video_root_dir", video_root_dir,
         "--base_model_path", model_path,
         "--controlnet_model_path", ckpt_path,

inference/v2v_data/get_anchor_videos.sh

@@ -6,7 +6,7 @@ target_pose_str="0_30_-0.6_0_0"
 traj_name="loop1"
 traj_txt="test/trajs/${traj_name}.txt"
 
-video="/home/user/app/data/test_v2v/videos/amalfi-coast_traj_loop2.mp4"
+video="/app/data/test_v2v/videos/amalfi-coast_traj_loop2.mp4"
 
 processed_data_name=$1
 # filename=$(basename "$video" .mp4)

inference/v2v_data/inference.py

@@ -99,7 +99,7 @@ def get_parser():
     parser.add_argument(
         '--model_name',
         type=str,
-        default='/home/user/app/pretrained/CogVideoX-Fun-V1.1-5b-InP',
+        default='/app/pretrained/CogVideoX-Fun-V1.1-5b-InP',
         help='Path to the model',
     )
     parser.add_argument(
@@ -113,7 +113,7 @@ def get_parser():
     parser.add_argument(
         '--transformer_path',
         type=str,
-        default="/home/user/app/pretrained/TrajectoryCrafter",
+        default="/app/pretrained/TrajectoryCrafter",
         help='Path to the pretrained transformer model',
     )
     parser.add_argument(
@@ -150,14 +150,14 @@ def get_parser():
         default=". The video is of high quality, and the view is very clear. ",
         help='Prompt for video generation',
     )
-    parser.add_argument('--qwen_path', type=str, default="/home/user/app/pretrained/Qwen2.5-VL-7B-Instruct")
+    parser.add_argument('--qwen_path', type=str, default="/app/pretrained/Qwen2.5-VL-7B-Instruct")
 
     ## depth
     # parser.add_argument('--unet_path', type=str, default='checkpoints/DepthCrafter', help='Path to the UNet model')
     parser.add_argument(
         '--unet_path',
         type=str,
-        default="/home/user/app/pretrained/DepthCrafter",
+        default="/app/pretrained/DepthCrafter",
         help='Path to the UNet model',
     )
 
@@ -165,7 +165,7 @@ def get_parser():
     parser.add_argument(
         '--pre_train_path',
         type=str,
-        default="/home/user/app/pretrained/stable-video-diffusion-img2vid",
+        default="/app/pretrained/stable-video-diffusion-img2vid",
         help='Path to the pre-trained model',
     )
     parser.add_argument(