v3 vulkan icd

Dockerfile

@@ -31,6 +31,9 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 ENV PYTHONUNBUFFERED=1 \
     PIP_NO_CACHE_DIR=1 \
     NVIDIA_DRIVER_CAPABILITIES=compute,utility,graphics \
+    HOME=/home/user \
+    PATH=/home/user/.local/bin:$PATH \
+    OMP_NUM_THREADS=1 \
     PORT=7860
 
 RUN useradd -m -u 1000 user
@@ -42,9 +45,10 @@ RUN python3 -m pip install --upgrade pip setuptools wheel \
 
 COPY --chown=user:user . .
 RUN python3 -m pip install -e .
-RUN chown -R user:user /home/user/app
+RUN chmod +x /home/user/app/docker-entrypoint.sh
 
 USER user
 EXPOSE 7860
+ENTRYPOINT ["./docker-entrypoint.sh"]
 
 CMD ["python3", "gradio-web/main.py"]

docker-entrypoint.sh

@@ -0,0 +1,49 @@
+#!/bin/sh
+set -eu
+
+pick_vulkan_icd() {
+    for candidate in \
+        /etc/vulkan/icd.d/nvidia_icd.json \
+        /etc/vulkan/icd.d/nvidia_icd.x86_64.json \
+        /usr/share/vulkan/icd.d/nvidia_icd.json \
+        /usr/share/vulkan/icd.d/nvidia_icd.x86_64.json
+    do
+        if [ -f "$candidate" ]; then
+            printf '%s\n' "$candidate"
+            return 0
+        fi
+    done
+    return 1
+}
+
+run_diagnostic() {
+    label="$1"
+    shift
+    echo "[entrypoint] $label"
+    if "$@"; then
+        return 0
+    fi
+    status=$?
+    echo "[entrypoint] $label failed with exit code $status"
+    return 0
+}
+
+if [ -z "${OMP_NUM_THREADS:-}" ]; then
+    export OMP_NUM_THREADS=1
+fi
+
+if [ -z "${VK_ICD_FILENAMES:-}" ]; then
+    if detected_icd="$(pick_vulkan_icd)"; then
+        export VK_ICD_FILENAMES="$detected_icd"
+        echo "[entrypoint] Using Vulkan ICD: $VK_ICD_FILENAMES"
+    else
+        echo "[entrypoint] Vulkan ICD file not found under /etc or /usr/share"
+    fi
+else
+    echo "[entrypoint] Respecting preset VK_ICD_FILENAMES: $VK_ICD_FILENAMES"
+fi
+
+echo "[entrypoint] OMP_NUM_THREADS=$OMP_NUM_THREADS"
+run_diagnostic "nvidia-smi" nvidia-smi
+run_diagnostic "vulkaninfo --summary" vulkaninfo --summary
+exec "$@"

gradio-web/main.py

@@ -59,6 +59,21 @@ def setup_logging() -> logging.Logger:
 LOGGER = setup_logging()
 
 
+def log_runtime_graphics_env():
+    """Log graphics-related runtime env so Spaces diagnostics are visible in stdout."""
+    keys = [
+        "CUDA_VISIBLE_DEVICES",
+        "NVIDIA_VISIBLE_DEVICES",
+        "NVIDIA_DRIVER_CAPABILITIES",
+        "VK_ICD_FILENAMES",
+        "OMP_NUM_THREADS",
+        "SAPIEN_RENDER_DEVICE",
+        "MUJOCO_GL",
+    ]
+    snapshot = {key: os.getenv(key) for key in keys}
+    LOGGER.info("Runtime graphics env: %s", snapshot)
+
+
 def ensure_media_dirs():
     """Ensure media temp directories exist before first write."""
     TEMP_DEMOS_DIR.mkdir(parents=True, exist_ok=True)
@@ -97,6 +112,7 @@ def main():
     from ui_layout import create_ui_blocks
 
     LOGGER.info("Starting Gradio real environment entrypoint: %s", __file__)
+    log_runtime_graphics_env()
     ensure_media_dirs()
     start_timeout_monitor()