Dockerfile: test with nvcr.io/nvidia/pytorch:24.07-py3 and install missing libraries

Dockerfile

@@ -1,43 +1,74 @@
-# ✅ PyTorch 2.2.2 (cu121), Python 3.10, cuDNN 9.5.1+
-FROM pytorch/pytorch:2.2.2-cuda12.1-cudnn8-runtime
+# ORI --> cuDNN error: CUDNN_STATUS_SUBLIBRARY_VERSION_MISMATCH
+# FROM nvidia/cuda:12.8.1-cudnn-runtime-ubuntu22.04
+# TESTS UNTIL RIGHT PYTHON VERSION
+# FROM nvidia/cuda:11.6.1-devel-ubuntu20.04
+# FROM nvidia/cuda:12.0.0-cudnn8-devel-ubuntu18.04
+# FROM nvidia/cuda:12.4.1-cudnn-runtime-ubuntu22.04
+# FROM nvidia/cuda:12.1.0-cudnn8-devel-ubuntu22.04
+# FROM nvidia/cuda:12.1.0-cudnn8-runtime-ubuntu22.04
+# FROM nvidia/cuda:12.4.1-cudnn-devel-ubuntu22.04
+FROM nvcr.io/nvidia/pytorch:24.07-py3
 
-ENV PYTHONUNBUFFERED=1
 ENV DEBIAN_FRONTEND=noninteractive
+ENV PYTHONUNBUFFERED=1
 
 WORKDIR /app
 
-# Optional: add caching envs (like HF cache) if needed
+ARG EXTRAS
+ARG HF_PRECACHE_DIR
+ARG HF_TKN_FILE
+
 ENV HF_HOME="/app/tmp/cache/huggingface"
-ENV XDG_CACHE_HOME="/app/tmp/cache"
+ENV HF_HUB_CACHE="/app/tmp/cache/huggingface/hub"
+ENV XDG_CACHE_HOME="/app/tmp/cache/huggingface"
 ENV LIBROSA_CACHE_DIR="/app/tmp/librosa_cache"
 ENV NUMBA_CACHE_DIR="/app/tmp/numba_cache"
 ENV MPLCONFIGDIR="/app/tmp/matplotlib"
 ENV HF_HUB_ETAG_TIMEOUT="600"
 ENV HF_HUB_DOWNLOAD_TIMEOUT="600"
+ENV LD_LIBRARY_PATH="/usr/local/cuda:/usr/lib/x86_64-linux-gnu:${LD_LIBRARY_PATH:-}"
 
-# 🧰 System dependencies
+# Install system dependencies
+#RUN apt-get update && \
+#    apt-get install -y ffmpeg git && \
+#    apt-get clean && \
+#    rm -rf /var/lib/apt/lists/*
+
+# 2) Install system dependencies + Python + pip
 RUN apt-get update && \
     apt-get install -y --no-install-recommends \
+        python3 \
+        python3-pip \
         ffmpeg \
         git \
-        libportaudio2 \
-        python3-dev \
-        portaudio19-dev && \
-    apt-get clean && \
+        libportaudio2 && \
     rm -rf /var/lib/apt/lists/*
 
-# 📦 Python packages
-# Prefer installing torch packages already bundled, to avoid conflicts
-RUN pip install --upgrade pip && \
-    pip install \
-        diart \
-        sounddevice
+RUN apt install -y libaio1 && apt install -y libaio-dev
+
+# RUN apt-get install -y --no-install-recommends libportaudio2
+# portaudio19-dev
+
+RUN pip install diart sounddevice
+RUN pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
 
 COPY . .
 
-# 🧼 Set permissions for Hugging Face cache
-RUN mkdir -p $HF_HOME $XDG_CACHE_HOME $LIBROSA_CACHE_DIR $NUMBA_CACHE_DIR /root/.cache /.cache && \
-    chmod -R 777 /app /root/.cache /.cache
+RUN mkdir -p $HF_HOME && \
+    mkdir -p $HF_HUB_CACHE && \
+    mkdir -p $XDG_CACHE_HOME && \
+    mkdir -p $LIBROSA_CACHE_DIR && \
+    mkdir -p $NUMBA_CACHE_DIR && \
+    mkdir -p $MPLCONFIGDIR && \
+    chmod 777 $HF_HOME && \
+    chmod 777 $HF_HUB_CACHE && \
+    chmod 777 $XDG_CACHE_HOME && \
+    chmod 777 $LIBROSA_CACHE_DIR && \
+    chmod 777 $NUMBA_CACHE_DIR && \
+    chmod 777 $MPLCONFIGDIR && \
+    chmod -R 777 /app
+
+RUN mkdir -p /.cache && chmod -R 777 /.cache
 
 # Install WhisperLiveKit directly, allowing for optional dependencies
 #   Note: For gates modedls, need to add your HF toke. See README.md
@@ -83,10 +114,11 @@ RUN if [ -n "$HF_TKN_FILE" ]; then \
     else \
       echo "No Hugging Face token file specified, skipping token setup"; \
     fi
-
-# ✅ Expose server port
+    
+# Expose port for the transcription server
 EXPOSE 8000
 
-# 🚀 Start the app
 ENTRYPOINT ["whisperlivekit-server", "--host", "0.0.0.0"]
-CMD ["--model", "tiny", "--diarization"]
+
+# Default args
+CMD ["--model", "tiny", "--diarization"]