Update Dockerfile

Dockerfile

@@ -30,36 +30,49 @@ ENV OMP_NUM_THREADS=8 MKL_NUM_THREADS=8 MAX_JOBS=160
 ENV PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:512,garbage_collection_threshold:0.8
 ENV CUDA_LAUNCH_BLOCKING=0 CUDA_CACHE_MAXSIZE=2147483648 CUDA_CACHE_DISABLE=0
 
-# Hugging Face caches
+# App paths
 ENV APP_HOME=/app
 WORKDIR $APP_HOME
 
+# Persistent data and caches in /data
+ENV HF_HOME=/data/.cache/huggingface
+ENV TORCH_HOME=/data/.cache/torch
+ENV HF_DATASETS_CACHE=/data/.cache/datasets
+ENV TRANSFORMERS_CACHE=/data/.cache/transformers
+ENV DIFFUSERS_CACHE=/data/.cache/diffusers
+ENV HF_HUB_ENABLE_HF_TRANSFER=1
+ENV TOKENIZERS_PARALLELISM=false
 
-ENV MODELS_DIR=/app/models
-RUN mkdir -p /data /data/.cache/huggingface /home/user/.cache/models && ln -sf /home/user/.cache/models /app/models
-
-
+# Create non-root user and data dirs early, fix ownership
+RUN useradd -m -u 1000 -s /bin/bash appuser && \
+    mkdir -p /data /data/models \
+             /data/.cache/huggingface /data/.cache/torch \
+             /data/.cache/datasets /data/.cache/transformers /data/.cache/diffusers && \
+    chown -R appuser:appuser /data
 
+# Models live in /data/models and are visible at /app/models
+ENV MODELS_DIR=/data/models
+RUN ln -sf /data/models /app/models
 
-# ---------------- Sistema & Python ----------------
+# ---------------- System & Python ----------------
 RUN apt-get update && apt-get install -y --no-install-recommends \
     build-essential gosu tree cmake git git-lfs curl wget ffmpeg ninja-build \
-    python3.10 python3.10-dev python3.10-distutils python3-pip  \
-    && apt-get clean && rm -rf /var/lib/apt/lists/*
+    python3.10 python3.10-dev python3.10-distutils python3-pip \
+    ca-certificates libglib2.0-0 libgl1 \
+ && apt-get clean && rm -rf /var/lib/apt/lists/*
 
 RUN ln -sf /usr/bin/python3.10 /usr/bin/python3 && \
-    ln -sf /usr/bin/python3.10 /usr/bin/python  && \
+    ln -sf /usr/bin/python3.10 /usr/bin/python && \
     python3 -m pip install --upgrade pip
 
-
-# ---------------- PyTorch cu128 (pinado) ----------------
+# ---------------- PyTorch cu128 (pinned) ----------------
 RUN pip install --index-url https://download.pytorch.org/whl/cu128 \
     torch==2.8.0+cu128 torchvision==0.23.0+cu128 torchaudio==2.8.0+cu128
 
-# ---------------- Toolchain, Triton, FA2 (sem bnb) ----------------
-RUN pip install packaging ninja cmake pybind11 scikit-build cython hf_transfer numpy>=1.24.4
+# ---------------- Toolchain, Triton, FA2 (no bnb build) ----------------
+RUN pip install packaging ninja cmake pybind11 scikit-build cython hf_transfer "numpy>=1.24.4"
 
-# Triton 3.x (sem triton.ops)
+# Triton 3.x (no triton.ops)
 RUN pip uninstall -y triton || true && \
     pip install -v --no-build-isolation triton==3.4.0
 
@@ -69,72 +82,44 @@ RUN pip install flash-attn==2.8.3 --no-build-isolation || \
     pip install flash-attn==2.8.1 --no-build-isolation || \
     pip install flash-attn==2.8.0.post2 --no-build-isolation
 
-# Diffusers/Transformers estáveis (sem dev)
-#RUN pip install --no-cache-dir diffusers>=0.31.0 transformers>=4.44.2 accelerate>=0.34.2 omegaconf>=2.3.0
-
-# Opcional: seu fork de otimizações
-#RUN pip install -U git+https://github.com/carlex22/diffusers-aduc-sdr
-
-# ---------------- Dependências da aplicação ----------------
+# ---------------- App dependencies ----------------
 COPY requirements.txt ./requirements.txt
 RUN pip install --no-cache-dir -r requirements.txt
 
-RUN pip install --upgrade bitsandbytes
-
+# Pin bnb to avoid surprise CUDA/PTX mismatches (adjust as needed)
+RUN pip install --upgrade "bitsandbytes==0.44.1" || pip install --upgrade "bitsandbytes==0.43.3"
 
-RUN apt-get update && apt-get install -y curl && \
-    echo "Instalando dependências .whl customizadas..." && \
+# Custom .whl (Apex + dropout_layer_norm)
+RUN echo "Installing custom wheels..." && \
     pip install --no-cache-dir \
-        "https://huggingface.co/euIaxs22/Aduc-sdr/resolve/main/apex-0.1-cp310-cp310-linux_x86_64.whl" \
-        "https://huggingface.co/euIaxs22/Aduc-sdr/resolve/main/dropout_layer_norm-0.1-cp310-cp310-linux_x86_64.whl" && \
-    apt-get purge -y curl && apt-get autoremove -y && rm -rf /var/lib/apt/lists/*
-
+      "https://huggingface.co/euIaxs22/Aduc-sdr/resolve/main/apex-0.1-cp310-cp310-linux_x86_64.whl" \
+      "https://huggingface.co/euIaxs22/Aduc-sdr/resolve/main/dropout_layer_norm-0.1-cp310-cp310-linux_x86_64.whl"
 
 # ====================================================================
-# <<< BLOCO ATUALIZADO PARA INSTALAR LTX-VIDEO E Q8_KERNELS (com .whl) >>>
-#RUN echo "Instalando q8_kernels (wheel) e LTX-Video (git)..." && \
-    # 1. Instala q8_kernels a partir do seu arquivo .whl pré-compilado
-    #pip install --no-cache-dir "https://huggingface.co/euIaxs22/Aduc-sdr/resolve/main/q8_kernels-0.0.5-cp310-cp310-linux_x86_64.whl" && \ 
-    
-    # 2. Instala a biblioteca LTX-Video a partir do fork (esta parte permanece igual)
-    #git clone https://github.com/Lightricks/LTX-Video.git /data/LTX-Video && \
-    #cd /data/LTX-Video && \
-    #python -m venv env && \
-    #source env/bin/activate  && \
-    #python -m pip install -e .\[inference\] && \
-    #cd ..
+# Optional: q8_kernels + LTX-Video (enable if needed; ensure wheel ABI)
+RUN pip install --no-cache-dir \
+   "https://huggingface.co/euIaxs22/Aduc-sdr/resolve/main/q8_kernels-0.0.5-cp310-cp310-linux_x86_64.whl"
+# RUN git clone https://github.com/Lightricks/LTX-Video.git /data/LTX-Video && \
+#     cd /data/LTX-Video && python -m pip install -e .[inference]
 # ====================================================================
 
-# Scripts e configs
+# Scripts and app
 COPY info.sh ./app/info.sh
 COPY builder.sh ./app/builder.sh
 COPY start.sh ./app/start.sh
 COPY entrypoint.sh ./app/entrypoint.sh
 
+# Copy the rest of the source last for better caching
 COPY . .
-RUN useradd -m -u 1000 -s /bin/bash appuser
-RUN chown -R appuser:appuser /app  && \
-    chmod 0755 /app/entrypoint.sh /app/start.sh /app/info.sh /app/builder.sh
-
 
+# Permissions on app tree
+RUN chown -R appuser:appuser /app /data && \
+    chmod 0755 /app/entrypoint.sh /app/start.sh /app/info.sh /app/builder.sh
 
-# Declara volume persistente para HF Spaces
 VOLUME /data
 
-# Env vars para caches em /data
-ENV HF_HOME=/data/.cache/huggingface
-ENV TORCH_HOME=/data/.cache/torch
-ENV HF_DATASETS_CACHE=/data/.cache/datasets
-ENV TRANSFORMERS_CACHE=/data/.cache/transformers
-ENV DIFFUSERS_CACHE=/data/.cache/diffusers
-ENV HF_HUB_ENABLE_HF_TRANSFER=1
-ENV TOKENIZERS_PARALLELISM=false
-
-
 ENTRYPOINT ["/app/entrypoint.sh"]
-
 USER appuser
 
 # ---------------- Entry ----------------
-
-CMD ["/app/start.sh"]
+CMD ["/app/start.sh"]