Create Dockerfile

Dockerfile

@@ -0,0 +1,43 @@
+# Use a lightweight Python base
+FROM python:3.10-slim
+
+# 1. Install system dependencies and jemalloc (prevents RAM fragmentation)
+RUN apt-get update && apt-get install -y \
+    build-essential \
+    libjemalloc-dev \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+
+# Set jemalloc as the memory allocator (crucial for CPU-only training)
+ENV LD_PRELOAD="/usr/lib/x86_64-linux-gnu/libjemalloc.so"
+
+# 2. Setup user for HF Spaces (non-root)
+RUN useradd -m -u 1000 user
+USER user
+ENV HOME=/home/user \
+    PATH=/home/user/.local/bin:$PATH
+WORKDIR $HOME/app
+
+# 3. Install Python dependencies 
+# IPEX is key for CPU speedup on Intel Xeon (common in HF Spaces)
+RUN pip install --no-cache-dir --upgrade pip && \
+    pip install --no-cache-dir \
+    torch \
+    intel-extension-for-pytorch \
+    transformers \
+    datasets \
+    accelerate \
+    trl \
+    sentencepiece
+
+# 4. Copy your training script and local files
+COPY --chown=user . $HOME/app
+
+# 5. Set Environment Variables for CPU Threading
+# Matches the standard 2vCPU or 4vCPU Space tiers
+ENV OMP_NUM_THREADS=2 \
+    MKL_NUM_THREADS=2 \
+    USE_CPU=1
+
+# 6. Run the training script
+CMD ["python", "train.py"]