feat: update and modernize finetuner Dockerfile

sd-finetuner/Dockerfile

@@ -1,20 +1,73 @@
-FROM gooseai/torch-base:6cfdc11
+# syntax=docker/dockerfile:1.4
+
+ARG BASE_IMAGE=ghcr.io/coreweave/ml-containers/torch:es-fa3-te-update-1b2ab29-nccl-cuda12.8.1-ubuntu22.04-nccl2.27.5-1-torch2.7.1-vision0.22.1-audio2.7.1-abi1
+# Dependencies requiring NVCC are built ahead of time in a separate stage
+# so that the ~2 GiB dev library installations don't have to be included
+# in the final finetuner image.
+# gcc-10/g++-10/lld do not need to be installed here, but they improve the build.
+# gfortran-10 is just for compiler_wrapper.f95.
+FROM ${BASE_IMAGE} as builder
+
+# Install build dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+        cuda-nvcc-12-8 \
+        cuda-nvml-dev-12-8 \
+        libcurand-dev-12-8 \
+        libcublas-dev-12-8 \
+        libcusparse-dev-12-8 \
+        libcusolver-dev-12-8 \
+        cuda-profiler-api-12-8 \
+        ninja-build \
+        gcc g++ gfortran && \
+    apt-get clean && \
+    rm -rf /var/lib/apt/lists/*
+
+# Create and pre-compile wheels for DeepSpeed
+RUN mkdir /wheels
+WORKDIR /wheels
+COPY requirements-precompilable.txt .
+RUN python3 -m pip install -U --no-cache-dir wheel && \
+    MAX_JOBS=16 DS_BUILD_UTILS=1 DS_BUILD_CPU_ADAM=1 \
+      TORCH_CUDA_ARCH_LIST="8.0 9.0+PTX" \
+      python3 -m pip wheel \
+      --no-cache-dir --no-build-isolation --no-deps \
+      -r requirements-precompilable.txt
+
+# Final image stage
+FROM ${BASE_IMAGE}
+
+# Update package lists and install dependencies
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends \
+    wget \
+    libglib2.0-0
 
-RUN apt-get install -y cuda-nvcc-11-3 cuda-nvml-dev-11-3 libcurand-dev-11-3 \
-                       libcublas-dev-11-3 libcusparse-dev-11-3 \
-                       libcusolver-dev-11-3 cuda-nvprof-11-3 \
-                       ninja-build git && \
-    apt-get clean
 
 RUN mkdir /app
 WORKDIR /app
 
 ARG COMMIT=master
-RUN git clone https://github.com/coreweave/kubernetes-cloud.git && \
-    cd kubernetes-cloud && \
+RUN git clone git@github.com:coreweave/finetuner && \
+    cd finetuner && \
     git checkout ${COMMIT} && \
-    cd ..
-RUN cp kubernetes-cloud/sd-finetuner-workflow/sd-finetuner/* .
+    cd .. && \
+    cp finetuner/* .  && \
+    rm -rf finetuner \
+
+# Install the pre-compiled DeepSpeed wheel
+RUN --mount=type=bind,from=builder,source=/wheels,target=/wheels \
+    pip3 install --no-cache-dir /wheels/*.whl
+
+# Install remaining Python requirements
+COPY requirements.txt .
+COPY requirements-precompilable.txt .
 RUN pip3 install --no-cache-dir -r requirements.txt
 
-CMD [ "/usr/bin/python3", "finetuner.py" ]
+# Copy application code
+COPY ds_config.json .
+COPY finetuner.py .
+COPY evaluator.py .
+COPY inference.py .
+COPY utils.py .
+
+CMD [ "/usr/bin/python3", "finetuner.py" ]