feat: python package build with abi (cross py3.10+) (#3571)

Signed-off-by: michaelfeil <[email protected]>

Author: michaelfeil

container/Dockerfile

@@ -15,7 +15,6 @@ ARG BASE_IMAGE="nvcr.io/nvidia/cuda-dl-base"
 ARG BASE_IMAGE_TAG="25.01-cuda12.8-devel-ubuntu24.04"
 
 # Build configuration
-ARG RELEASE_BUILD=false
 ARG ENABLE_KVBM=false
 ARG CARGO_BUILD_JOBS
 
@@ -280,7 +279,6 @@ FROM quay.io/pypa/manylinux_2_28_${ARCH_ALT} AS wheel_builder
 ARG ARCH
 ARG ARCH_ALT
 ARG CARGO_BUILD_JOBS
-ARG RELEASE_BUILD
 ARG ENABLE_KVBM
 ARG USE_SCCACHE
 ARG SCCACHE_BUCKET
@@ -356,10 +354,6 @@ RUN --mount=type=secret,id=aws-key-id,env=AWS_ACCESS_KEY_ID \
     else \
         maturin build --release --out /opt/dynamo/dist; \
     fi && \
-    if [ "$RELEASE_BUILD" = "true" ]; then \
-        uv run --python 3.11 maturin build --release --out /opt/dynamo/dist && \
-        uv run --python 3.10 maturin build --release --out /opt/dynamo/dist; \
-    fi && \
     /tmp/use-sccache.sh show-stats "Dynamo"
 
 ##############################################
@@ -382,7 +376,7 @@ COPY --from=wheel_builder $CARGO_HOME $CARGO_HOME
 # Install Python packages
 COPY benchmarks/ /opt/dynamo/benchmarks/
 RUN uv pip install \
-    /opt/dynamo/wheelhouse/ai_dynamo_runtime*cp312*.whl \
+    /opt/dynamo/wheelhouse/ai_dynamo_runtime*.whl \
     /opt/dynamo/wheelhouse/ai_dynamo*any.whl \
     /opt/dynamo/wheelhouse/nixl/nixl*.whl \
     && cd /opt/dynamo/benchmarks \

container/Dockerfile.sglang

@@ -9,7 +9,6 @@ ARG BASE_IMAGE="nvcr.io/nvidia/cuda-dl-base"
 # for details and reproducer to manually test if the image
 # can be updated to later versions.
 ARG BASE_IMAGE_TAG="25.01-cuda12.8-devel-ubuntu24.04"
-ARG RELEASE_BUILD
 ARG RUNTIME_IMAGE="nvcr.io/nvidia/cuda"
 ARG RUNTIME_IMAGE_TAG="12.8.1-runtime-ubuntu24.04"
 
@@ -204,7 +203,7 @@ COPY --from=framework ${VIRTUAL_ENV} ${VIRTUAL_ENV}
 COPY benchmarks/ /opt/dynamo/benchmarks/
 COPY --from=dynamo_base /opt/dynamo/wheelhouse/ /opt/dynamo/wheelhouse/
 RUN uv pip install \
-    /opt/dynamo/wheelhouse/ai_dynamo_runtime*cp312*.whl \
+    /opt/dynamo/wheelhouse/ai_dynamo_runtime*.whl \
     /opt/dynamo/wheelhouse/ai_dynamo*any.whl \
     /opt/dynamo/wheelhouse/nixl/nixl*.whl \
     && cd /opt/dynamo/benchmarks \

container/Dockerfile.trtllm

@@ -3,7 +3,6 @@
 
 ARG BASE_IMAGE="nvcr.io/nvidia/pytorch"
 ARG BASE_IMAGE_TAG="25.06-py3"
-ARG RELEASE_BUILD
 ARG ENABLE_KVBM=false
 ARG RUNTIME_IMAGE="nvcr.io/nvidia/cuda"
 ARG RUNTIME_IMAGE_TAG="12.9.1-runtime-ubuntu24.04"
@@ -242,7 +241,7 @@ ENV LD_LIBRARY_PATH=${TENSORRT_LIB_DIR}:${LD_LIBRARY_PATH}
 COPY benchmarks/ /opt/dynamo/benchmarks/
 COPY --from=dynamo_base /opt/dynamo/wheelhouse/ /opt/dynamo/wheelhouse/
 RUN uv pip install \
-    /opt/dynamo/wheelhouse/ai_dynamo_runtime*cp312*.whl \
+    /opt/dynamo/wheelhouse/ai_dynamo_runtime*.whl \
     /opt/dynamo/wheelhouse/ai_dynamo*any.whl \
     /opt/dynamo/wheelhouse/nixl/nixl*.whl \
     && cd /opt/dynamo/benchmarks \

container/Dockerfile.vllm

@@ -8,7 +8,6 @@ ARG BASE_IMAGE="nvcr.io/nvidia/cuda-dl-base"
 # for details and reproducer to manually test if the image
 # can be updated to later versions.
 ARG BASE_IMAGE_TAG="25.01-cuda12.8-devel-ubuntu24.04"
-ARG RELEASE_BUILD
 ARG ENABLE_KVBM=false
 ARG RUNTIME_IMAGE="nvcr.io/nvidia/cuda"
 ARG RUNTIME_IMAGE_TAG="12.8.1-runtime-ubuntu24.04"
@@ -254,7 +253,7 @@ COPY --from=framework ${VIRTUAL_ENV} ${VIRTUAL_ENV}
 COPY benchmarks/ /opt/dynamo/benchmarks/
 COPY --from=dynamo_base /opt/dynamo/wheelhouse/ /opt/dynamo/wheelhouse/
 RUN uv pip install \
-    /opt/dynamo/wheelhouse/ai_dynamo_runtime*cp312*.whl \
+    /opt/dynamo/wheelhouse/ai_dynamo_runtime*.whl \
     /opt/dynamo/wheelhouse/ai_dynamo*any.whl \
     /opt/dynamo/wheelhouse/nixl/nixl*.whl \
     && cd /opt/dynamo/benchmarks \

container/build.sh

@@ -303,9 +303,6 @@ get_options() {
                 missing_requirement "$1"
             fi
             ;;
-        --release-build)
-            RELEASE_BUILD=true
-            ;;
         --enable-kvbm)
             ENABLE_KVBM=true
             ;;
@@ -705,10 +702,6 @@ fi
 if [ -n "${HF_TOKEN}" ]; then
     BUILD_ARGS+=" --build-arg HF_TOKEN=${HF_TOKEN} "
 fi
-if [  ! -z ${RELEASE_BUILD} ]; then
-    echo "Performing a release build!"
-    BUILD_ARGS+=" --build-arg RELEASE_BUILD=${RELEASE_BUILD} "
-fi
 
 if [[ $FRAMEWORK == "VLLM" ]] || [[ $FRAMEWORK == "TRTLLM" ]]; then
     echo "Forcing enable_kvbm to true in ${FRAMEWORK} image build"

lib/bindings/python/Cargo.toml

@@ -60,7 +60,7 @@ pyo3 = { version = "0.23.4", default-features = false, features = [
   "experimental-inspect",
   "extension-module",
   "py-clone",
-  # "abi3-py310" # TODO: Add abi feature in follow-up, since docker build can be simplified.
+  "abi3-py310"
 ] }
 
 pyo3-async-runtimes = { version = "0.23.0", default-features = false, features = [