Skip to content

[circle-mlir/infra] Revise u2204/Dockerfile #15166

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
seanshpark merged 1 commit into from
Apr 16, 2025
Merged

[circle-mlir/infra] Revise u2204/Dockerfile #15166

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
69 changes: 13 additions & 56 deletions circle-mlir/infra/docker/u2204/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,14 +11,6 @@ RUN apt-get update \
&& apt-get -qqy install build-essential cmake git fakeroot autoconf automake libtool unzip wget \
devscripts debmake debhelper lcov

# Install clang-format
RUN apt-get update \
&& apt-get install -qqy gnupg2 \
&& wget -O - https://apt.llvm.org/llvm-snapshot.gpg.key | apt-key add - \
&& add-apt-repository "deb http://apt.llvm.org/jammy/ llvm-toolchain-jammy-16 main" \
&& apt-get update \
&& apt-get install -qqy clang-format-16

# additonal tools
RUN apt-get update \
&& apt-get -qqy install tree tmux
Expand All @@ -27,16 +19,14 @@ RUN apt-get update \
RUN apt-get update \
&& apt-get -qqy install python3 python3-pip python3-venv python3-dev python3-all dh-python
RUN python3 -m pip install --upgrade pip setuptools
RUN python3 -m pip install yapf==0.43.0 numpy==1.24.4 h5py==3.8.0 einops
RUN python3 -m pip install yapf==0.43.0 numpy==2.2.4 h5py==3.8.0 einops

# TODO upgrade
ARG VER_TORCH=2.1.0
ARG VER_ONNX=1.16.0
ARG VER_ONNXRUNTIME=1.18.0
ARG VER_TORCH=2.6.0
ARG VER_ONNX=1.17.0
ARG VER_ONNXRUNTIME=1.21.0

RUN python3 -m pip install torch==${VER_TORCH}
RUN python3 -m pip install onnx==${VER_ONNX}
RUN python3 -m pip install onnxruntime==${VER_ONNXRUNTIME}
RUN python3 -m pip install torch==${VER_TORCH} onnx==${VER_ONNX} onnxruntime==${VER_ONNXRUNTIME}

# Clean archives (to reduce image size)
RUN apt-get clean -y
Expand All @@ -48,47 +38,15 @@ ARG NPROC=4
ARG WORK_DIR=/workdir
ARG FLATBUFFERS_HASH=a078130c
ARG ABSEIL_CPP_HASH=21510581
ARG LLVM_PROJECT_HASH=91088978d712cd7b33610c59f69d87d5a39e3113
ARG ONNX_MLIR_HASH=40615b362fe474d6cce4fd9c31a7ec51acd8f88a
ARG LLVM_PROJECT_HASH=b270525f730be6e7196667925f5a9bfa153262e9
ARG ONNX_MLIR_HASH=660bd8efa664bd2d05801a7a793fe29925313b4b

WORKDIR ${WORK_DIR}

# Generate experimental patch file for onnx-mlir

RUN set -x \
&& { \
echo 'diff --git a/src/Dialect/ONNX/ONNXOps.td.inc b/src/Dialect/ONNX/ONNXOps.td.inc'; \
echo 'index d75ee6b6..f63a2910 100644'; \
echo '--- a/src/Dialect/ONNX/ONNXOps.td.inc'; \
echo '+++ b/src/Dialect/ONNX/ONNXOps.td.inc'; \
echo '@@ -1779,9 +1779,9 @@ def ONNXDequantizeLinearOp:ONNX_Op<"DequantizeLinear",'; \
echo ' `zero-point` is usually not used in the case of float8e4m3fn, float8e4m3fnuz, float8e5m2, float8e5m2fnuz quantization,'; \
echo " but the dequantization formula remains the same for consistency and 'x_scale' still determines the output type."; \
echo ' }];'; \
echo '- let arguments = (ins AnyTypeOf<[TensorOf<[I8]>, TensorOf<[UI8]>, TensorOf<[I32]>, TensorOf<[F8E4M3FN]>, TensorOf<[F8E4M3FNUZ]>, TensorOf<[F8E5M2]>, TensorOf<[F8E5M2FNUZ]>]>:$x,'; \
echo '+ let arguments = (ins AnyTypeOf<[TensorOf<[I8]>, TensorOf<[UI8]>, TensorOf<[I16]>, TensorOf<[I32]>, TensorOf<[F8E4M3FN]>, TensorOf<[F8E4M3FNUZ]>, TensorOf<[F8E5M2]>, TensorOf<[F8E5M2FNUZ]>]>:$x,'; \
echo ' AnyTypeOf<[TensorOf<[F32]>, TensorOf<[F16]>, TensorOf<[BF16]>]>:$x_scale,'; \
echo '- AnyTypeOf<[TensorOf<[I8]>, TensorOf<[UI8]>, TensorOf<[I32]>, TensorOf<[F8E4M3FN]>, TensorOf<[F8E4M3FNUZ]>, TensorOf<[F8E5M2]>, TensorOf<[F8E5M2FNUZ]>, NoneType]>:$x_zero_point,'; \
echo '+ AnyTypeOf<[TensorOf<[I8]>, TensorOf<[UI8]>, TensorOf<[I16]>, TensorOf<[I32]>, TensorOf<[F8E4M3FN]>, TensorOf<[F8E4M3FNUZ]>, TensorOf<[F8E5M2]>, TensorOf<[F8E5M2FNUZ]>, NoneType]>:$x_zero_point,'; \
echo ' DefaultValuedAttr<SI64Attr, "1">:$axis);'; \
echo ' let results = (outs AnyTypeOf<[TensorOf<[F32]>, TensorOf<[F16]>, TensorOf<[BF16]>]>:$y);'; \
echo ' let extraClassDeclaration = [{'; \
echo '@@ -5721,10 +5721,10 @@ def ONNXQuantizeLinearOp:ONNX_Op<"QuantizeLinear",'; \
echo ' }];'; \
echo ' let arguments = (ins AnyTypeOf<[TensorOf<[F32]>, TensorOf<[F16]>, TensorOf<[BF16]>, TensorOf<[I32]>]>:$x,'; \
echo ' AnyTypeOf<[TensorOf<[F32]>, TensorOf<[F16]>, TensorOf<[BF16]>, TensorOf<[I32]>]>:$y_scale,'; \
echo '- AnyTypeOf<[TensorOf<[I8]>, TensorOf<[UI8]>, TensorOf<[F8E4M3FN]>, TensorOf<[F8E4M3FNUZ]>, TensorOf<[F8E5M2]>, TensorOf<[F8E5M2FNUZ]>, NoneType]>:$y_zero_point,'; \
echo '+ AnyTypeOf<[TensorOf<[I8]>, TensorOf<[UI8]>, TensorOf<[I16]>, TensorOf<[F8E4M3FN]>, TensorOf<[F8E4M3FNUZ]>, TensorOf<[F8E5M2]>, TensorOf<[F8E5M2FNUZ]>, NoneType]>:$y_zero_point,'; \
echo ' DefaultValuedAttr<SI64Attr, "1">:$axis,'; \
echo ' DefaultValuedAttr<SI64Attr, "1">:$saturate);'; \
echo '- let results = (outs AnyTypeOf<[TensorOf<[I8]>, TensorOf<[UI8]>, TensorOf<[F8E4M3FN]>, TensorOf<[F8E4M3FNUZ]>, TensorOf<[F8E5M2]>, TensorOf<[F8E5M2FNUZ]>]>:$y);'; \
echo '+ let results = (outs AnyTypeOf<[TensorOf<[I8]>, TensorOf<[UI8]>, TensorOf<[I16]>, TensorOf<[F8E4M3FN]>, TensorOf<[F8E4M3FNUZ]>, TensorOf<[F8E5M2]>, TensorOf<[F8E5M2FNUZ]>]>:$y);'; \
echo ' let extraClassDeclaration = [{'; \
echo ' static int getNumberOfOperands() {'; \
echo ' return 3;'; \
echo ''; \
} > ${WORK_DIR}/onnx_quantdequant_i16.patch \
&& cat ${WORK_DIR}/onnx_quantdequant_i16.patch
# Download experimental patch file for onnx-mlir

RUN wget https://raw.githubusercontent.com/Samsung/ONE/refs/heads/master/circle-mlir/externals/onnx_mlir_0_5_0_0.diff \
-O ${WORK_DIR}/onnx_mlir_0_5_0_0.diff

# FLATBUFFERS build from source
RUN git clone --recursive https://github.com/google/flatbuffers.git \
Expand Down Expand Up @@ -165,9 +123,9 @@ RUN mkdir llvm-project-build && cd llvm-project-build \
RUN cd llvm-project-build && cmake --build . && cmake --install .

# ONNX-MLIR build from source
# Patch to accept INT16 for QuantizeLinear/DequantizeLinear
# Patch to accept INT16 for QuantizeLinear/DequantizeLinear and fix ONNX passes
RUN cd onnx-mlir \
&& patch -p1 < ${WORK_DIR}/onnx_quantdequant_i16.patch
&& patch -p1 < ${WORK_DIR}/onnx_mlir_0_5_0_0.diff

RUN mkdir onnx-mlir-build && cd onnx-mlir-build \
&& cmake -G Ninja -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=${WORK_DIR} \
Expand All @@ -176,7 +134,6 @@ RUN mkdir onnx-mlir-build && cd onnx-mlir-build \
-DCMAKE_JOB_POOL_COMPILE:STRING=compile -DCMAKE_JOB_POOL_LINK:STRING=link \
-DCMAKE_JOB_POOLS:STRING='compile=4;link=1' \
-DONNX_MLIR_BUILD_TESTS=OFF -DONNX_MLIR_ENABLE_MHLO=OFF -DLLVM_ENABLE_ASSERTIONS=ON \
-DONNX_MLIR_DECOMP_ONNX_CONVTRANSPOSE=OFF \
../onnx-mlir

RUN cd onnx-mlir-build && cmake --build . && cmake --install .
Expand Down