diff --git a/.ci/docker/build.sh b/.ci/docker/build.sh index f857147c364..a3387eddc78 100755 --- a/.ci/docker/build.sh +++ b/.ci/docker/build.sh @@ -86,6 +86,10 @@ CMAKE_VERSION=3.18.5 _UCX_COMMIT=7bb2722ff2187a0cad557ae4a6afa090569f83fb _UCC_COMMIT=20eae37090a4ce1b32bcce6144ccad0b49943e0b +if [[ "$image" == *rocm* ]]; then + _UCX_COMMIT=cc312eaa4655c0cc5c2bcd796db938f90563bcf6 + _UCC_COMMIT=0c0fc21559835044ab107199e334f7157d6a0d3d +fi # It's annoying to rename jobs every time you want to rewrite a # configuration, so we hardcode everything here rather than do it @@ -206,7 +210,7 @@ case "$image" in ;; pytorch-linux-focal-rocm-n-1-py3) ANACONDA_PYTHON_VERSION=3.10 - GCC_VERSION=9 + GCC_VERSION=11 PROTOBUF=yes DB=yes VISION=yes @@ -214,10 +218,14 @@ case "$image" in NINJA_VERSION=1.9.0 CONDA_CMAKE=yes TRITON=yes + KATEX=yes + UCX_COMMIT=${_UCX_COMMIT} + UCC_COMMIT=${_UCC_COMMIT} + INDUCTOR_BENCHMARKS=yes ;; pytorch-linux-focal-rocm-n-py3) ANACONDA_PYTHON_VERSION=3.10 - GCC_VERSION=9 + GCC_VERSION=11 PROTOBUF=yes DB=yes VISION=yes @@ -225,6 +233,10 @@ case "$image" in NINJA_VERSION=1.9.0 CONDA_CMAKE=yes TRITON=yes + KATEX=yes + UCX_COMMIT=${_UCX_COMMIT} + UCC_COMMIT=${_UCC_COMMIT} + INDUCTOR_BENCHMARKS=yes ;; pytorch-linux-jammy-xpu-2024.0-py3) ANACONDA_PYTHON_VERSION=3.9 diff --git a/.ci/docker/common/install_ucc.sh b/.ci/docker/common/install_ucc.sh index 2224811bd98..b7f884ea964 100755 --- a/.ci/docker/common/install_ucc.sh +++ b/.ci/docker/common/install_ucc.sh @@ -8,6 +8,12 @@ else with_cuda=no fi +if [[ -d "/opt/rocm" ]]; then + with_rocm=/opt/rocm +else + with_rocm=no +fi + function install_ucx() { set -ex git clone --recursive https://github.com/openucx/ucx.git @@ -19,6 +25,7 @@ function install_ucx() { ./configure --prefix=$UCX_HOME \ --enable-mt \ --with-cuda=$with_cuda \ + --with-rocm=$with_rocm \ --enable-profiling \ --enable-stats time make -j @@ -36,12 +43,29 @@ function install_ucc() { git submodule update --init --recursive ./autogen.sh + # We only run distributed tests on Tesla M60 and A10G NVCC_GENCODE="-gencode=arch=compute_52,code=sm_52 -gencode=arch=compute_86,code=compute_86" + + if [[ -n "$ROCM_VERSION" ]]; then + if [[ -n "$PYTORCH_ROCM_ARCH" ]]; then + amdgpu_targets=`echo $PYTORCH_ROCM_ARCH | sed 's/;/ /g'` + else + amdgpu_targets=`rocm_agent_enumerator | grep -v gfx000 | sort -u | xargs` + fi + for arch in $amdgpu_targets; do + HIP_OFFLOAD="$HIP_OFFLOAD --offload-arch=$arch" + done + else + HIP_OFFLOAD="all-arch-no-native" + fi + ./configure --prefix=$UCC_HOME \ --with-ucx=$UCX_HOME \ --with-cuda=$with_cuda \ - --with-nvcc-gencode="${NVCC_GENCODE}" + --with-nvcc-gencode="${NVCC_GENCODE}" \ + --with-rocm=$with_rocm \ + --with-rocm-arch="${HIP_OFFLOAD}" time make -j sudo make install diff --git a/.ci/docker/ubuntu-rocm/Dockerfile b/.ci/docker/ubuntu-rocm/Dockerfile index faad1d91918..70ea39b5c7b 100644 --- a/.ci/docker/ubuntu-rocm/Dockerfile +++ b/.ci/docker/ubuntu-rocm/Dockerfile @@ -14,21 +14,20 @@ ENV PYTORCH_ROCM_ARCH ${PYTORCH_ROCM_ARCH} COPY ./common/install_base.sh install_base.sh RUN bash ./install_base.sh && rm install_base.sh -# Install clang -ARG LLVMDEV -ARG CLANG_VERSION -COPY ./common/install_clang.sh install_clang.sh -RUN bash ./install_clang.sh && rm install_clang.sh - # Install user COPY ./common/install_user.sh install_user.sh RUN bash ./install_user.sh && rm install_user.sh +# Install katex +ARG KATEX +COPY ./common/install_docs_reqs.sh install_docs_reqs.sh +RUN bash ./install_docs_reqs.sh && rm install_docs_reqs.sh + # Install conda and other packages (e.g., numpy, pytest) ARG ANACONDA_PYTHON_VERSION -ARG CONDA_CMAKE ENV ANACONDA_PYTHON_VERSION=$ANACONDA_PYTHON_VERSION ENV PATH /opt/conda/envs/py_$ANACONDA_PYTHON_VERSION/bin:/opt/conda/bin:$PATH +ARG CONDA_CMAKE COPY requirements-ci.txt /opt/conda/requirements-ci.txt COPY ./common/install_conda.sh install_conda.sh COPY ./common/common_utils.sh common_utils.sh @@ -39,6 +38,11 @@ ARG GCC_VERSION COPY ./common/install_gcc.sh install_gcc.sh RUN bash ./install_gcc.sh && rm install_gcc.sh +# Install clang +ARG CLANG_VERSION +COPY ./common/install_clang.sh install_clang.sh +RUN bash ./install_clang.sh && rm install_clang.sh + # (optional) Install protobuf for ONNX ARG PROTOBUF COPY ./common/install_protobuf.sh install_protobuf.sh @@ -85,6 +89,32 @@ COPY ./common/install_amdsmi.sh install_amdsmi.sh RUN bash ./install_amdsmi.sh RUN rm install_amdsmi.sh +# (optional) Install UCC +ARG UCX_COMMIT +ARG UCC_COMMIT +ENV UCX_COMMIT $UCX_COMMIT +ENV UCC_COMMIT $UCC_COMMIT +ENV UCX_HOME /usr +ENV UCC_HOME /usr +ADD ./common/install_ucc.sh install_ucc.sh +RUN if [ -n "${UCX_COMMIT}" ] && [ -n "${UCC_COMMIT}" ]; then bash ./install_ucc.sh; fi +RUN rm install_ucc.sh + +COPY ./common/install_openssl.sh install_openssl.sh +ENV OPENSSL_ROOT_DIR /opt/openssl +RUN bash ./install_openssl.sh +ENV OPENSSL_DIR /opt/openssl + +ARG INDUCTOR_BENCHMARKS +ARG ANACONDA_PYTHON_VERSION +ENV ANACONDA_PYTHON_VERSION=$ANACONDA_PYTHON_VERSION +COPY ./common/install_inductor_benchmark_deps.sh install_inductor_benchmark_deps.sh +COPY ./common/common_utils.sh common_utils.sh +COPY ci_commit_pins/huggingface.txt huggingface.txt +COPY ci_commit_pins/timm.txt timm.txt +RUN if [ -n "${INDUCTOR_BENCHMARKS}" ]; then bash ./install_inductor_benchmark_deps.sh; fi +RUN rm install_inductor_benchmark_deps.sh common_utils.sh timm.txt huggingface.txt + # (optional) Install non-default CMake version ARG CMAKE_VERSION COPY ./common/install_cmake.sh install_cmake.sh @@ -107,17 +137,17 @@ COPY triton_version.txt triton_version.txt RUN if [ -n "${TRITON}" ]; then bash ./install_triton.sh; fi RUN rm install_triton.sh common_utils.sh triton.txt triton_version.txt -# This is needed by sccache -COPY ./common/install_openssl.sh install_openssl.sh -ENV OPENSSL_ROOT_DIR /opt/openssl -RUN bash ./install_openssl.sh -ENV OPENSSL_DIR /opt/openssl # Install ccache/sccache (do this last, so we get priority in PATH) COPY ./common/install_cache.sh install_cache.sh ENV PATH /opt/cache/bin:$PATH RUN bash ./install_cache.sh && rm install_cache.sh +# Install Open MPI for ROCm +COPY ./common/install_openmpi.sh install_openmpi.sh +RUN if [ -n "${CUDA_VERSION}" ]; then bash install_openmpi.sh; fi +RUN rm install_openmpi.sh + # Include BUILD_ENVIRONMENT environment variable in image ARG BUILD_ENVIRONMENT ENV BUILD_ENVIRONMENT ${BUILD_ENVIRONMENT} diff --git a/.ci/pytorch/common_utils.sh b/.ci/pytorch/common_utils.sh index 1f7093fba29..de2449d47c8 100644 --- a/.ci/pytorch/common_utils.sh +++ b/.ci/pytorch/common_utils.sh @@ -169,13 +169,35 @@ function install_torchrec_and_fbgemm() { torchrec_commit=$(get_pinned_commit torchrec) local fbgemm_commit fbgemm_commit=$(get_pinned_commit fbgemm) + if [[ "$BUILD_ENVIRONMENT" == *rocm* ]] ; then + fbgemm_commit=$(get_pinned_commit fbgemm_rocm) + fi pip_uninstall torchrec-nightly pip_uninstall fbgemm-gpu-nightly pip_install setuptools-git-versioning scikit-build pyre-extensions - # See https://github.com/pytorch/pytorch/issues/106971 - CUDA_PATH=/usr/local/cuda-12.1 pip_install --no-use-pep517 --user "git+https://github.com/pytorch/FBGEMM.git@${fbgemm_commit}#egg=fbgemm-gpu&subdirectory=fbgemm_gpu" - pip_install --no-use-pep517 --user "git+https://github.com/pytorch/torchrec.git@${torchrec_commit}" + if [[ "$BUILD_ENVIRONMENT" == *rocm* ]] ; then + # install torchrec first because it installs fbgemm nightly on top of rocm fbgemm + pip_install --no-use-pep517 --user "git+https://github.com/pytorch/torchrec.git@${torchrec_commit}" + pip_uninstall fbgemm-gpu-nightly + + pip_install tabulate # needed for newer fbgemm + pip_install patchelf # needed for rocm fbgemm + git clone --recursive https://github.com/pytorch/fbgemm + pushd fbgemm/fbgemm_gpu + git checkout "${fbgemm_commit}" + python setup.py install \ + --package_variant=rocm \ + -DHIP_ROOT_DIR="${ROCM_PATH}" \ + -DCMAKE_C_FLAGS="-DTORCH_USE_HIP_DSA" \ + -DCMAKE_CXX_FLAGS="-DTORCH_USE_HIP_DSA" + popd + rm -rf fbgemm + else + # See https://github.com/pytorch/pytorch/issues/106971 + CUDA_PATH=/usr/local/cuda-12.1 pip_install --no-use-pep517 --user "git+https://github.com/pytorch/FBGEMM.git@${fbgemm_commit}#egg=fbgemm-gpu&subdirectory=fbgemm_gpu" + pip_install --no-use-pep517 --user "git+https://github.com/pytorch/torchrec.git@${torchrec_commit}" + fi } function clone_pytorch_xla() { diff --git a/.ci/pytorch/test.sh b/.ci/pytorch/test.sh index 4213c911b1e..f3129922c0c 100755 --- a/.ci/pytorch/test.sh +++ b/.ci/pytorch/test.sh @@ -177,6 +177,9 @@ if [[ "$BUILD_ENVIRONMENT" == *rocm* ]]; then # Print GPU info rocminfo rocminfo | grep -E 'Name:.*\sgfx|Marketing' + + # for benchmarks/dynamo/check_accuracy.py, we need to put results in a rocm specific directory to avoid clashes with cuda + MAYBE_ROCM="rocm/" fi if [[ "$BUILD_ENVIRONMENT" == *xpu* ]]; then @@ -429,7 +432,7 @@ test_inductor_cpp_wrapper_shard() { --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_training.csv" python benchmarks/dynamo/check_accuracy.py \ --actual "$TEST_REPORTS_DIR/inductor_cpp_wrapper_training.csv" \ - --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_timm_training.csv" + --expected "benchmarks/dynamo/ci_expected_accuracy/${MAYBE_ROCM}inductor_timm_training.csv" python benchmarks/dynamo/torchbench.py --device cuda --accuracy \ --bfloat16 --inference --inductor --only hf_T5 --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv" @@ -439,7 +442,7 @@ test_inductor_cpp_wrapper_shard() { --bfloat16 --inference --inductor --only moco --output "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv" python benchmarks/dynamo/check_accuracy.py \ --actual "$TEST_REPORTS_DIR/inductor_cpp_wrapper_inference.csv" \ - --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_torchbench_inference.csv" + --expected "benchmarks/dynamo/ci_expected_accuracy/${MAYBE_ROCM}inductor_torchbench_inference.csv" fi } @@ -638,10 +641,10 @@ test_single_dynamo_benchmark() { --output "$TEST_REPORTS_DIR/${name}_${suite}.csv" python benchmarks/dynamo/check_accuracy.py \ --actual "$TEST_REPORTS_DIR/${name}_$suite.csv" \ - --expected "benchmarks/dynamo/ci_expected_accuracy/${TEST_CONFIG}_${name}.csv" + --expected "benchmarks/dynamo/ci_expected_accuracy/${MAYBE_ROCM}${TEST_CONFIG}_${name}.csv" python benchmarks/dynamo/check_graph_breaks.py \ --actual "$TEST_REPORTS_DIR/${name}_$suite.csv" \ - --expected "benchmarks/dynamo/ci_expected_accuracy/${TEST_CONFIG}_${name}.csv" + --expected "benchmarks/dynamo/ci_expected_accuracy/${MAYBE_ROCM}${TEST_CONFIG}_${name}.csv" fi } @@ -728,7 +731,7 @@ test_inductor_torchbench_smoketest_perf() { --only $test --output "$TEST_REPORTS_DIR/inductor_warm_start_smoketest_$test.csv" python benchmarks/dynamo/check_accuracy.py \ --actual "$TEST_REPORTS_DIR/inductor_warm_start_smoketest_$test.csv" \ - --expected "benchmarks/dynamo/ci_expected_accuracy/inductor_huggingface_training.csv" + --expected "benchmarks/dynamo/ci_expected_accuracy/${MAYBE_ROCM}inductor_huggingface_training.csv" done } diff --git a/.github/ci_commit_pins/fbgemm_rocm.txt b/.github/ci_commit_pins/fbgemm_rocm.txt new file mode 100644 index 00000000000..fa11e10ca6b --- /dev/null +++ b/.github/ci_commit_pins/fbgemm_rocm.txt @@ -0,0 +1 @@ +5fb5024118e9bb9decf96c2b0b1a8f0010bf56be diff --git a/.github/workflows/inductor-perf-test-nightly-rocm.yml b/.github/workflows/inductor-perf-test-nightly-rocm.yml index 1316297d52c..90f7b4f4fbb 100644 --- a/.github/workflows/inductor-perf-test-nightly-rocm.yml +++ b/.github/workflows/inductor-perf-test-nightly-rocm.yml @@ -77,8 +77,7 @@ jobs: curr_ref_type: ${{ github.ref_type }} linux-focal-rocm6_3-py3_10-inductor-benchmark-build: - # TODO (huydhn): Remove this once https://github.com/pytorch/pytorch/pull/144594 lands - if: false + if: github.repository_owner == 'pytorch' name: rocm6_3-py3_10-inductor-benchmark-build uses: ./.github/workflows/_linux-build.yml with: @@ -86,18 +85,18 @@ jobs: docker-image-name: pytorch-linux-focal-rocm-n-py3 test-matrix: | { include: [ - { config: "inductor_huggingface_perf_rocm", shard: 1, num_shards: 3, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_huggingface_perf_rocm", shard: 2, num_shards: 3, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_huggingface_perf_rocm", shard: 3, num_shards: 3, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_timm_perf_rocm", shard: 1, num_shards: 5, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_timm_perf_rocm", shard: 2, num_shards: 5, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_timm_perf_rocm", shard: 3, num_shards: 5, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_timm_perf_rocm", shard: 4, num_shards: 5, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_timm_perf_rocm", shard: 5, num_shards: 5, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_torchbench_perf_rocm", shard: 1, num_shards: 4, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_torchbench_perf_rocm", shard: 2, num_shards: 4, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_torchbench_perf_rocm", shard: 3, num_shards: 4, runner: "linux.rocm.gpu.mi300.2" }, - { config: "inductor_torchbench_perf_rocm", shard: 4, num_shards: 4, runner: "linux.rocm.gpu.mi300.2" }, + { config: "inductor_huggingface_perf_rocm", shard: 1, num_shards: 3, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "inductor_huggingface_perf_rocm", shard: 2, num_shards: 3, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "inductor_huggingface_perf_rocm", shard: 3, num_shards: 3, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "inductor_timm_perf_rocm", shard: 1, num_shards: 5, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "inductor_timm_perf_rocm", shard: 2, num_shards: 5, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "inductor_timm_perf_rocm", shard: 3, num_shards: 5, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "inductor_timm_perf_rocm", shard: 4, num_shards: 5, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "inductor_timm_perf_rocm", shard: 5, num_shards: 5, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "inductor_torchbench_perf_rocm", shard: 1, num_shards: 4, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "inductor_torchbench_perf_rocm", shard: 2, num_shards: 4, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "inductor_torchbench_perf_rocm", shard: 3, num_shards: 4, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "inductor_torchbench_perf_rocm", shard: 4, num_shards: 4, runner: "linux.rocm.gpu.mi300.2.c2" }, ]} secrets: inherit diff --git a/.github/workflows/inductor-periodic.yml b/.github/workflows/inductor-periodic.yml index d27d80f7aec..9528307a4c1 100644 --- a/.github/workflows/inductor-periodic.yml +++ b/.github/workflows/inductor-periodic.yml @@ -67,6 +67,47 @@ jobs: test-matrix: ${{ needs.linux-focal-cuda12_4-py3_10-gcc9-periodic-dynamo-benchmarks-build.outputs.test-matrix }} secrets: inherit + linux-focal-rocm6_3-py3_10-periodic-dynamo-benchmarks-build: + if: github.repository_owner == 'pytorch' + name: rocm6_3-py3_10-periodic-dynamo-benchmarks + uses: ./.github/workflows/_linux-build.yml + with: + build-environment: linux-focal-rocm6_3-py3_10 + docker-image-name: pytorch-linux-focal-rocm-n-py3 + sync-tag: rocm-build + test-matrix: | + { include: [ + { config: "dynamo_eager_torchbench", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "dynamo_eager_torchbench", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "dynamo_eager_huggingface", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "dynamo_eager_timm", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "dynamo_eager_timm", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "aot_eager_torchbench", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "aot_eager_torchbench", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "aot_eager_huggingface", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "aot_eager_timm", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "aot_eager_timm", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "dynamic_aot_eager_torchbench", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "dynamic_aot_eager_torchbench", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "dynamic_aot_eager_huggingface", shard: 1, num_shards: 1, runner: "linux.rocm.gpu.mi300.2.c2" }, + { config: "dynamic_aot_eager_timm", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c1" }, + { config: "dynamic_aot_eager_timm", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.mi300.2.c2" }, + ]} + secrets: inherit + + linux-focal-rocm6_3-py3_10-periodic-dynamo-benchmarks-test: + permissions: + id-token: write + contents: read + name: rocm6_3-py3_10-periodic-dynamo-benchmarks + uses: ./.github/workflows/_rocm-test.yml + needs: linux-focal-rocm6_3-py3_10-periodic-dynamo-benchmarks-build + with: + build-environment: linux-focal-rocm6_3-py3_10 + docker-image: ${{ needs.linux-focal-rocm6_3-py3_10-periodic-dynamo-benchmarks-build.outputs.docker-image }} + test-matrix: ${{ needs.linux-focal-rocm6_3-py3_10-periodic-dynamo-benchmarks-build.outputs.test-matrix }} + secrets: inherit + linux-focal-cuda12_4-py3_10-gcc9-inductor-build-gcp: name: cuda12.4-py3.10-gcc9-sm80 uses: ./.github/workflows/_linux-build.yml diff --git a/benchmarks/dynamo/check_accuracy.py b/benchmarks/dynamo/check_accuracy.py index 359289219dc..92b7f552fc3 100644 --- a/benchmarks/dynamo/check_accuracy.py +++ b/benchmarks/dynamo/check_accuracy.py @@ -27,6 +27,50 @@ def check_accuracy(actual_csv, expected_csv, expected_filename): failed = [] improved = [] + if "rocm" in expected_filename: + flaky_models.update( + { + "alexnet", + "cait_m36_384", + "demucs", + "densenet121", + "detectron2_fcos_r_50_fpn", + "doctr_det_predictor", + "doctr_reco_predictor", + "hf_BigBird", + "hf_Longformer", + "hf_Reformer", + "hf_Roberta_base", + "hf_T5", + "hf_T5_base", + "levit_128", + "llava", + "microbench_unbacked_tolist_sum", + "mnasnet1_0", + "mobilenet_v2", + "pytorch_CycleGAN_and_pix2pix", + "pytorch_stargan", + "resnet152", + "resnet18", + "resnet50", + "resnext50_32x4d", + "sam", + "sam_fast", + "shufflenet_v2_x1_0", + "squeezenet1_1", + "stable_diffusion_text_encoder", + "stable_diffusion_unet", + "timm_efficientdet", + "timm_efficientnet", + "timm_nfnet", + "timm_regnet", + "timm_resnest", + "timm_vovnet", + "torchrec_dlrm", + "vgg16", + } + ) + for model in actual_csv["name"]: accuracy = get_field(actual_csv, model, "accuracy") expected_accuracy = get_field(expected_csv, model, "accuracy") diff --git a/benchmarks/dynamo/check_graph_breaks.py b/benchmarks/dynamo/check_graph_breaks.py index c5279bb6bcf..8a62538c96d 100644 --- a/benchmarks/dynamo/check_graph_breaks.py +++ b/benchmarks/dynamo/check_graph_breaks.py @@ -6,6 +6,16 @@ import textwrap import pandas as pd +# Hack to have something similar to DISABLED_TEST. These models are flaky. + +flaky_models = { + "yolov3", + "gluon_inception_v3", + "detectron2_maskrcnn_r_101_c4", + "XGLMForCausalLM", # discovered in https://github.com/pytorch/pytorch/pull/128148 +} + + def get_field(csv, model_name: str, field: str): try: return csv.loc[csv["name"] == model_name][field].item() @@ -17,23 +27,62 @@ def check_graph_breaks(actual_csv, expected_csv, expected_filename): failed = [] improved = [] + if "rocm" in expected_filename: + flaky_models.update( + { + "alexnet", + "cait_m36_384", + "demucs", + "densenet121", + "detectron2_fcos_r_50_fpn", + "doctr_det_predictor", + "doctr_reco_predictor", + "hf_BigBird", + "hf_Longformer", + "hf_Reformer", + "hf_Roberta_base", + "hf_T5", + "hf_T5_base", + "levit_128", + "llava", + "microbench_unbacked_tolist_sum", + "sam", + "sam_fast", + "stable_diffusion_text_encoder", + "stable_diffusion_unet", + "timm_efficientdet", + "timm_nfnet", + "torchrec_dlrm", + "vgg16", + } + ) + for model in actual_csv["name"]: graph_breaks = get_field(actual_csv, model, "graph_breaks") expected_graph_breaks = get_field(expected_csv, model, "graph_breaks") + flaky = model in flaky_models - if graph_breaks == expected_graph_breaks: - status = "PASS" + if expected_graph_breaks is None: + status = "MISSING:" + improved.append(model) + elif graph_breaks == expected_graph_breaks: + status = "PASS_BUT_FLAKY" if flaky else "PASS" print(f"{model:34} {status}") continue - elif graph_breaks > expected_graph_breaks: - status = "FAIL:" - failed.append(model) + if flaky: + status = "FAIL_BUT_FLAKY:" + else: + status = "FAIL:" + failed.append(model) elif graph_breaks < expected_graph_breaks: - status = "IMPROVED:" - improved.append(model) + if flaky: + status = "IMPROVED_BUT_FLAKY:" + else: + status = "IMPROVED:" + improved.append(model) print( - f"{model:34} {status:9} graph_breaks={graph_breaks}, expected={expected_graph_breaks}" + f"{model:34} {status:19} graph_breaks={graph_breaks}, expected={expected_graph_breaks}" ) msg = "" diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_huggingface_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_huggingface_inference.csv new file mode 100644 index 00000000000..fd57a3b4cbf --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_huggingface_inference.csv @@ -0,0 +1,185 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,0 + + + +AlbertForQuestionAnswering,pass,0 + + + +AllenaiLongformerBase,pass,4 + + + +BartForCausalLM,pass,0 + + + +BartForConditionalGeneration,pass,0 + + + +BertForMaskedLM,pass,0 + + + +BertForQuestionAnswering,pass,0 + + + +BlenderbotForCausalLM,pass_due_to_skip,0 + + + +BlenderbotSmallForCausalLM,pass,0 + + + +BlenderbotSmallForConditionalGeneration,pass,0 + + + +CamemBert,pass,0 + + + +DebertaForMaskedLM,pass,0 + + + +DebertaForQuestionAnswering,pass,0 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,pass,0 + + + +DistilBertForMaskedLM,pass,0 + + + +DistilBertForQuestionAnswering,pass,0 + + + +DistillGPT2,pass,0 + + + +ElectraForCausalLM,pass,0 + + + +ElectraForQuestionAnswering,pass,0 + + + +GPT2ForSequenceClassification,pass,0 + + + +GoogleFnet,pass,0 + + + +LayoutLMForMaskedLM,pass,0 + + + +LayoutLMForSequenceClassification,pass,0 + + + +M2M100ForConditionalGeneration,pass,0 + + + +MBartForCausalLM,pass,0 + + + +MBartForConditionalGeneration,pass,0 + + + +MT5ForConditionalGeneration,pass,0 + + + +MegatronBertForCausalLM,pass,0 + + + +MegatronBertForQuestionAnswering,pass,0 + + + +MobileBertForMaskedLM,pass,0 + + + +MobileBertForQuestionAnswering,pass,0 + + + +OPTForCausalLM,pass,0 + + + +PLBartForCausalLM,pass,0 + + + +PLBartForConditionalGeneration,pass,0 + + + +PegasusForCausalLM,pass,0 + + + +PegasusForConditionalGeneration,pass,0 + + + +RobertaForCausalLM,pass,0 + + + +RobertaForQuestionAnswering,pass,0 + + + +Speech2Text2ForCausalLM,pass,0 + + + +T5ForConditionalGeneration,pass,0 + + + +T5Small,pass,0 + + + +TrOCRForCausalLM,pass,0 + + + +XGLMForCausalLM,pass,0 + + + +XLNetLMHeadModel,pass,0 + + + +YituTechConvBert,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_huggingface_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_huggingface_training.csv new file mode 100644 index 00000000000..b54c6a84bc2 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_huggingface_training.csv @@ -0,0 +1,185 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,4 + + + +AlbertForQuestionAnswering,pass,5 + + + +AllenaiLongformerBase,pass,9 + + + +BartForCausalLM,pass,6 + + + +BartForConditionalGeneration,pass,8 + + + +BertForMaskedLM,pass,5 + + + +BertForQuestionAnswering,pass,5 + + + +BlenderbotForCausalLM,pass_due_to_skip,0 + + + +BlenderbotSmallForCausalLM,pass,6 + + + +BlenderbotSmallForConditionalGeneration,pass,8 + + + +CamemBert,pass,5 + + + +DebertaForMaskedLM,pass,5 + + + +DebertaForQuestionAnswering,pass,5 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,pass,4 + + + +DistilBertForMaskedLM,pass,5 + + + +DistilBertForQuestionAnswering,pass,5 + + + +DistillGPT2,pass,5 + + + +ElectraForCausalLM,pass,4 + + + +ElectraForQuestionAnswering,pass,5 + + + +GPT2ForSequenceClassification,pass,5 + + + +GoogleFnet,pass,5 + + + +LayoutLMForMaskedLM,pass,5 + + + +LayoutLMForSequenceClassification,pass,5 + + + +M2M100ForConditionalGeneration,pass,4 + + + +MBartForCausalLM,pass,6 + + + +MBartForConditionalGeneration,pass,8 + + + +MT5ForConditionalGeneration,pass,5 + + + +MegatronBertForCausalLM,pass,5 + + + +MegatronBertForQuestionAnswering,pass,5 + + + +MobileBertForMaskedLM,pass,3 + + + +MobileBertForQuestionAnswering,pass,3 + + + +OPTForCausalLM,pass,6 + + + +PLBartForCausalLM,pass,6 + + + +PLBartForConditionalGeneration,pass,8 + + + +PegasusForCausalLM,pass,6 + + + +PegasusForConditionalGeneration,pass,7 + + + +RobertaForCausalLM,pass,5 + + + +RobertaForQuestionAnswering,pass,5 + + + +Speech2Text2ForCausalLM,pass,6 + + + +T5ForConditionalGeneration,pass,5 + + + +T5Small,pass,5 + + + +TrOCRForCausalLM,pass,6 + + + +XGLMForCausalLM,pass,6 + + + +XLNetLMHeadModel,pass,5 + + + +YituTechConvBert,pass,5 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_timm_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_timm_inference.csv new file mode 100644 index 00000000000..6727624d0b0 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_timm_inference.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,0 + + + +beit_base_patch16_224,pass,0 + + + +botnet26t_256,pass,0 + + + +cait_m36_384,pass,0 + + + +coat_lite_mini,pass,0 + + + +convit_base,pass,0 + + + +convmixer_768_32,pass,0 + + + +convnext_base,pass,0 + + + +crossvit_9_240,pass,0 + + + +cspdarknet53,pass,0 + + + +deit_base_distilled_patch16_224,pass,0 + + + +dla102,pass,0 + + + +dm_nfnet_f0,pass,0 + + + +dpn107,pass,0 + + + +eca_botnext26ts_256,pass,0 + + + +eca_halonext26ts,pass,0 + + + +ese_vovnet19b_dw,pass,0 + + + +fbnetc_100,pass,0 + + + +fbnetv3_b,pass,0 + + + +gernet_l,pass,0 + + + +ghostnet_100,pass,0 + + + +gluon_inception_v3,pass,0 + + + +gmixer_24_224,pass,0 + + + +gmlp_s16_224,pass,0 + + + +hrnet_w18,pass,0 + + + +inception_v3,pass,0 + + + +jx_nest_base,pass,0 + + + +lcnet_050,pass,0 + + + +levit_128,eager_fail_to_run,0 + + + +mixer_b16_224,pass,0 + + + +mixnet_l,pass,0 + + + +mnasnet_100,pass,0 + + + +mobilenetv2_100,pass,0 + + + +mobilenetv3_large_100,pass,0 + + + +mobilevit_s,pass,0 + + + +nfnet_l0,pass,0 + + + +pit_b_224,pass,0 + + + +pnasnet5large,pass,0 + + + +poolformer_m36,pass,0 + + + +regnety_002,pass,0 + + + +repvgg_a2,pass,0 + + + +res2net101_26w_4s,pass,0 + + + +res2net50_14w_8s,pass,0 + + + +res2next50,pass,0 + + + +resmlp_12_224,pass,0 + + + +resnest101e,pass,0 + + + +rexnet_100,pass,0 + + + +sebotnet33ts_256,pass,0 + + + +selecsls42b,pass,0 + + + +spnasnet_100,pass,0 + + + +swin_base_patch4_window7_224,pass,0 + + + +swsl_resnext101_32x16d,pass,0 + + + +tf_efficientnet_b0,pass,0 + + + +tf_mixnet_l,pass,0 + + + +tinynet_a,pass,0 + + + +tnt_s_patch16_224,pass,0 + + + +twins_pcpvt_base,pass,0 + + + +visformer_small,pass,0 + + + +vit_base_patch16_224,pass,0 + + + +volo_d1_224,pass,0 + + + +xcit_large_24_p8_224,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_timm_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_timm_training.csv new file mode 100644 index 00000000000..afd846df7cb --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_timm_training.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,6 + + + +beit_base_patch16_224,pass,7 + + + +botnet26t_256,pass,6 + + + +cait_m36_384,pass,6 + + + +coat_lite_mini,pass,6 + + + +convit_base,pass,7 + + + +convmixer_768_32,pass,5 + + + +convnext_base,pass,7 + + + +crossvit_9_240,pass,7 + + + +cspdarknet53,pass,7 + + + +deit_base_distilled_patch16_224,pass,7 + + + +dla102,pass,7 + + + +dm_nfnet_f0,pass,6 + + + +dpn107,pass,6 + + + +eca_botnext26ts_256,pass,7 + + + +eca_halonext26ts,pass,7 + + + +ese_vovnet19b_dw,pass,7 + + + +fbnetc_100,pass,7 + + + +fbnetv3_b,pass,6 + + + +gernet_l,pass,6 + + + +ghostnet_100,pass,6 + + + +gluon_inception_v3,pass,7 + + + +gmixer_24_224,pass,6 + + + +gmlp_s16_224,pass,7 + + + +hrnet_w18,pass,5 + + + +inception_v3,pass,6 + + + +jx_nest_base,pass,7 + + + +lcnet_050,pass,6 + + + +levit_128,eager_fail_to_run,7 + + + +mixer_b16_224,pass,7 + + + +mixnet_l,pass,6 + + + +mnasnet_100,pass,7 + + + +mobilenetv2_100,pass,7 + + + +mobilenetv3_large_100,pass,7 + + + +mobilevit_s,pass,6 + + + +nfnet_l0,pass,7 + + + +pit_b_224,pass,6 + + + +pnasnet5large,pass,5 + + + +poolformer_m36,pass,6 + + + +regnety_002,pass,6 + + + +repvgg_a2,pass,7 + + + +res2net101_26w_4s,pass,6 + + + +res2net50_14w_8s,pass,6 + + + +res2next50,pass,6 + + + +resmlp_12_224,pass,6 + + + +resnest101e,pass,6 + + + +rexnet_100,pass,7 + + + +sebotnet33ts_256,pass,6 + + + +selecsls42b,pass,6 + + + +spnasnet_100,pass,7 + + + +swin_base_patch4_window7_224,pass,7 + + + +swsl_resnext101_32x16d,pass,6 + + + +tf_efficientnet_b0,pass,6 + + + +tf_mixnet_l,pass,6 + + + +tinynet_a,pass,6 + + + +tnt_s_patch16_224,pass,7 + + + +twins_pcpvt_base,pass,7 + + + +visformer_small,pass,7 + + + +vit_base_patch16_224,pass,7 + + + +volo_d1_224,pass,7 + + + +xcit_large_24_p8_224,pass_due_to_skip,7 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_torchbench_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_torchbench_inference.csv new file mode 100644 index 00000000000..148e0331849 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_torchbench_inference.csv @@ -0,0 +1,396 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,eager_fail_to_run,0 + + + +BERT_pytorch,pass,0 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,0 + + + +Super_SloMo,pass,0 + + + +alexnet,pass,0 + + + +basic_gnn_edgecnn,pass,0 + + + +basic_gnn_gcn,pass,6 + + + +basic_gnn_gin,pass,0 + + + +basic_gnn_sage,pass,0 + + + +cm3leon_generate,pass,4 + + + +dcgan,pass,0 + + + +demucs,pass,3 + + + +densenet121,pass,0 + + + +detectron2_fasterrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_fpn,eager_fail_to_run,0 + + + +detectron2_fcos_r_50_fpn,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_fpn,eager_fail_to_run,0 + + + +dlrm,pass,0 + + + +doctr_det_predictor,eager_fail_to_run,5 + + + +doctr_reco_predictor,eager_fail_to_run,4 + + + +drq,pass,0 + + + +fastNLP_Bert,pass,4 + + + +functorch_dp_cifar10,pass,0 + + + +functorch_maml_omniglot,pass,0 + + + +hf_Albert,pass,0 + + + +hf_Bart,pass,0 + + + +hf_Bert,pass,0 + + + +hf_Bert_large,pass,0 + + + +hf_BigBird,pass,9 + + + +hf_DistilBert,pass,0 + + + +hf_GPT2,pass,0 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_Longformer,pass,4 + + + +hf_Reformer,pass,5 + + + +hf_Roberta_base,pass,0 + + +hf_T5,pass,0 + + + +hf_T5_base,pass,0 + + + +hf_T5_generate,pass,5 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,0 + + + +hf_distil_whisper,pass,0 + + + +lennard_jones,pass,0 + + + +llama,pass,0 + + + +llama_v2_7b_16h,model_fail_to_load,0 + + + +llava,pass,0 + + + +maml,pass_due_to_skip,0 + + + +maml_omniglot,pass,0 + + + +microbench_unbacked_tolist_sum,pass,1 + + + +mnasnet1_0,pass,0 + + + +mobilenet_v2,pass,0 + + + +mobilenet_v2_quantized_qat,model_fail_to_load,0 + + + +mobilenet_v3_large,pass,0 + + + +moco,pass,7 + + + +moondream,pass,0 + + + +nanogpt,pass,0 + + + +nvidia_deeprecommender,pass,0 + + + +opacus_cifar10,pass,0 + + + +phlippe_densenet,pass,0 + + + +phlippe_resnet,pass,0 + + + +pyhpc_equation_of_state,pass,0 + + + +pyhpc_isoneutral_mixing,pass,0 + + + +pyhpc_turbulent_kinetic_energy,pass,0 + + + +pytorch_CycleGAN_and_pix2pix,pass,0 + + + +pytorch_stargan,pass,0 + + + +pytorch_unet,pass,0 + + + +resnet152,pass,0 + + + +resnet18,pass,0 + + + +resnet50,pass,0 + + + +resnet50_quantized_qat,model_fail_to_load,0 + + + +resnext50_32x4d,pass,0 + + + +sam,pass,0 + + + +sam_fast,pass,0 + + + +shufflenet_v2_x1_0,pass,0 + + + +soft_actor_critic,pass,0 + + + +speech_transformer,pass,10 + + + +squeezenet1_1,pass,0 + + + +stable_diffusion_text_encoder,model_fail_to_load,0 + + + +stable_diffusion_unet,model_fail_to_load,0 + + + +timm_efficientdet,pass,2 + + + +timm_efficientnet,pass,0 + + + +timm_nfnet,pass,0 + + + +timm_regnet,pass,0 + + + +timm_resnest,pass,0 + + + +timm_vision_transformer,pass,0 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,0 + + + +torch_multimodal_clip,pass,0 + + + +tts_angular,pass,2 + + + +vgg16,eager_two_runs_differ,0 + + + +vision_maskrcnn,pass,18 + + + +yolov3,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_torchbench_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_torchbench_training.csv new file mode 100644 index 00000000000..0f68cba8f87 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_eager_torchbench_training.csv @@ -0,0 +1,305 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,pass,6 + + + +BERT_pytorch,pass,6 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,6 + + + +Super_SloMo,pass,7 + + + +alexnet,pass,6 + + + +basic_gnn_edgecnn,pass,20 + + + +basic_gnn_gcn,pass,13 + + + +basic_gnn_gin,pass,7 + + + +basic_gnn_sage,pass,7 + + + +dcgan,pass,6 + + + +demucs,pass,9 + + + +densenet121,pass,6 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +dlrm,pass,6 + + + +drq,pass,7 + + + +fastNLP_Bert,pass,10 + + + +functorch_dp_cifar10,pass,7 + + + +functorch_maml_omniglot,pass,7 + + + +hf_Albert,pass,6 + + + +hf_Bart,pass,6 + + + +hf_Bert,pass,6 + + + +hf_Bert_large,pass,6 + + + +hf_BigBird,pass,6 + + + +hf_DistilBert,pass,6 + + + +hf_GPT2,pass,6 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_Longformer,pass,4 + + + +hf_Reformer,pass,23 + + + +hf_Roberta_base,pass,6 + + + +hf_T5,pass,0 + + + +hf_T5_base,eager_2nd_run_OOM,0 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,6 + + + +hf_distil_whisper,model_fail_to_load,0 + + + +lennard_jones,pass,7 + + + +llava,model_fail_to_load,0 + + + +maml_omniglot,pass,7 + + + +mnasnet1_0,pass,7 + + + +mobilenet_v2,pass,6 + + + +mobilenet_v2_quantized_qat,eager_fail_to_run,0 + + + +mobilenet_v3_large,pass,7 + + + +moco,pass,13 + + + +nanogpt,pass,7 + + + +nvidia_deeprecommender,pass,7 + + + +opacus_cifar10,eager_fail_to_run,0 + + + +phlippe_densenet,pass,6 + + + +phlippe_resnet,pass,6 + + + +pytorch_CycleGAN_and_pix2pix,pass,6 + + + +pytorch_stargan,pass,6 + + + +pytorch_unet,pass_due_to_skip,7 + + + +resnet152,pass,7 + + + +resnet18,pass,6 + + + +resnet50,pass,6 + + + +resnet50_quantized_qat,eager_fail_to_run,0 + + + +resnext50_32x4d,pass,7 + + + +sam,eager_fail_to_run,0 + + + +shufflenet_v2_x1_0,pass,6 + + + +soft_actor_critic,pass,6 + + + +speech_transformer,pass,16 + + + +squeezenet1_1,pass,6 + + + +stable_diffusion_text_encoder,pass,0 + + + +stable_diffusion_unet,pass_due_to_skip,0 + + + +timm_efficientdet,pass,2 + + + +timm_efficientnet,pass,7 + + + +timm_nfnet,pass,0 + + + +timm_regnet,pass,6 + + + +timm_resnest,pass,7 + + + +timm_vision_transformer,pass,6 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,6 + + + +torch_multimodal_clip,pass,7 + + + +tts_angular,pass,9 + + + +vgg16,pass,6 + + + +vision_maskrcnn,pass,35 + + + +yolov3,pass,8 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_inductor_huggingface_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_inductor_huggingface_inference.csv new file mode 100644 index 00000000000..1cafcbe5567 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_inductor_huggingface_inference.csv @@ -0,0 +1,181 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,0 + + + +AlbertForQuestionAnswering,pass,0 + + + +BartForCausalLM,pass,0 + + + +BartForConditionalGeneration,pass,0 + + + +BertForMaskedLM,pass,0 + + + +BertForQuestionAnswering,pass,0 + + + +BlenderbotForCausalLM,pass_due_to_skip,0 + + + +BlenderbotSmallForCausalLM,pass,0 + + + +BlenderbotSmallForConditionalGeneration,pass,0 + + + +CamemBert,pass,0 + + + +DebertaForMaskedLM,pass,0 + + + +DebertaForQuestionAnswering,pass,0 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,pass,0 + + + +DistilBertForMaskedLM,pass,0 + + + +DistilBertForQuestionAnswering,pass,0 + + + +DistillGPT2,pass,0 + + + +ElectraForCausalLM,pass,0 + + + +ElectraForQuestionAnswering,pass,0 + + + +GPT2ForSequenceClassification,pass,0 + + + +GoogleFnet,pass,0 + + + +LayoutLMForMaskedLM,pass,0 + + + +LayoutLMForSequenceClassification,pass,0 + + + +M2M100ForConditionalGeneration,pass,0 + + + +MBartForCausalLM,pass,0 + + + +MBartForConditionalGeneration,pass,0 + + + +MT5ForConditionalGeneration,pass,0 + + + +MegatronBertForCausalLM,pass,0 + + + +MegatronBertForQuestionAnswering,pass,0 + + + +MobileBertForMaskedLM,pass,0 + + + +MobileBertForQuestionAnswering,pass,0 + + + +OPTForCausalLM,pass,0 + + + +PLBartForCausalLM,pass,0 + + + +PLBartForConditionalGeneration,pass,0 + + + +PegasusForCausalLM,pass,0 + + + +PegasusForConditionalGeneration,pass,0 + + + +RobertaForCausalLM,pass,0 + + + +RobertaForQuestionAnswering,pass,0 + + + +Speech2Text2ForCausalLM,pass,0 + + + +T5ForConditionalGeneration,pass,0 + + + +T5Small,pass,0 + + + +TrOCRForCausalLM,pass,0 + + + +XGLMForCausalLM,pass,0 + + + +XLNetLMHeadModel,pass,0 + + + +YituTechConvBert,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_inductor_timm_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_inductor_timm_inference.csv new file mode 100644 index 00000000000..c889ba0e8d2 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_inductor_timm_inference.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,0 + + + +beit_base_patch16_224,pass,0 + + + +botnet26t_256,pass,0 + + + +cait_m36_384,pass,0 + + + +coat_lite_mini,pass,0 + + + +convit_base,pass,0 + + + +convmixer_768_32,pass,0 + + + +convnext_base,pass,0 + + + +crossvit_9_240,pass,0 + + + +cspdarknet53,pass,0 + + + +deit_base_distilled_patch16_224,pass,0 + + + +dla102,pass,0 + + + +dm_nfnet_f0,pass,0 + + + +dpn107,pass,0 + + + +eca_botnext26ts_256,pass,0 + + + +eca_halonext26ts,pass,0 + + + +ese_vovnet19b_dw,pass,0 + + + +fbnetc_100,pass,0 + + + +fbnetv3_b,pass,0 + + + +gernet_l,pass,0 + + + +ghostnet_100,pass,0 + + + +gluon_inception_v3,pass,0 + + + +gmixer_24_224,pass,0 + + + +gmlp_s16_224,pass,0 + + + +hrnet_w18,pass,0 + + + +inception_v3,pass,0 + + + +jx_nest_base,pass,0 + + + +lcnet_050,pass,0 + + + +levit_128,pass,0 + + + +mixer_b16_224,pass,0 + + + +mixnet_l,pass,0 + + + +mnasnet_100,pass,0 + + + +mobilenetv2_100,pass,0 + + + +mobilenetv3_large_100,pass,0 + + + +mobilevit_s,pass,0 + + + +nfnet_l0,pass,0 + + + +pit_b_224,pass,0 + + + +pnasnet5large,pass,0 + + + +poolformer_m36,pass,0 + + + +regnety_002,pass,0 + + + +repvgg_a2,pass,0 + + + +res2net101_26w_4s,pass,0 + + + +res2net50_14w_8s,pass,0 + + + +res2next50,pass,0 + + + +resmlp_12_224,pass,0 + + + +resnest101e,pass,0 + + + +rexnet_100,pass,0 + + + +sebotnet33ts_256,pass,0 + + + +selecsls42b,pass,0 + + + +spnasnet_100,pass,0 + + + +swin_base_patch4_window7_224,pass,0 + + + +swsl_resnext101_32x16d,pass,0 + + + +tf_efficientnet_b0,pass,0 + + + +tf_mixnet_l,pass,0 + + + +tinynet_a,pass,0 + + + +tnt_s_patch16_224,pass,0 + + + +twins_pcpvt_base,pass,0 + + + +visformer_small,pass,0 + + + +vit_base_patch16_224,pass,0 + + + +volo_d1_224,pass,0 + + + +xcit_large_24_p8_224,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_inductor_torchbench_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_inductor_torchbench_inference.csv new file mode 100644 index 00000000000..cc4ef192ca5 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/aot_inductor_torchbench_inference.csv @@ -0,0 +1,349 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,eager_fail_to_run,0 + + + +BERT_pytorch,pass,0 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,0 + + + +Super_SloMo,pass,0 + + + +alexnet,pass,0 + + + +basic_gnn_edgecnn,pass,0 + + + +basic_gnn_gcn,pass,0 + + + +basic_gnn_gin,pass,0 + + + +basic_gnn_sage,pass,0 + + + +dcgan,pass,0 + + + +demucs,pass,0 + + + +densenet121,pass,0 + + + +detectron2_fasterrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_fpn,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_fpn,eager_fail_to_run,0 + + + +dlrm,pass,0 + + + +doctr_det_predictor,fail_to_run,0 + + + +doctr_reco_predictor,fail_to_run,0 + + + +drq,fail_to_run,0 + + + +functorch_dp_cifar10,pass,0 + + + +functorch_maml_omniglot,pass,0 + + + +hf_Albert,pass,0 + + + +hf_Bart,pass,0 + + + +hf_Bert,pass,0 + + + +hf_Bert_large,pass,0 + + + +hf_BigBird,fail_accuracy,0 + + + +hf_DistilBert,pass,0 + + + +hf_GPT2,pass,0 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_T5,pass,0 + + + +hf_T5_base,eager_fail_to_run,0 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,0 + + + +hf_distil_whisper,pass,0 + + + +lennard_jones,pass,0 + + + +llama,fail_to_run,0 + + + +llama_v2_7b_16h,model_fail_to_load,0 + + + +llava,model_fail_to_load,0 + + + +maml,pass_due_to_skip,0 + + + +maml_omniglot,pass,0 + + + +mnasnet1_0,pass,0 + + + +mobilenet_v2,pass,0 + + + +mobilenet_v2_quantized_qat,model_fail_to_load,0 + + + +mobilenet_v3_large,pass,0 + + + +moco,fail_to_run,0 + + + +moondream,model_fail_to_load,0 + + + +nanogpt,pass,0 + + + +nvidia_deeprecommender,pass,0 + + + +phlippe_densenet,pass,0 + + + +phlippe_resnet,pass,0 + + + +pyhpc_equation_of_state,pass,0 + + + +pyhpc_isoneutral_mixing,pass,0 + + + +pyhpc_turbulent_kinetic_energy,pass,0 + + + +pytorch_CycleGAN_and_pix2pix,pass,0 + + + +pytorch_stargan,pass,0 + + + +pytorch_unet,pass,0 + + + +resnet152,pass,0 + + + +resnet18,pass,0 + + + +resnet50,pass,0 + + + +resnet50_quantized_qat,model_fail_to_load,0 + + + +resnext50_32x4d,pass,0 + + + +sam,pass,0 + + + +sam_fast,fail_to_run,0 + + + +shufflenet_v2_x1_0,pass,0 + + + +soft_actor_critic,fail_to_run,0 + + + +squeezenet1_1,pass,0 + + + +stable_diffusion_text_encoder,pass,0 + + + +stable_diffusion_unet,pass_due_to_skip,0 + + + +timm_efficientnet,pass,0 + + + +timm_regnet,pass,0 + + + +timm_resnest,pass,0 + + + +timm_vision_transformer,pass,0 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,0 + + + +torch_multimodal_clip,pass,0 + + + +tts_angular,fail_to_run,0 + + + +vgg16,pass,0 + + + +vision_maskrcnn,fail_to_run,0 + + + +yolov3,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_huggingface_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_huggingface_inference.csv new file mode 100644 index 00000000000..fd57a3b4cbf --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_huggingface_inference.csv @@ -0,0 +1,185 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,0 + + + +AlbertForQuestionAnswering,pass,0 + + + +AllenaiLongformerBase,pass,4 + + + +BartForCausalLM,pass,0 + + + +BartForConditionalGeneration,pass,0 + + + +BertForMaskedLM,pass,0 + + + +BertForQuestionAnswering,pass,0 + + + +BlenderbotForCausalLM,pass_due_to_skip,0 + + + +BlenderbotSmallForCausalLM,pass,0 + + + +BlenderbotSmallForConditionalGeneration,pass,0 + + + +CamemBert,pass,0 + + + +DebertaForMaskedLM,pass,0 + + + +DebertaForQuestionAnswering,pass,0 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,pass,0 + + + +DistilBertForMaskedLM,pass,0 + + + +DistilBertForQuestionAnswering,pass,0 + + + +DistillGPT2,pass,0 + + + +ElectraForCausalLM,pass,0 + + + +ElectraForQuestionAnswering,pass,0 + + + +GPT2ForSequenceClassification,pass,0 + + + +GoogleFnet,pass,0 + + + +LayoutLMForMaskedLM,pass,0 + + + +LayoutLMForSequenceClassification,pass,0 + + + +M2M100ForConditionalGeneration,pass,0 + + + +MBartForCausalLM,pass,0 + + + +MBartForConditionalGeneration,pass,0 + + + +MT5ForConditionalGeneration,pass,0 + + + +MegatronBertForCausalLM,pass,0 + + + +MegatronBertForQuestionAnswering,pass,0 + + + +MobileBertForMaskedLM,pass,0 + + + +MobileBertForQuestionAnswering,pass,0 + + + +OPTForCausalLM,pass,0 + + + +PLBartForCausalLM,pass,0 + + + +PLBartForConditionalGeneration,pass,0 + + + +PegasusForCausalLM,pass,0 + + + +PegasusForConditionalGeneration,pass,0 + + + +RobertaForCausalLM,pass,0 + + + +RobertaForQuestionAnswering,pass,0 + + + +Speech2Text2ForCausalLM,pass,0 + + + +T5ForConditionalGeneration,pass,0 + + + +T5Small,pass,0 + + + +TrOCRForCausalLM,pass,0 + + + +XGLMForCausalLM,pass,0 + + + +XLNetLMHeadModel,pass,0 + + + +YituTechConvBert,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_huggingface_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_huggingface_training.csv new file mode 100644 index 00000000000..b54c6a84bc2 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_huggingface_training.csv @@ -0,0 +1,185 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,4 + + + +AlbertForQuestionAnswering,pass,5 + + + +AllenaiLongformerBase,pass,9 + + + +BartForCausalLM,pass,6 + + + +BartForConditionalGeneration,pass,8 + + + +BertForMaskedLM,pass,5 + + + +BertForQuestionAnswering,pass,5 + + + +BlenderbotForCausalLM,pass_due_to_skip,0 + + + +BlenderbotSmallForCausalLM,pass,6 + + + +BlenderbotSmallForConditionalGeneration,pass,8 + + + +CamemBert,pass,5 + + + +DebertaForMaskedLM,pass,5 + + + +DebertaForQuestionAnswering,pass,5 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,pass,4 + + + +DistilBertForMaskedLM,pass,5 + + + +DistilBertForQuestionAnswering,pass,5 + + + +DistillGPT2,pass,5 + + + +ElectraForCausalLM,pass,4 + + + +ElectraForQuestionAnswering,pass,5 + + + +GPT2ForSequenceClassification,pass,5 + + + +GoogleFnet,pass,5 + + + +LayoutLMForMaskedLM,pass,5 + + + +LayoutLMForSequenceClassification,pass,5 + + + +M2M100ForConditionalGeneration,pass,4 + + + +MBartForCausalLM,pass,6 + + + +MBartForConditionalGeneration,pass,8 + + + +MT5ForConditionalGeneration,pass,5 + + + +MegatronBertForCausalLM,pass,5 + + + +MegatronBertForQuestionAnswering,pass,5 + + + +MobileBertForMaskedLM,pass,3 + + + +MobileBertForQuestionAnswering,pass,3 + + + +OPTForCausalLM,pass,6 + + + +PLBartForCausalLM,pass,6 + + + +PLBartForConditionalGeneration,pass,8 + + + +PegasusForCausalLM,pass,6 + + + +PegasusForConditionalGeneration,pass,7 + + + +RobertaForCausalLM,pass,5 + + + +RobertaForQuestionAnswering,pass,5 + + + +Speech2Text2ForCausalLM,pass,6 + + + +T5ForConditionalGeneration,pass,5 + + + +T5Small,pass,5 + + + +TrOCRForCausalLM,pass,6 + + + +XGLMForCausalLM,pass,6 + + + +XLNetLMHeadModel,pass,5 + + + +YituTechConvBert,pass,5 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_timm_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_timm_inference.csv new file mode 100644 index 00000000000..6727624d0b0 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_timm_inference.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,0 + + + +beit_base_patch16_224,pass,0 + + + +botnet26t_256,pass,0 + + + +cait_m36_384,pass,0 + + + +coat_lite_mini,pass,0 + + + +convit_base,pass,0 + + + +convmixer_768_32,pass,0 + + + +convnext_base,pass,0 + + + +crossvit_9_240,pass,0 + + + +cspdarknet53,pass,0 + + + +deit_base_distilled_patch16_224,pass,0 + + + +dla102,pass,0 + + + +dm_nfnet_f0,pass,0 + + + +dpn107,pass,0 + + + +eca_botnext26ts_256,pass,0 + + + +eca_halonext26ts,pass,0 + + + +ese_vovnet19b_dw,pass,0 + + + +fbnetc_100,pass,0 + + + +fbnetv3_b,pass,0 + + + +gernet_l,pass,0 + + + +ghostnet_100,pass,0 + + + +gluon_inception_v3,pass,0 + + + +gmixer_24_224,pass,0 + + + +gmlp_s16_224,pass,0 + + + +hrnet_w18,pass,0 + + + +inception_v3,pass,0 + + + +jx_nest_base,pass,0 + + + +lcnet_050,pass,0 + + + +levit_128,eager_fail_to_run,0 + + + +mixer_b16_224,pass,0 + + + +mixnet_l,pass,0 + + + +mnasnet_100,pass,0 + + + +mobilenetv2_100,pass,0 + + + +mobilenetv3_large_100,pass,0 + + + +mobilevit_s,pass,0 + + + +nfnet_l0,pass,0 + + + +pit_b_224,pass,0 + + + +pnasnet5large,pass,0 + + + +poolformer_m36,pass,0 + + + +regnety_002,pass,0 + + + +repvgg_a2,pass,0 + + + +res2net101_26w_4s,pass,0 + + + +res2net50_14w_8s,pass,0 + + + +res2next50,pass,0 + + + +resmlp_12_224,pass,0 + + + +resnest101e,pass,0 + + + +rexnet_100,pass,0 + + + +sebotnet33ts_256,pass,0 + + + +selecsls42b,pass,0 + + + +spnasnet_100,pass,0 + + + +swin_base_patch4_window7_224,pass,0 + + + +swsl_resnext101_32x16d,pass,0 + + + +tf_efficientnet_b0,pass,0 + + + +tf_mixnet_l,pass,0 + + + +tinynet_a,pass,0 + + + +tnt_s_patch16_224,pass,0 + + + +twins_pcpvt_base,pass,0 + + + +visformer_small,pass,0 + + + +vit_base_patch16_224,pass,0 + + + +volo_d1_224,pass,0 + + + +xcit_large_24_p8_224,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_timm_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_timm_training.csv new file mode 100644 index 00000000000..afd846df7cb --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_timm_training.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,6 + + + +beit_base_patch16_224,pass,7 + + + +botnet26t_256,pass,6 + + + +cait_m36_384,pass,6 + + + +coat_lite_mini,pass,6 + + + +convit_base,pass,7 + + + +convmixer_768_32,pass,5 + + + +convnext_base,pass,7 + + + +crossvit_9_240,pass,7 + + + +cspdarknet53,pass,7 + + + +deit_base_distilled_patch16_224,pass,7 + + + +dla102,pass,7 + + + +dm_nfnet_f0,pass,6 + + + +dpn107,pass,6 + + + +eca_botnext26ts_256,pass,7 + + + +eca_halonext26ts,pass,7 + + + +ese_vovnet19b_dw,pass,7 + + + +fbnetc_100,pass,7 + + + +fbnetv3_b,pass,6 + + + +gernet_l,pass,6 + + + +ghostnet_100,pass,6 + + + +gluon_inception_v3,pass,7 + + + +gmixer_24_224,pass,6 + + + +gmlp_s16_224,pass,7 + + + +hrnet_w18,pass,5 + + + +inception_v3,pass,6 + + + +jx_nest_base,pass,7 + + + +lcnet_050,pass,6 + + + +levit_128,eager_fail_to_run,7 + + + +mixer_b16_224,pass,7 + + + +mixnet_l,pass,6 + + + +mnasnet_100,pass,7 + + + +mobilenetv2_100,pass,7 + + + +mobilenetv3_large_100,pass,7 + + + +mobilevit_s,pass,6 + + + +nfnet_l0,pass,7 + + + +pit_b_224,pass,6 + + + +pnasnet5large,pass,5 + + + +poolformer_m36,pass,6 + + + +regnety_002,pass,6 + + + +repvgg_a2,pass,7 + + + +res2net101_26w_4s,pass,6 + + + +res2net50_14w_8s,pass,6 + + + +res2next50,pass,6 + + + +resmlp_12_224,pass,6 + + + +resnest101e,pass,6 + + + +rexnet_100,pass,7 + + + +sebotnet33ts_256,pass,6 + + + +selecsls42b,pass,6 + + + +spnasnet_100,pass,7 + + + +swin_base_patch4_window7_224,pass,7 + + + +swsl_resnext101_32x16d,pass,6 + + + +tf_efficientnet_b0,pass,6 + + + +tf_mixnet_l,pass,6 + + + +tinynet_a,pass,6 + + + +tnt_s_patch16_224,pass,7 + + + +twins_pcpvt_base,pass,7 + + + +visformer_small,pass,7 + + + +vit_base_patch16_224,pass,7 + + + +volo_d1_224,pass,7 + + + +xcit_large_24_p8_224,pass_due_to_skip,7 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_torchbench_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_torchbench_inference.csv new file mode 100644 index 00000000000..0ceb0901dee --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_torchbench_inference.csv @@ -0,0 +1,396 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,eager_fail_to_run,0 + + + +BERT_pytorch,pass,0 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,0 + + + +Super_SloMo,pass,0 + + + +alexnet,eager_two_runs_differ,0 + + + +basic_gnn_edgecnn,pass,0 + + + +basic_gnn_gcn,pass,6 + + + +basic_gnn_gin,pass,0 + + + +basic_gnn_sage,pass,0 + + + +cm3leon_generate,pass,4 + + + +dcgan,pass,0 + + + +demucs,pass,3 + + + +densenet121,pass,0 + + + +detectron2_fasterrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_fpn,eager_fail_to_run,0 + + + +detectron2_fcos_r_50_fpn,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_fpn,eager_fail_to_run,0 + + + +dlrm,pass,0 + + + +doctr_det_predictor,eager_fail_to_run,5 + + + +doctr_reco_predictor,eager_fail_to_run,4 + + + +drq,pass,0 + + + +fastNLP_Bert,pass,4 + + + +functorch_dp_cifar10,pass,0 + + + +functorch_maml_omniglot,pass,0 + + + +hf_Albert,pass,0 + + + +hf_Bart,pass,0 + + + +hf_Bert,pass,0 + + + +hf_Bert_large,pass,0 + + + +hf_BigBird,pass,9 + + + +hf_DistilBert,pass,0 + + + +hf_GPT2,pass,0 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_Longformer,pass,4 + + + +hf_Reformer,pass,5 + + + +hf_Roberta_base,pass,0 + + +hf_T5,pass,0 + + + +hf_T5_base,pass,0 + + + +hf_T5_generate,pass,5 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,0 + + + +hf_distil_whisper,pass,0 + + + +lennard_jones,pass,0 + + + +llama,pass,0 + + + +llama_v2_7b_16h,model_fail_to_load,0 + + + +llava,pass,0 + + + +maml,pass_due_to_skip,0 + + + +maml_omniglot,pass,0 + + + +microbench_unbacked_tolist_sum,pass,1 + + + +mnasnet1_0,pass,0 + + + +mobilenet_v2,pass,0 + + + +mobilenet_v2_quantized_qat,model_fail_to_load,0 + + + +mobilenet_v3_large,pass,0 + + + +moco,pass,7 + + + +moondream,pass,0 + + + +nanogpt,pass,0 + + + +nvidia_deeprecommender,pass,0 + + + +opacus_cifar10,pass,0 + + + +phlippe_densenet,pass,0 + + + +phlippe_resnet,pass,0 + + + +pyhpc_equation_of_state,pass,0 + + + +pyhpc_isoneutral_mixing,pass,0 + + + +pyhpc_turbulent_kinetic_energy,pass,0 + + + +pytorch_CycleGAN_and_pix2pix,pass,0 + + + +pytorch_stargan,pass,0 + + + +pytorch_unet,pass,0 + + + +resnet152,pass,0 + + + +resnet18,pass,0 + + + +resnet50,pass,0 + + + +resnet50_quantized_qat,model_fail_to_load,0 + + + +resnext50_32x4d,pass,0 + + + +sam,pass,0 + + + +sam_fast,eager_fail_to_run,0 + + + +shufflenet_v2_x1_0,pass,0 + + + +soft_actor_critic,pass,0 + + + +speech_transformer,pass,10 + + + +squeezenet1_1,pass,0 + + + +stable_diffusion_text_encoder,model_fail_to_load,0 + + + +stable_diffusion_unet,model_fail_to_load,0 + + + +timm_efficientdet,pass,2 + + + +timm_efficientnet,pass,0 + + + +timm_nfnet,pass,0 + + + +timm_regnet,pass,0 + + + +timm_resnest,pass,0 + + + +timm_vision_transformer,pass,0 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,0 + + + +torch_multimodal_clip,pass,0 + + + +tts_angular,pass,2 + + + +vgg16,eager_two_runs_differ,0 + + + +vision_maskrcnn,pass,18 + + + +yolov3,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_torchbench_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_torchbench_training.csv new file mode 100644 index 00000000000..17346089a4e --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_aot_eager_torchbench_training.csv @@ -0,0 +1,301 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,fail_to_run,3 + + + +BERT_pytorch,pass,6 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,6 + + + +Super_SloMo,pass,7 + + + +alexnet,pass,6 + + + +basic_gnn_edgecnn,pass,20 + + + +basic_gnn_gcn,pass,13 + + + +basic_gnn_gin,pass,7 + + + +basic_gnn_sage,pass,7 + + + +dcgan,pass,6 + + + +demucs,pass,9 + + + +densenet121,pass,6 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +dlrm,pass,6 + + + +drq,pass,7 + + + +fastNLP_Bert,pass,10 + + + +functorch_dp_cifar10,pass,7 + + + +functorch_maml_omniglot,pass,7 + + + +hf_Albert,pass,6 + + + +hf_Bart,pass,6 + + + +hf_Bert,pass,6 + + + +hf_Bert_large,pass,6 + + + +hf_BigBird,fail_to_run,3 + + + +hf_DistilBert,pass,6 + + + +hf_GPT2,pass,6 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_Longformer,pass,4 + + + +hf_Reformer,pass,23 + + + +hf_Roberta_base,pass,6 + + + +hf_T5,pass,0 + + + +hf_T5_base,eager_2nd_run_OOM,0 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,6 + + + +hf_distil_whisper,model_fail_to_load,0 + + + +lennard_jones,pass,7 + + + +llava,model_fail_to_load,0 + + + +maml_omniglot,pass,7 + + + +mnasnet1_0,pass,7 + + + +mobilenet_v2,pass,6 + + + +mobilenet_v2_quantized_qat,eager_fail_to_run,0 + + + +mobilenet_v3_large,pass,7 + + + +moco,pass,13 + + + +nanogpt,pass,7 + + + +nvidia_deeprecommender,pass,7 + + + +opacus_cifar10,eager_fail_to_run,0 + + + +phlippe_densenet,pass,6 + + + +phlippe_resnet,pass,6 + + + +pytorch_CycleGAN_and_pix2pix,pass,6 + + + +pytorch_stargan,pass,6 + + + +pytorch_unet,pass_due_to_skip,7 + + + +resnet152,pass,7 + + + +resnet18,pass,6 + + + +resnet50,pass,6 + + + +resnet50_quantized_qat,eager_fail_to_run,0 + + + +resnext50_32x4d,pass,7 + + + +sam,eager_fail_to_run,0 + + + +shufflenet_v2_x1_0,pass,6 + + + +soft_actor_critic,pass,6 + + + +squeezenet1_1,pass,6 + + + +stable_diffusion_text_encoder,pass,0 + + + +stable_diffusion_unet,pass_due_to_skip,0 + + + +timm_efficientdet,pass,2 + + + +timm_efficientnet,pass,7 + + + +timm_nfnet,pass,0 + + + +timm_regnet,pass,6 + + + +timm_resnest,pass,7 + + + +timm_vision_transformer,pass,6 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,6 + + + +torch_multimodal_clip,pass,7 + + + +tts_angular,pass,9 + + + +vgg16,pass,6 + + + +vision_maskrcnn,pass,35 + + + +yolov3,pass,8 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_huggingface_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_huggingface_inference.csv new file mode 100644 index 00000000000..fd57a3b4cbf --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_huggingface_inference.csv @@ -0,0 +1,185 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,0 + + + +AlbertForQuestionAnswering,pass,0 + + + +AllenaiLongformerBase,pass,4 + + + +BartForCausalLM,pass,0 + + + +BartForConditionalGeneration,pass,0 + + + +BertForMaskedLM,pass,0 + + + +BertForQuestionAnswering,pass,0 + + + +BlenderbotForCausalLM,pass_due_to_skip,0 + + + +BlenderbotSmallForCausalLM,pass,0 + + + +BlenderbotSmallForConditionalGeneration,pass,0 + + + +CamemBert,pass,0 + + + +DebertaForMaskedLM,pass,0 + + + +DebertaForQuestionAnswering,pass,0 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,pass,0 + + + +DistilBertForMaskedLM,pass,0 + + + +DistilBertForQuestionAnswering,pass,0 + + + +DistillGPT2,pass,0 + + + +ElectraForCausalLM,pass,0 + + + +ElectraForQuestionAnswering,pass,0 + + + +GPT2ForSequenceClassification,pass,0 + + + +GoogleFnet,pass,0 + + + +LayoutLMForMaskedLM,pass,0 + + + +LayoutLMForSequenceClassification,pass,0 + + + +M2M100ForConditionalGeneration,pass,0 + + + +MBartForCausalLM,pass,0 + + + +MBartForConditionalGeneration,pass,0 + + + +MT5ForConditionalGeneration,pass,0 + + + +MegatronBertForCausalLM,pass,0 + + + +MegatronBertForQuestionAnswering,pass,0 + + + +MobileBertForMaskedLM,pass,0 + + + +MobileBertForQuestionAnswering,pass,0 + + + +OPTForCausalLM,pass,0 + + + +PLBartForCausalLM,pass,0 + + + +PLBartForConditionalGeneration,pass,0 + + + +PegasusForCausalLM,pass,0 + + + +PegasusForConditionalGeneration,pass,0 + + + +RobertaForCausalLM,pass,0 + + + +RobertaForQuestionAnswering,pass,0 + + + +Speech2Text2ForCausalLM,pass,0 + + + +T5ForConditionalGeneration,pass,0 + + + +T5Small,pass,0 + + + +TrOCRForCausalLM,pass,0 + + + +XGLMForCausalLM,pass,0 + + + +XLNetLMHeadModel,pass,0 + + + +YituTechConvBert,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_huggingface_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_huggingface_training.csv new file mode 100644 index 00000000000..8202281ed9b --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_huggingface_training.csv @@ -0,0 +1,185 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,4 + + + +AlbertForQuestionAnswering,pass,5 + + + +AllenaiLongformerBase,pass,9 + + + +BartForCausalLM,pass,6 + + + +BartForConditionalGeneration,pass,8 + + + +BertForMaskedLM,pass,5 + + + +BertForQuestionAnswering,pass,5 + + + +BlenderbotForCausalLM,eager_fail_to_run,0 + + + +BlenderbotSmallForCausalLM,pass,6 + + + +BlenderbotSmallForConditionalGeneration,pass,8 + + + +CamemBert,pass,5 + + + +DebertaForMaskedLM,pass,5 + + + +DebertaForQuestionAnswering,pass,5 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,eager_1st_run_OOM,0 + + + +DistilBertForMaskedLM,pass,5 + + + +DistilBertForQuestionAnswering,pass,5 + + + +DistillGPT2,pass,5 + + + +ElectraForCausalLM,pass,4 + + + +ElectraForQuestionAnswering,pass,5 + + + +GPT2ForSequenceClassification,pass,5 + + + +GoogleFnet,pass,5 + + + +LayoutLMForMaskedLM,pass,5 + + + +LayoutLMForSequenceClassification,pass,5 + + + +M2M100ForConditionalGeneration,pass,4 + + + +MBartForCausalLM,pass,6 + + + +MBartForConditionalGeneration,pass,8 + + + +MT5ForConditionalGeneration,pass,5 + + + +MegatronBertForCausalLM,pass,5 + + + +MegatronBertForQuestionAnswering,pass,5 + + + +MobileBertForMaskedLM,pass,3 + + + +MobileBertForQuestionAnswering,pass,3 + + + +OPTForCausalLM,pass,6 + + + +PLBartForCausalLM,pass,6 + + + +PLBartForConditionalGeneration,pass,8 + + + +PegasusForCausalLM,pass,6 + + + +PegasusForConditionalGeneration,pass,7 + + + +RobertaForCausalLM,pass,5 + + + +RobertaForQuestionAnswering,pass,5 + + + +Speech2Text2ForCausalLM,pass,6 + + + +T5ForConditionalGeneration,pass,5 + + + +T5Small,pass,5 + + + +TrOCRForCausalLM,pass,6 + + + +XGLMForCausalLM,pass,6 + + + +XLNetLMHeadModel,pass,5 + + + +YituTechConvBert,pass,5 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_timm_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_timm_inference.csv new file mode 100644 index 00000000000..c889ba0e8d2 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_timm_inference.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,0 + + + +beit_base_patch16_224,pass,0 + + + +botnet26t_256,pass,0 + + + +cait_m36_384,pass,0 + + + +coat_lite_mini,pass,0 + + + +convit_base,pass,0 + + + +convmixer_768_32,pass,0 + + + +convnext_base,pass,0 + + + +crossvit_9_240,pass,0 + + + +cspdarknet53,pass,0 + + + +deit_base_distilled_patch16_224,pass,0 + + + +dla102,pass,0 + + + +dm_nfnet_f0,pass,0 + + + +dpn107,pass,0 + + + +eca_botnext26ts_256,pass,0 + + + +eca_halonext26ts,pass,0 + + + +ese_vovnet19b_dw,pass,0 + + + +fbnetc_100,pass,0 + + + +fbnetv3_b,pass,0 + + + +gernet_l,pass,0 + + + +ghostnet_100,pass,0 + + + +gluon_inception_v3,pass,0 + + + +gmixer_24_224,pass,0 + + + +gmlp_s16_224,pass,0 + + + +hrnet_w18,pass,0 + + + +inception_v3,pass,0 + + + +jx_nest_base,pass,0 + + + +lcnet_050,pass,0 + + + +levit_128,pass,0 + + + +mixer_b16_224,pass,0 + + + +mixnet_l,pass,0 + + + +mnasnet_100,pass,0 + + + +mobilenetv2_100,pass,0 + + + +mobilenetv3_large_100,pass,0 + + + +mobilevit_s,pass,0 + + + +nfnet_l0,pass,0 + + + +pit_b_224,pass,0 + + + +pnasnet5large,pass,0 + + + +poolformer_m36,pass,0 + + + +regnety_002,pass,0 + + + +repvgg_a2,pass,0 + + + +res2net101_26w_4s,pass,0 + + + +res2net50_14w_8s,pass,0 + + + +res2next50,pass,0 + + + +resmlp_12_224,pass,0 + + + +resnest101e,pass,0 + + + +rexnet_100,pass,0 + + + +sebotnet33ts_256,pass,0 + + + +selecsls42b,pass,0 + + + +spnasnet_100,pass,0 + + + +swin_base_patch4_window7_224,pass,0 + + + +swsl_resnext101_32x16d,pass,0 + + + +tf_efficientnet_b0,pass,0 + + + +tf_mixnet_l,pass,0 + + + +tinynet_a,pass,0 + + + +tnt_s_patch16_224,pass,0 + + + +twins_pcpvt_base,pass,0 + + + +visformer_small,pass,0 + + + +vit_base_patch16_224,pass,0 + + + +volo_d1_224,pass,0 + + + +xcit_large_24_p8_224,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_timm_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_timm_training.csv new file mode 100644 index 00000000000..e5464160d32 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_timm_training.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,6 + + + +beit_base_patch16_224,pass,7 + + + +botnet26t_256,pass,6 + + + +cait_m36_384,eager_fail_to_run,0 + + + +coat_lite_mini,pass,6 + + + +convit_base,pass,7 + + + +convmixer_768_32,pass,5 + + + +convnext_base,pass,7 + + + +crossvit_9_240,pass,7 + + + +cspdarknet53,pass,7 + + + +deit_base_distilled_patch16_224,pass,7 + + + +dla102,pass,7 + + + +dm_nfnet_f0,pass,6 + + + +dpn107,pass,6 + + + +eca_botnext26ts_256,pass,7 + + + +eca_halonext26ts,pass,7 + + + +ese_vovnet19b_dw,pass,7 + + + +fbnetc_100,pass,7 + + + +fbnetv3_b,pass,6 + + + +gernet_l,pass,6 + + + +ghostnet_100,pass,6 + + + +gluon_inception_v3,pass,7 + + + +gmixer_24_224,pass,6 + + + +gmlp_s16_224,pass,7 + + + +hrnet_w18,pass,5 + + + +inception_v3,pass,6 + + + +jx_nest_base,pass,7 + + + +lcnet_050,pass,6 + + + +levit_128,pass,7 + + + +mixer_b16_224,pass,7 + + + +mixnet_l,pass,6 + + + +mnasnet_100,pass,7 + + + +mobilenetv2_100,pass,7 + + + +mobilenetv3_large_100,pass,7 + + + +mobilevit_s,pass,6 + + + +nfnet_l0,pass,7 + + + +pit_b_224,pass,6 + + + +pnasnet5large,pass,5 + + + +poolformer_m36,pass,6 + + + +regnety_002,pass,6 + + + +repvgg_a2,pass,7 + + + +res2net101_26w_4s,pass,6 + + + +res2net50_14w_8s,pass,6 + + + +res2next50,pass,6 + + + +resmlp_12_224,pass,6 + + + +resnest101e,pass,6 + + + +rexnet_100,pass,7 + + + +sebotnet33ts_256,pass,6 + + + +selecsls42b,pass,6 + + + +spnasnet_100,pass,7 + + + +swin_base_patch4_window7_224,pass,7 + + + +swsl_resnext101_32x16d,pass,6 + + + +tf_efficientnet_b0,pass,6 + + + +tf_mixnet_l,pass,6 + + + +tinynet_a,pass,6 + + + +tnt_s_patch16_224,pass,7 + + + +twins_pcpvt_base,pass,7 + + + +visformer_small,pass,7 + + + +vit_base_patch16_224,pass,7 + + + +volo_d1_224,pass,7 + + + +xcit_large_24_p8_224,pass_due_to_skip,7 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_torchbench_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_torchbench_inference.csv new file mode 100644 index 00000000000..4a379642b49 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_torchbench_inference.csv @@ -0,0 +1,373 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,eager_fail_to_run,0 + + + +BERT_pytorch,pass,0 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,0 + + + +Super_SloMo,pass,0 + + + +alexnet,pass,0 + + + +basic_gnn_edgecnn,pass,0 + + + +basic_gnn_gcn,pass,6 + + + +basic_gnn_gin,pass,0 + + + +basic_gnn_sage,pass,0 + + + +cm3leon_generate,pass,4 + + + +dcgan,pass,0 + + + +demucs,pass,3 + + + +densenet121,pass,0 + + + +detectron2_fasterrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_fpn,eager_fail_to_run,0 + + + +detectron2_fcos_r_50_fpn,pass,22 + + + +detectron2_maskrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_fpn,eager_fail_to_run,0 + + + +dlrm,pass,0 + + + +doctr_det_predictor,eager_fail_to_run,5 + + + +doctr_reco_predictor,eager_fail_to_run,4 + + + +drq,pass,0 + + + +fastNLP_Bert,pass,4 + + + +functorch_dp_cifar10,pass,0 + + + +functorch_maml_omniglot,pass,0 + + + +hf_Albert,pass,0 + + + +hf_Bart,pass,0 + + + +hf_Bert,pass,0 + + + +hf_Bert_large,pass,0 + + + +hf_BigBird,fail_to_run,0 + + + +hf_DistilBert,pass,0 + + + +hf_GPT2,pass,0 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_Reformer,pass,5 + + + +hf_T5,pass,0 + + + +hf_T5_base,eager_fail_to_run,0 + + + +hf_T5_generate,pass,5 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,0 + + + +hf_distil_whisper,pass,0 + + + +lennard_jones,pass,0 + + + +llama,pass,0 + + + +llama_v2_7b_16h,model_fail_to_load,0 + + + +llava,model_fail_to_load,0 + + + +maml,pass_due_to_skip,0 + + + +maml_omniglot,pass,0 + + + +mnasnet1_0,pass,0 + + + +mobilenet_v2,pass,0 + + + +mobilenet_v2_quantized_qat,model_fail_to_load,0 + + + +mobilenet_v3_large,pass,0 + + + +moco,pass,7 + + + +moondream,model_fail_to_load,0 + + + +nanogpt,pass,0 + + + +nvidia_deeprecommender,pass,0 + + + +opacus_cifar10,pass,0 + + + +phlippe_densenet,pass,0 + + + +phlippe_resnet,pass,0 + + + +pyhpc_equation_of_state,pass,0 + + + +pyhpc_isoneutral_mixing,pass,0 + + + +pyhpc_turbulent_kinetic_energy,pass,0 + + + +pytorch_CycleGAN_and_pix2pix,pass,0 + + + +pytorch_stargan,pass,0 + + + +pytorch_unet,pass,0 + + + +resnet152,pass,0 + + + +resnet18,pass,0 + + + +resnet50,pass,0 + + + +resnet50_quantized_qat,model_fail_to_load,0 + + + +resnext50_32x4d,pass,0 + + + +sam,pass,0 + + + +shufflenet_v2_x1_0,pass,0 + + + +soft_actor_critic,pass,0 + + + +speech_transformer,pass,10 + + + +squeezenet1_1,pass,0 + + + +stable_diffusion_text_encoder,pass,0 + + + +stable_diffusion_unet,pass_due_to_skip,0 + + + +timm_efficientnet,pass,0 + + + +timm_regnet,pass,0 + + + +timm_resnest,pass,0 + + + +timm_vision_transformer,pass,0 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,0 + + + +torch_multimodal_clip,pass,0 + + + +tts_angular,pass,2 + + + +vgg16,pass,0 + + + +vision_maskrcnn,pass,18 + + + +yolov3,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_torchbench_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_torchbench_training.csv new file mode 100644 index 00000000000..14b44da8bbc --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamic_inductor_torchbench_training.csv @@ -0,0 +1,281 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,fail_to_run,3 + + + +BERT_pytorch,pass,6 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,6 + + + +Super_SloMo,pass,7 + + + +alexnet,pass,6 + + + +basic_gnn_edgecnn,pass,20 + + + +basic_gnn_gcn,pass,13 + + + +basic_gnn_gin,pass,7 + + + +basic_gnn_sage,pass,7 + + + +dcgan,pass,6 + + + +demucs,fail_to_run,4 + + + +densenet121,pass,6 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +dlrm,pass,6 + + + +drq,pass,7 + + + +fastNLP_Bert,pass,10 + + + +functorch_dp_cifar10,pass,7 + + + +functorch_maml_omniglot,pass,7 + + + +hf_Albert,pass,6 + + + +hf_Bart,pass,6 + + + +hf_Bert,pass,6 + + + +hf_Bert_large,pass,6 + + + +hf_BigBird,fail_to_run,3 + + + +hf_DistilBert,pass,6 + + + +hf_GPT2,pass,6 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_Reformer,pass,23 + + + +hf_T5_base,eager_2nd_run_OOM,0 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,6 + + + +hf_distil_whisper,model_fail_to_load,0 + + + +lennard_jones,pass,7 + + + +llava,model_fail_to_load,0 + + + +maml_omniglot,pass,7 + + + +mnasnet1_0,pass,7 + + + +mobilenet_v2,pass,6 + + + +mobilenet_v2_quantized_qat,eager_fail_to_run,0 + + + +mobilenet_v3_large,pass,7 + + + +moco,pass,13 + + + +nanogpt,pass,7 + + + +nvidia_deeprecommender,pass,7 + + + +opacus_cifar10,eager_fail_to_run,0 + + + +phlippe_densenet,pass,6 + + + +phlippe_resnet,pass,6 + + + +pytorch_CycleGAN_and_pix2pix,pass,6 + + + +pytorch_stargan,pass,6 + + + +pytorch_unet,pass_due_to_skip,7 + + + +resnet152,pass,7 + + + +resnet18,pass,6 + + + +resnet50,pass,6 + + + +resnet50_quantized_qat,eager_fail_to_run,0 + + + +resnext50_32x4d,pass,7 + + + +sam,eager_fail_to_run,0 + + + +shufflenet_v2_x1_0,pass,6 + + + +soft_actor_critic,pass,6 + + + +squeezenet1_1,pass,6 + + + +stable_diffusion_text_encoder,pass,5 + + + +stable_diffusion_unet,pass_due_to_skip,0 + + + +timm_efficientnet,pass,7 + + + +timm_regnet,pass,6 + + + +timm_resnest,pass,7 + + + +timm_vision_transformer,pass,6 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,6 + + + +torch_multimodal_clip,pass,7 + + + +tts_angular,pass,9 + + + +vgg16,pass,6 + + + +vision_maskrcnn,pass,35 + + + +yolov3,pass,8 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_huggingface_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_huggingface_inference.csv new file mode 100644 index 00000000000..fd57a3b4cbf --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_huggingface_inference.csv @@ -0,0 +1,185 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,0 + + + +AlbertForQuestionAnswering,pass,0 + + + +AllenaiLongformerBase,pass,4 + + + +BartForCausalLM,pass,0 + + + +BartForConditionalGeneration,pass,0 + + + +BertForMaskedLM,pass,0 + + + +BertForQuestionAnswering,pass,0 + + + +BlenderbotForCausalLM,pass_due_to_skip,0 + + + +BlenderbotSmallForCausalLM,pass,0 + + + +BlenderbotSmallForConditionalGeneration,pass,0 + + + +CamemBert,pass,0 + + + +DebertaForMaskedLM,pass,0 + + + +DebertaForQuestionAnswering,pass,0 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,pass,0 + + + +DistilBertForMaskedLM,pass,0 + + + +DistilBertForQuestionAnswering,pass,0 + + + +DistillGPT2,pass,0 + + + +ElectraForCausalLM,pass,0 + + + +ElectraForQuestionAnswering,pass,0 + + + +GPT2ForSequenceClassification,pass,0 + + + +GoogleFnet,pass,0 + + + +LayoutLMForMaskedLM,pass,0 + + + +LayoutLMForSequenceClassification,pass,0 + + + +M2M100ForConditionalGeneration,pass,0 + + + +MBartForCausalLM,pass,0 + + + +MBartForConditionalGeneration,pass,0 + + + +MT5ForConditionalGeneration,pass,0 + + + +MegatronBertForCausalLM,pass,0 + + + +MegatronBertForQuestionAnswering,pass,0 + + + +MobileBertForMaskedLM,pass,0 + + + +MobileBertForQuestionAnswering,pass,0 + + + +OPTForCausalLM,pass,0 + + + +PLBartForCausalLM,pass,0 + + + +PLBartForConditionalGeneration,pass,0 + + + +PegasusForCausalLM,pass,0 + + + +PegasusForConditionalGeneration,pass,0 + + + +RobertaForCausalLM,pass,0 + + + +RobertaForQuestionAnswering,pass,0 + + + +Speech2Text2ForCausalLM,pass,0 + + + +T5ForConditionalGeneration,pass,0 + + + +T5Small,pass,0 + + + +TrOCRForCausalLM,pass,0 + + + +XGLMForCausalLM,pass,0 + + + +XLNetLMHeadModel,pass,0 + + + +YituTechConvBert,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_huggingface_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_huggingface_training.csv new file mode 100644 index 00000000000..b54c6a84bc2 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_huggingface_training.csv @@ -0,0 +1,185 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,4 + + + +AlbertForQuestionAnswering,pass,5 + + + +AllenaiLongformerBase,pass,9 + + + +BartForCausalLM,pass,6 + + + +BartForConditionalGeneration,pass,8 + + + +BertForMaskedLM,pass,5 + + + +BertForQuestionAnswering,pass,5 + + + +BlenderbotForCausalLM,pass_due_to_skip,0 + + + +BlenderbotSmallForCausalLM,pass,6 + + + +BlenderbotSmallForConditionalGeneration,pass,8 + + + +CamemBert,pass,5 + + + +DebertaForMaskedLM,pass,5 + + + +DebertaForQuestionAnswering,pass,5 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,pass,4 + + + +DistilBertForMaskedLM,pass,5 + + + +DistilBertForQuestionAnswering,pass,5 + + + +DistillGPT2,pass,5 + + + +ElectraForCausalLM,pass,4 + + + +ElectraForQuestionAnswering,pass,5 + + + +GPT2ForSequenceClassification,pass,5 + + + +GoogleFnet,pass,5 + + + +LayoutLMForMaskedLM,pass,5 + + + +LayoutLMForSequenceClassification,pass,5 + + + +M2M100ForConditionalGeneration,pass,4 + + + +MBartForCausalLM,pass,6 + + + +MBartForConditionalGeneration,pass,8 + + + +MT5ForConditionalGeneration,pass,5 + + + +MegatronBertForCausalLM,pass,5 + + + +MegatronBertForQuestionAnswering,pass,5 + + + +MobileBertForMaskedLM,pass,3 + + + +MobileBertForQuestionAnswering,pass,3 + + + +OPTForCausalLM,pass,6 + + + +PLBartForCausalLM,pass,6 + + + +PLBartForConditionalGeneration,pass,8 + + + +PegasusForCausalLM,pass,6 + + + +PegasusForConditionalGeneration,pass,7 + + + +RobertaForCausalLM,pass,5 + + + +RobertaForQuestionAnswering,pass,5 + + + +Speech2Text2ForCausalLM,pass,6 + + + +T5ForConditionalGeneration,pass,5 + + + +T5Small,pass,5 + + + +TrOCRForCausalLM,pass,6 + + + +XGLMForCausalLM,pass,6 + + + +XLNetLMHeadModel,pass,5 + + + +YituTechConvBert,pass,5 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_timm_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_timm_inference.csv new file mode 100644 index 00000000000..6727624d0b0 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_timm_inference.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,0 + + + +beit_base_patch16_224,pass,0 + + + +botnet26t_256,pass,0 + + + +cait_m36_384,pass,0 + + + +coat_lite_mini,pass,0 + + + +convit_base,pass,0 + + + +convmixer_768_32,pass,0 + + + +convnext_base,pass,0 + + + +crossvit_9_240,pass,0 + + + +cspdarknet53,pass,0 + + + +deit_base_distilled_patch16_224,pass,0 + + + +dla102,pass,0 + + + +dm_nfnet_f0,pass,0 + + + +dpn107,pass,0 + + + +eca_botnext26ts_256,pass,0 + + + +eca_halonext26ts,pass,0 + + + +ese_vovnet19b_dw,pass,0 + + + +fbnetc_100,pass,0 + + + +fbnetv3_b,pass,0 + + + +gernet_l,pass,0 + + + +ghostnet_100,pass,0 + + + +gluon_inception_v3,pass,0 + + + +gmixer_24_224,pass,0 + + + +gmlp_s16_224,pass,0 + + + +hrnet_w18,pass,0 + + + +inception_v3,pass,0 + + + +jx_nest_base,pass,0 + + + +lcnet_050,pass,0 + + + +levit_128,eager_fail_to_run,0 + + + +mixer_b16_224,pass,0 + + + +mixnet_l,pass,0 + + + +mnasnet_100,pass,0 + + + +mobilenetv2_100,pass,0 + + + +mobilenetv3_large_100,pass,0 + + + +mobilevit_s,pass,0 + + + +nfnet_l0,pass,0 + + + +pit_b_224,pass,0 + + + +pnasnet5large,pass,0 + + + +poolformer_m36,pass,0 + + + +regnety_002,pass,0 + + + +repvgg_a2,pass,0 + + + +res2net101_26w_4s,pass,0 + + + +res2net50_14w_8s,pass,0 + + + +res2next50,pass,0 + + + +resmlp_12_224,pass,0 + + + +resnest101e,pass,0 + + + +rexnet_100,pass,0 + + + +sebotnet33ts_256,pass,0 + + + +selecsls42b,pass,0 + + + +spnasnet_100,pass,0 + + + +swin_base_patch4_window7_224,pass,0 + + + +swsl_resnext101_32x16d,pass,0 + + + +tf_efficientnet_b0,pass,0 + + + +tf_mixnet_l,pass,0 + + + +tinynet_a,pass,0 + + + +tnt_s_patch16_224,pass,0 + + + +twins_pcpvt_base,pass,0 + + + +visformer_small,pass,0 + + + +vit_base_patch16_224,pass,0 + + + +volo_d1_224,pass,0 + + + +xcit_large_24_p8_224,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_timm_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_timm_training.csv new file mode 100644 index 00000000000..afd846df7cb --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_timm_training.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,6 + + + +beit_base_patch16_224,pass,7 + + + +botnet26t_256,pass,6 + + + +cait_m36_384,pass,6 + + + +coat_lite_mini,pass,6 + + + +convit_base,pass,7 + + + +convmixer_768_32,pass,5 + + + +convnext_base,pass,7 + + + +crossvit_9_240,pass,7 + + + +cspdarknet53,pass,7 + + + +deit_base_distilled_patch16_224,pass,7 + + + +dla102,pass,7 + + + +dm_nfnet_f0,pass,6 + + + +dpn107,pass,6 + + + +eca_botnext26ts_256,pass,7 + + + +eca_halonext26ts,pass,7 + + + +ese_vovnet19b_dw,pass,7 + + + +fbnetc_100,pass,7 + + + +fbnetv3_b,pass,6 + + + +gernet_l,pass,6 + + + +ghostnet_100,pass,6 + + + +gluon_inception_v3,pass,7 + + + +gmixer_24_224,pass,6 + + + +gmlp_s16_224,pass,7 + + + +hrnet_w18,pass,5 + + + +inception_v3,pass,6 + + + +jx_nest_base,pass,7 + + + +lcnet_050,pass,6 + + + +levit_128,eager_fail_to_run,7 + + + +mixer_b16_224,pass,7 + + + +mixnet_l,pass,6 + + + +mnasnet_100,pass,7 + + + +mobilenetv2_100,pass,7 + + + +mobilenetv3_large_100,pass,7 + + + +mobilevit_s,pass,6 + + + +nfnet_l0,pass,7 + + + +pit_b_224,pass,6 + + + +pnasnet5large,pass,5 + + + +poolformer_m36,pass,6 + + + +regnety_002,pass,6 + + + +repvgg_a2,pass,7 + + + +res2net101_26w_4s,pass,6 + + + +res2net50_14w_8s,pass,6 + + + +res2next50,pass,6 + + + +resmlp_12_224,pass,6 + + + +resnest101e,pass,6 + + + +rexnet_100,pass,7 + + + +sebotnet33ts_256,pass,6 + + + +selecsls42b,pass,6 + + + +spnasnet_100,pass,7 + + + +swin_base_patch4_window7_224,pass,7 + + + +swsl_resnext101_32x16d,pass,6 + + + +tf_efficientnet_b0,pass,6 + + + +tf_mixnet_l,pass,6 + + + +tinynet_a,pass,6 + + + +tnt_s_patch16_224,pass,7 + + + +twins_pcpvt_base,pass,7 + + + +visformer_small,pass,7 + + + +vit_base_patch16_224,pass,7 + + + +volo_d1_224,pass,7 + + + +xcit_large_24_p8_224,pass_due_to_skip,7 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_torchbench_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_torchbench_inference.csv new file mode 100644 index 00000000000..148e0331849 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_torchbench_inference.csv @@ -0,0 +1,396 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,eager_fail_to_run,0 + + + +BERT_pytorch,pass,0 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,0 + + + +Super_SloMo,pass,0 + + + +alexnet,pass,0 + + + +basic_gnn_edgecnn,pass,0 + + + +basic_gnn_gcn,pass,6 + + + +basic_gnn_gin,pass,0 + + + +basic_gnn_sage,pass,0 + + + +cm3leon_generate,pass,4 + + + +dcgan,pass,0 + + + +demucs,pass,3 + + + +densenet121,pass,0 + + + +detectron2_fasterrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_fpn,eager_fail_to_run,0 + + + +detectron2_fcos_r_50_fpn,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_fpn,eager_fail_to_run,0 + + + +dlrm,pass,0 + + + +doctr_det_predictor,eager_fail_to_run,5 + + + +doctr_reco_predictor,eager_fail_to_run,4 + + + +drq,pass,0 + + + +fastNLP_Bert,pass,4 + + + +functorch_dp_cifar10,pass,0 + + + +functorch_maml_omniglot,pass,0 + + + +hf_Albert,pass,0 + + + +hf_Bart,pass,0 + + + +hf_Bert,pass,0 + + + +hf_Bert_large,pass,0 + + + +hf_BigBird,pass,9 + + + +hf_DistilBert,pass,0 + + + +hf_GPT2,pass,0 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_Longformer,pass,4 + + + +hf_Reformer,pass,5 + + + +hf_Roberta_base,pass,0 + + +hf_T5,pass,0 + + + +hf_T5_base,pass,0 + + + +hf_T5_generate,pass,5 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,0 + + + +hf_distil_whisper,pass,0 + + + +lennard_jones,pass,0 + + + +llama,pass,0 + + + +llama_v2_7b_16h,model_fail_to_load,0 + + + +llava,pass,0 + + + +maml,pass_due_to_skip,0 + + + +maml_omniglot,pass,0 + + + +microbench_unbacked_tolist_sum,pass,1 + + + +mnasnet1_0,pass,0 + + + +mobilenet_v2,pass,0 + + + +mobilenet_v2_quantized_qat,model_fail_to_load,0 + + + +mobilenet_v3_large,pass,0 + + + +moco,pass,7 + + + +moondream,pass,0 + + + +nanogpt,pass,0 + + + +nvidia_deeprecommender,pass,0 + + + +opacus_cifar10,pass,0 + + + +phlippe_densenet,pass,0 + + + +phlippe_resnet,pass,0 + + + +pyhpc_equation_of_state,pass,0 + + + +pyhpc_isoneutral_mixing,pass,0 + + + +pyhpc_turbulent_kinetic_energy,pass,0 + + + +pytorch_CycleGAN_and_pix2pix,pass,0 + + + +pytorch_stargan,pass,0 + + + +pytorch_unet,pass,0 + + + +resnet152,pass,0 + + + +resnet18,pass,0 + + + +resnet50,pass,0 + + + +resnet50_quantized_qat,model_fail_to_load,0 + + + +resnext50_32x4d,pass,0 + + + +sam,pass,0 + + + +sam_fast,pass,0 + + + +shufflenet_v2_x1_0,pass,0 + + + +soft_actor_critic,pass,0 + + + +speech_transformer,pass,10 + + + +squeezenet1_1,pass,0 + + + +stable_diffusion_text_encoder,model_fail_to_load,0 + + + +stable_diffusion_unet,model_fail_to_load,0 + + + +timm_efficientdet,pass,2 + + + +timm_efficientnet,pass,0 + + + +timm_nfnet,pass,0 + + + +timm_regnet,pass,0 + + + +timm_resnest,pass,0 + + + +timm_vision_transformer,pass,0 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,0 + + + +torch_multimodal_clip,pass,0 + + + +tts_angular,pass,2 + + + +vgg16,eager_two_runs_differ,0 + + + +vision_maskrcnn,pass,18 + + + +yolov3,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_torchbench_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_torchbench_training.csv new file mode 100644 index 00000000000..44064a074ba --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/dynamo_eager_torchbench_training.csv @@ -0,0 +1,305 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,pass,6 + + + +BERT_pytorch,pass,6 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,6 + + + +Super_SloMo,pass,7 + + + +alexnet,pass,0 + + + +basic_gnn_edgecnn,pass,20 + + + +basic_gnn_gcn,pass,13 + + + +basic_gnn_gin,pass,7 + + + +basic_gnn_sage,pass,7 + + + +dcgan,pass,6 + + + +demucs,pass,9 + + + +densenet121,pass,6 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +dlrm,pass,6 + + + +drq,pass,7 + + + +fastNLP_Bert,pass,10 + + + +functorch_dp_cifar10,pass,7 + + + +functorch_maml_omniglot,pass,7 + + + +hf_Albert,pass,6 + + + +hf_Bart,pass,6 + + + +hf_Bert,pass,6 + + + +hf_Bert_large,pass,6 + + + +hf_BigBird,pass,15 + + + +hf_DistilBert,pass,6 + + + +hf_GPT2,pass,6 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_Longformer,pass,4 + + + +hf_Reformer,pass,23 + + + +hf_Roberta_base,pass,6 + + + +hf_T5,pass,0 + + + +hf_T5_base,eager_2nd_run_OOM,0 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,6 + + + +hf_distil_whisper,model_fail_to_load,0 + + + +lennard_jones,pass,7 + + + +llava,model_fail_to_load,0 + + + +maml_omniglot,pass,7 + + + +mnasnet1_0,pass,7 + + + +mobilenet_v2,pass,6 + + + +mobilenet_v2_quantized_qat,eager_fail_to_run,0 + + + +mobilenet_v3_large,pass,7 + + + +moco,pass,13 + + + +nanogpt,pass,7 + + + +nvidia_deeprecommender,pass,7 + + + +opacus_cifar10,eager_fail_to_run,0 + + + +phlippe_densenet,pass,6 + + + +phlippe_resnet,pass,6 + + + +pytorch_CycleGAN_and_pix2pix,pass,6 + + + +pytorch_stargan,pass,6 + + + +pytorch_unet,pass_due_to_skip,7 + + + +resnet152,pass,7 + + + +resnet18,pass,6 + + + +resnet50,pass,6 + + + +resnet50_quantized_qat,eager_fail_to_run,0 + + + +resnext50_32x4d,pass,7 + + + +sam,eager_fail_to_run,0 + + + +shufflenet_v2_x1_0,pass,6 + + + +soft_actor_critic,pass,6 + + + +speech_transformer,pass,16 + + + +squeezenet1_1,pass,6 + + + +stable_diffusion_text_encoder,pass,0 + + + +stable_diffusion_unet,pass_due_to_skip,0 + + + +timm_efficientdet,pass,2 + + + +timm_efficientnet,pass,7 + + + +timm_nfnet,pass,0 + + + +timm_regnet,pass,6 + + + +timm_resnest,pass,7 + + + +timm_vision_transformer,pass,6 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,6 + + + +torch_multimodal_clip,pass,7 + + + +tts_angular,pass,9 + + + +vgg16,pass,6 + + + +vision_maskrcnn,pass,35 + + + +yolov3,pass,8 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_huggingface_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_huggingface_inference.csv new file mode 100644 index 00000000000..fd57a3b4cbf --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_huggingface_inference.csv @@ -0,0 +1,185 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,0 + + + +AlbertForQuestionAnswering,pass,0 + + + +AllenaiLongformerBase,pass,4 + + + +BartForCausalLM,pass,0 + + + +BartForConditionalGeneration,pass,0 + + + +BertForMaskedLM,pass,0 + + + +BertForQuestionAnswering,pass,0 + + + +BlenderbotForCausalLM,pass_due_to_skip,0 + + + +BlenderbotSmallForCausalLM,pass,0 + + + +BlenderbotSmallForConditionalGeneration,pass,0 + + + +CamemBert,pass,0 + + + +DebertaForMaskedLM,pass,0 + + + +DebertaForQuestionAnswering,pass,0 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,pass,0 + + + +DistilBertForMaskedLM,pass,0 + + + +DistilBertForQuestionAnswering,pass,0 + + + +DistillGPT2,pass,0 + + + +ElectraForCausalLM,pass,0 + + + +ElectraForQuestionAnswering,pass,0 + + + +GPT2ForSequenceClassification,pass,0 + + + +GoogleFnet,pass,0 + + + +LayoutLMForMaskedLM,pass,0 + + + +LayoutLMForSequenceClassification,pass,0 + + + +M2M100ForConditionalGeneration,pass,0 + + + +MBartForCausalLM,pass,0 + + + +MBartForConditionalGeneration,pass,0 + + + +MT5ForConditionalGeneration,pass,0 + + + +MegatronBertForCausalLM,pass,0 + + + +MegatronBertForQuestionAnswering,pass,0 + + + +MobileBertForMaskedLM,pass,0 + + + +MobileBertForQuestionAnswering,pass,0 + + + +OPTForCausalLM,pass,0 + + + +PLBartForCausalLM,pass,0 + + + +PLBartForConditionalGeneration,pass,0 + + + +PegasusForCausalLM,pass,0 + + + +PegasusForConditionalGeneration,pass,0 + + + +RobertaForCausalLM,pass,0 + + + +RobertaForQuestionAnswering,pass,0 + + + +Speech2Text2ForCausalLM,pass,0 + + + +T5ForConditionalGeneration,pass,0 + + + +T5Small,pass,0 + + + +TrOCRForCausalLM,pass,0 + + + +XGLMForCausalLM,pass,0 + + + +XLNetLMHeadModel,pass,0 + + + +YituTechConvBert,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_huggingface_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_huggingface_training.csv new file mode 100644 index 00000000000..8202281ed9b --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_huggingface_training.csv @@ -0,0 +1,185 @@ +name,accuracy,graph_breaks + + + +AlbertForMaskedLM,pass,4 + + + +AlbertForQuestionAnswering,pass,5 + + + +AllenaiLongformerBase,pass,9 + + + +BartForCausalLM,pass,6 + + + +BartForConditionalGeneration,pass,8 + + + +BertForMaskedLM,pass,5 + + + +BertForQuestionAnswering,pass,5 + + + +BlenderbotForCausalLM,eager_fail_to_run,0 + + + +BlenderbotSmallForCausalLM,pass,6 + + + +BlenderbotSmallForConditionalGeneration,pass,8 + + + +CamemBert,pass,5 + + + +DebertaForMaskedLM,pass,5 + + + +DebertaForQuestionAnswering,pass,5 + + + +DebertaV2ForMaskedLM,pass_due_to_skip,0 + + + +DebertaV2ForQuestionAnswering,eager_1st_run_OOM,0 + + + +DistilBertForMaskedLM,pass,5 + + + +DistilBertForQuestionAnswering,pass,5 + + + +DistillGPT2,pass,5 + + + +ElectraForCausalLM,pass,4 + + + +ElectraForQuestionAnswering,pass,5 + + + +GPT2ForSequenceClassification,pass,5 + + + +GoogleFnet,pass,5 + + + +LayoutLMForMaskedLM,pass,5 + + + +LayoutLMForSequenceClassification,pass,5 + + + +M2M100ForConditionalGeneration,pass,4 + + + +MBartForCausalLM,pass,6 + + + +MBartForConditionalGeneration,pass,8 + + + +MT5ForConditionalGeneration,pass,5 + + + +MegatronBertForCausalLM,pass,5 + + + +MegatronBertForQuestionAnswering,pass,5 + + + +MobileBertForMaskedLM,pass,3 + + + +MobileBertForQuestionAnswering,pass,3 + + + +OPTForCausalLM,pass,6 + + + +PLBartForCausalLM,pass,6 + + + +PLBartForConditionalGeneration,pass,8 + + + +PegasusForCausalLM,pass,6 + + + +PegasusForConditionalGeneration,pass,7 + + + +RobertaForCausalLM,pass,5 + + + +RobertaForQuestionAnswering,pass,5 + + + +Speech2Text2ForCausalLM,pass,6 + + + +T5ForConditionalGeneration,pass,5 + + + +T5Small,pass,5 + + + +TrOCRForCausalLM,pass,6 + + + +XGLMForCausalLM,pass,6 + + + +XLNetLMHeadModel,pass,5 + + + +YituTechConvBert,pass,5 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_timm_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_timm_inference.csv new file mode 100644 index 00000000000..c889ba0e8d2 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_timm_inference.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,0 + + + +beit_base_patch16_224,pass,0 + + + +botnet26t_256,pass,0 + + + +cait_m36_384,pass,0 + + + +coat_lite_mini,pass,0 + + + +convit_base,pass,0 + + + +convmixer_768_32,pass,0 + + + +convnext_base,pass,0 + + + +crossvit_9_240,pass,0 + + + +cspdarknet53,pass,0 + + + +deit_base_distilled_patch16_224,pass,0 + + + +dla102,pass,0 + + + +dm_nfnet_f0,pass,0 + + + +dpn107,pass,0 + + + +eca_botnext26ts_256,pass,0 + + + +eca_halonext26ts,pass,0 + + + +ese_vovnet19b_dw,pass,0 + + + +fbnetc_100,pass,0 + + + +fbnetv3_b,pass,0 + + + +gernet_l,pass,0 + + + +ghostnet_100,pass,0 + + + +gluon_inception_v3,pass,0 + + + +gmixer_24_224,pass,0 + + + +gmlp_s16_224,pass,0 + + + +hrnet_w18,pass,0 + + + +inception_v3,pass,0 + + + +jx_nest_base,pass,0 + + + +lcnet_050,pass,0 + + + +levit_128,pass,0 + + + +mixer_b16_224,pass,0 + + + +mixnet_l,pass,0 + + + +mnasnet_100,pass,0 + + + +mobilenetv2_100,pass,0 + + + +mobilenetv3_large_100,pass,0 + + + +mobilevit_s,pass,0 + + + +nfnet_l0,pass,0 + + + +pit_b_224,pass,0 + + + +pnasnet5large,pass,0 + + + +poolformer_m36,pass,0 + + + +regnety_002,pass,0 + + + +repvgg_a2,pass,0 + + + +res2net101_26w_4s,pass,0 + + + +res2net50_14w_8s,pass,0 + + + +res2next50,pass,0 + + + +resmlp_12_224,pass,0 + + + +resnest101e,pass,0 + + + +rexnet_100,pass,0 + + + +sebotnet33ts_256,pass,0 + + + +selecsls42b,pass,0 + + + +spnasnet_100,pass,0 + + + +swin_base_patch4_window7_224,pass,0 + + + +swsl_resnext101_32x16d,pass,0 + + + +tf_efficientnet_b0,pass,0 + + + +tf_mixnet_l,pass,0 + + + +tinynet_a,pass,0 + + + +tnt_s_patch16_224,pass,0 + + + +twins_pcpvt_base,pass,0 + + + +visformer_small,pass,0 + + + +vit_base_patch16_224,pass,0 + + + +volo_d1_224,pass,0 + + + +xcit_large_24_p8_224,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_timm_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_timm_training.csv new file mode 100644 index 00000000000..e5464160d32 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_timm_training.csv @@ -0,0 +1,245 @@ +name,accuracy,graph_breaks + + + +adv_inception_v3,pass,6 + + + +beit_base_patch16_224,pass,7 + + + +botnet26t_256,pass,6 + + + +cait_m36_384,eager_fail_to_run,0 + + + +coat_lite_mini,pass,6 + + + +convit_base,pass,7 + + + +convmixer_768_32,pass,5 + + + +convnext_base,pass,7 + + + +crossvit_9_240,pass,7 + + + +cspdarknet53,pass,7 + + + +deit_base_distilled_patch16_224,pass,7 + + + +dla102,pass,7 + + + +dm_nfnet_f0,pass,6 + + + +dpn107,pass,6 + + + +eca_botnext26ts_256,pass,7 + + + +eca_halonext26ts,pass,7 + + + +ese_vovnet19b_dw,pass,7 + + + +fbnetc_100,pass,7 + + + +fbnetv3_b,pass,6 + + + +gernet_l,pass,6 + + + +ghostnet_100,pass,6 + + + +gluon_inception_v3,pass,7 + + + +gmixer_24_224,pass,6 + + + +gmlp_s16_224,pass,7 + + + +hrnet_w18,pass,5 + + + +inception_v3,pass,6 + + + +jx_nest_base,pass,7 + + + +lcnet_050,pass,6 + + + +levit_128,pass,7 + + + +mixer_b16_224,pass,7 + + + +mixnet_l,pass,6 + + + +mnasnet_100,pass,7 + + + +mobilenetv2_100,pass,7 + + + +mobilenetv3_large_100,pass,7 + + + +mobilevit_s,pass,6 + + + +nfnet_l0,pass,7 + + + +pit_b_224,pass,6 + + + +pnasnet5large,pass,5 + + + +poolformer_m36,pass,6 + + + +regnety_002,pass,6 + + + +repvgg_a2,pass,7 + + + +res2net101_26w_4s,pass,6 + + + +res2net50_14w_8s,pass,6 + + + +res2next50,pass,6 + + + +resmlp_12_224,pass,6 + + + +resnest101e,pass,6 + + + +rexnet_100,pass,7 + + + +sebotnet33ts_256,pass,6 + + + +selecsls42b,pass,6 + + + +spnasnet_100,pass,7 + + + +swin_base_patch4_window7_224,pass,7 + + + +swsl_resnext101_32x16d,pass,6 + + + +tf_efficientnet_b0,pass,6 + + + +tf_mixnet_l,pass,6 + + + +tinynet_a,pass,6 + + + +tnt_s_patch16_224,pass,7 + + + +twins_pcpvt_base,pass,7 + + + +visformer_small,pass,7 + + + +vit_base_patch16_224,pass,7 + + + +volo_d1_224,pass,7 + + + +xcit_large_24_p8_224,pass_due_to_skip,7 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_torchbench_inference.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_torchbench_inference.csv new file mode 100644 index 00000000000..996fb77dfb2 --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_torchbench_inference.csv @@ -0,0 +1,377 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,eager_fail_to_run,0 + + + +BERT_pytorch,pass,0 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,0 + + + +Super_SloMo,pass,0 + + + +alexnet,pass,0 + + + +basic_gnn_edgecnn,pass,0 + + + +basic_gnn_gcn,pass,6 + + + +basic_gnn_gin,pass,0 + + + +basic_gnn_sage,pass,0 + + + +cm3leon_generate,pass,4 + + + +dcgan,pass,0 + + + +demucs,pass,3 + + + +densenet121,pass,0 + + + +detectron2_fasterrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_dc5,eager_fail_to_run,0 + + + +detectron2_fasterrcnn_r_50_fpn,eager_fail_to_run,0 + + + +detectron2_fcos_r_50_fpn,pass,22 + + + +detectron2_maskrcnn_r_101_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_101_fpn,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +detectron2_maskrcnn_r_50_fpn,eager_fail_to_run,0 + + + +dlrm,pass,0 + + + +doctr_det_predictor,eager_fail_to_run,5 + + + +doctr_reco_predictor,eager_fail_to_run,4 + + + +drq,pass,0 + + + +fastNLP_Bert,pass,4 + + + +functorch_dp_cifar10,pass,0 + + + +functorch_maml_omniglot,pass,0 + + + +hf_Albert,pass,0 + + + +hf_Bart,pass,0 + + + +hf_Bert,pass,0 + + + +hf_Bert_large,pass,0 + + + +hf_BigBird,fail_accuracy,0 + + + +hf_DistilBert,pass,0 + + + +hf_GPT2,pass,0 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_Reformer,pass,5 + + + +hf_T5,pass,0 + + + +hf_T5_base,eager_fail_to_run,0 + + + +hf_T5_generate,pass,5 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,0 + + + +hf_distil_whisper,pass,0 + + + +lennard_jones,pass,0 + + + +llama,pass,0 + + + +llama_v2_7b_16h,model_fail_to_load,0 + + + +llava,model_fail_to_load,0 + + + +maml,pass_due_to_skip,0 + + + +maml_omniglot,pass,0 + + + +mnasnet1_0,pass,0 + + + +mobilenet_v2,pass,0 + + + +mobilenet_v2_quantized_qat,model_fail_to_load,0 + + + +mobilenet_v3_large,pass,0 + + + +moco,pass,7 + + + +moondream,model_fail_to_load,0 + + + +nanogpt,pass,0 + + + +nvidia_deeprecommender,pass,0 + + + +opacus_cifar10,pass,0 + + + +phlippe_densenet,pass,0 + + + +phlippe_resnet,pass,0 + + + +pyhpc_equation_of_state,pass,0 + + + +pyhpc_isoneutral_mixing,pass,0 + + + +pyhpc_turbulent_kinetic_energy,pass,0 + + + +pytorch_CycleGAN_and_pix2pix,pass,0 + + + +pytorch_stargan,pass,0 + + + +pytorch_unet,pass,0 + + + +resnet152,pass,0 + + + +resnet18,pass,0 + + + +resnet50,pass,0 + + + +resnet50_quantized_qat,model_fail_to_load,0 + + + +resnext50_32x4d,pass,0 + + + +sam,pass,0 + + + +sam_fast,pass,0 + + + +shufflenet_v2_x1_0,pass,0 + + + +soft_actor_critic,pass,0 + + + +speech_transformer,pass,10 + + + +squeezenet1_1,pass,0 + + + +stable_diffusion_text_encoder,pass,0 + + + +stable_diffusion_unet,pass_due_to_skip,0 + + + +timm_efficientnet,pass,0 + + + +timm_regnet,pass,0 + + + +timm_resnest,pass,0 + + + +timm_vision_transformer,pass,0 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,0 + + + +torch_multimodal_clip,pass,0 + + + +tts_angular,pass,2 + + + +vgg16,pass,0 + + + +vision_maskrcnn,pass,18 + + + +yolov3,pass,0 diff --git a/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_torchbench_training.csv b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_torchbench_training.csv new file mode 100644 index 00000000000..2e5fdd341da --- /dev/null +++ b/benchmarks/dynamo/ci_expected_accuracy/rocm/inductor_torchbench_training.csv @@ -0,0 +1,285 @@ +name,accuracy,graph_breaks + + + +torchrec_dlrm,pass,6 + + + +BERT_pytorch,pass,6 + + + +Background_Matting,pass_due_to_skip,0 + + + +LearningToPaint,pass,6 + + + +Super_SloMo,pass,7 + + + +alexnet,pass,6 + + + +basic_gnn_edgecnn,pass,20 + + + +basic_gnn_gcn,pass,13 + + + +basic_gnn_gin,pass,7 + + + +basic_gnn_sage,pass,7 + + + +dcgan,pass,6 + + + +demucs,pass,9 + + + +densenet121,pass,6 + + + +detectron2_maskrcnn_r_50_c4,eager_fail_to_run,0 + + + +dlrm,pass,6 + + + +drq,pass,7 + + + +fastNLP_Bert,pass,10 + + + +functorch_dp_cifar10,pass,7 + + + +functorch_maml_omniglot,pass,7 + + + +hf_Albert,pass,6 + + + +hf_Bart,pass,6 + + + +hf_Bert,pass,6 + + + +hf_Bert_large,pass,6 + + + +hf_BigBird,pass,6 + + + +hf_DistilBert,pass,6 + + + +hf_GPT2,pass,6 + + + +hf_GPT2_large,pass_due_to_skip,0 + + + +hf_Reformer,pass,23 + + + +hf_T5_base,eager_2nd_run_OOM,0 + + + +hf_T5_large,pass_due_to_skip,0 + + + +hf_Whisper,pass,6 + + + +hf_distil_whisper,model_fail_to_load,0 + + + +lennard_jones,pass,7 + + + +llava,model_fail_to_load,0 + + + +maml_omniglot,pass,7 + + + +mnasnet1_0,pass,7 + + + +mobilenet_v2,pass,6 + + + +mobilenet_v2_quantized_qat,eager_fail_to_run,0 + + + +mobilenet_v3_large,pass,7 + + + +moco,pass,13 + + + +nanogpt,pass,7 + + + +nvidia_deeprecommender,pass,7 + + + +opacus_cifar10,eager_fail_to_run,0 + + + +phlippe_densenet,pass,6 + + + +phlippe_resnet,pass,6 + + + +pytorch_CycleGAN_and_pix2pix,pass,6 + + + +pytorch_stargan,pass,6 + + + +pytorch_unet,pass_due_to_skip,7 + + + +resnet152,pass,7 + + + +resnet18,pass,6 + + + +resnet50,pass,6 + + + +resnet50_quantized_qat,eager_fail_to_run,0 + + + +resnext50_32x4d,pass,7 + + + +sam,eager_fail_to_run,0 + + + +shufflenet_v2_x1_0,pass,6 + + + +soft_actor_critic,pass,6 + + + +speech_transformer,pass,16 + + + +squeezenet1_1,pass,6 + + + +stable_diffusion_text_encoder,pass,5 + + + +stable_diffusion_unet,pass_due_to_skip,0 + + + +timm_efficientnet,pass,7 + + + +timm_regnet,pass,6 + + + +timm_resnest,pass,7 + + + +timm_vision_transformer,pass,6 + + + +timm_vision_transformer_large,pass_due_to_skip,0 + + + +timm_vovnet,pass,6 + + + +torch_multimodal_clip,pass,7 + + + +tts_angular,pass,9 + + + +vgg16,pass,6 + + + +vision_maskrcnn,pass,35 + + + +yolov3,pass,8