Failure to build jaxlib, AMD GPU #25204
Comments
|
python3 ./build/build.py build --wheels=jaxlib,jax-rocm-plugin,jax-rocm-pjrt --rocm_path=/opt/rocm-6.2.4 2024-12-03 00:21:06,886 - INFO - Building jaxlib for linux x86_64... 2024-12-03 00:40:05,229 - INFO - Building jax-rocm-plugin for linux x86_64... I tried again with no --local_xla_path flag. |
|
I already tried bazel clean --expunge |
|
I've been seeing the same issue. Any chance you've found a solution? |
|
Not yet, and give up to use pytorch. |
|
Just adding some investigation findings. Looks like the error comes from TSL adding this flag to its compiler options: https://github.com/google/tsl/pull/2944/files. I double-checked that I'm setting In the meantime, Wintoplay, if you're using an AMD GPU you can get a Docker container that has JAX in it or install it via pip: https://hub.docker.com/r/rocm/jax and https://github.com/ROCm/jax/releases. Curious as to what your use-case is for JAX that you're building from source and not using one of the AMD-provided ways of getting JAX. |
|
clang-18: error: unknown argument: '-fno-canonical-system-headers' So it is the result of my clang package having the name clang-18 and it think I am using gcc? How to work around that? and for why I want to build it, I just want to have the access to the latest version as ROCM build lack behind |
|
Thanks for using JAX on ROCm. We identified a Clang tooling issue in the XLA repository that is causing this problem. We are actively working to get this patch upstream. In the meantime, could you cherry-pick the following commit into your custom XLA repository? Once cherry-picked, you can use the following command (change accordingly) to build JAX + XLA: Please let us know if you encounter any issues or need further assistance. |
Description
I can't build jaxlib.
python3 ./build/build.py build --wheels=jax-rocm-plugin --rocm_version=60 --rocm_path=/opt/rocm --local_xla_path=/media/win/storing/xla
2024-12-03 00:08:13,627 - INFO -
_ _ __ __
| | / \ \ / /
_ | |/ _ \ \ /
| || / ___ /
___// /_/_\
2024-12-03 00:08:13,630 - INFO - Building jax-rocm-plugin for linux x86_64...
2024-12-03 00:08:13,646 - INFO - Bazel options written to .jax_configure.bazelrc
2024-12-03 00:08:13,647 - INFO - [EXECUTING] ./bazel-6.5.0-linux-x86_64 run --repo_env=HERMETIC_PYTHON_VERSION=3.10 --verbose_failures=true --override_repository=xla="/media/win/storing/xla" --action_env=CLANG_COMPILER_PATH="/usr/lib/llvm-14/bin/clang" --repo_env=CC="/usr/lib/llvm-14/bin/clang" --repo_env=BAZEL_COMPILER="/usr/lib/llvm-14/bin/clang" --config=clang --config=mkl_open_source_only --config=avx_posix --config=rocm_base --config=rocm --action_env=CLANG_COMPILER_PATH="/usr/lib/llvm-14/bin/clang" --action_env=ROCM_PATH="/opt/rocm" --action_env=TF_ROCM_AMDGPU_TARGETS=gfx900,gfx906,gfx908,gfx90a,gfx940,gfx941,gfx942,gfx1030,gfx1100 //jaxlib/tools:build_gpu_kernels_wheel -- --output_path="/media/win/storing/jax/dist" --cpu=x86_64 --enable-rocm=True --platform_version=60 --jaxlib_git_hash=aff7714dc0f49cc0097e4db08e028b68182c8ab9
2024-12-03 00:08:13,683 - INFO - WARNING: The following configs were expanded more than once: [rocm_base, clang, mkl_open_source_only, avx_posix, rocm]. For repeatable flags, repeats are counted twice and may lead to unexpected behavior.
2024-12-03 00:08:13,683 - INFO - INFO: Options provided by the client:
2024-12-03 00:08:13,683 - INFO - Inherited 'common' options: --isatty=0 --terminal_columns=80
2024-12-03 00:08:13,683 - INFO - INFO: Reading rc options for 'run' from /media/win/storing/jax/.bazelrc:
2024-12-03 00:08:13,683 - INFO - Inherited 'common' options: --experimental_repo_remote_exec
2024-12-03 00:08:13,683 - INFO - INFO: Reading rc options for 'run' from /media/win/storing/jax/.bazelrc:
2024-12-03 00:08:13,683 - INFO - Inherited 'build' options: --announce_rc --spawn_strategy=local --enable_platform_specific_config --experimental_cc_shared_library --define=grpc_no_ares=true --define=tsl_link_protobuf=true -c opt --output_filter=DONT_MATCH_ANYTHING --copt=-DMLIR_PYTHON_PACKAGE_PREFIX=jaxlib.mlir.
2024-12-03 00:08:13,683 - INFO - INFO: Reading rc options for 'run' from /media/win/storing/jax/.jax_configure.bazelrc:
2024-12-03 00:08:13,683 - INFO - Inherited 'build' options: --repo_env=HERMETIC_PYTHON_VERSION=3.10 --verbose_failures=true --override_repository=xla=/media/win/storing/xla --action_env=CLANG_COMPILER_PATH=/usr/lib/llvm-14/bin/clang --repo_env=CC=/usr/lib/llvm-14/bin/clang --repo_env=BAZEL_COMPILER=/usr/lib/llvm-14/bin/clang --config=clang --config=mkl_open_source_only --config=avx_posix --config=rocm_base --config=rocm --action_env=CLANG_COMPILER_PATH=/usr/lib/llvm-14/bin/clang --action_env=ROCM_PATH=/opt/rocm --action_env=TF_ROCM_AMDGPU_TARGETS=gfx900,gfx906,gfx908,gfx90a,gfx940,gfx941,gfx942,gfx1030,gfx1100
2024-12-03 00:08:13,683 - INFO - INFO: Found applicable config definition build:clang in file /media/win/storing/jax/.bazelrc: --copt=-Wno-gnu-offsetof-extensions --copt=-Qunused-arguments
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:mkl_open_source_only in file /media/win/storing/jax/.bazelrc: --define=tensorflow_mkldnn_contraction_kernel=1
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:avx_posix in file /media/win/storing/jax/.bazelrc: --copt=-mavx --host_copt=-mavx
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:rocm_base in file /media/win/storing/jax/.bazelrc: --crosstool_top=@local_config_rocm//crosstool:toolchain --define=using_rocm=true --define=using_rocm_hipcc=true --repo_env TF_NEED_ROCM=1 --action_env TF_ROCM_AMDGPU_TARGETS=gfx900,gfx906,gfx908,gfx90a,gfx940,gfx941,gfx942,gfx1030,gfx1100
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:rocm in file /media/win/storing/jax/.bazelrc: --config=rocm_base --action_env=TF_ROCM_CLANG=1 --action_env=CLANG_COMPILER_PATH=/usr/lib/llvm-18/bin/clang --copt=-Wno-gnu-offsetof-extensions --copt=-Qunused-arguments --action_env=TF_HIPCC_CLANG=1
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:rocm_base in file /media/win/storing/jax/.bazelrc: --crosstool_top=@local_config_rocm//crosstool:toolchain --define=using_rocm=true --define=using_rocm_hipcc=true --repo_env TF_NEED_ROCM=1 --action_env TF_ROCM_AMDGPU_TARGETS=gfx900,gfx906,gfx908,gfx90a,gfx940,gfx941,gfx942,gfx1030,gfx1100
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:clang in file /media/win/storing/jax/.bazelrc: --copt=-Wno-gnu-offsetof-extensions --copt=-Qunused-arguments
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:mkl_open_source_only in file /media/win/storing/jax/.bazelrc: --define=tensorflow_mkldnn_contraction_kernel=1
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:avx_posix in file /media/win/storing/jax/.bazelrc: --copt=-mavx --host_copt=-mavx
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:rocm_base in file /media/win/storing/jax/.bazelrc: --crosstool_top=@local_config_rocm//crosstool:toolchain --define=using_rocm=true --define=using_rocm_hipcc=true --repo_env TF_NEED_ROCM=1 --action_env TF_ROCM_AMDGPU_TARGETS=gfx900,gfx906,gfx908,gfx90a,gfx940,gfx941,gfx942,gfx1030,gfx1100
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:rocm in file /media/win/storing/jax/.bazelrc: --config=rocm_base --action_env=TF_ROCM_CLANG=1 --action_env=CLANG_COMPILER_PATH=/usr/lib/llvm-18/bin/clang --copt=-Wno-gnu-offsetof-extensions --copt=-Qunused-arguments --action_env=TF_HIPCC_CLANG=1
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:rocm_base in file /media/win/storing/jax/.bazelrc: --crosstool_top=@local_config_rocm//crosstool:toolchain --define=using_rocm=true --define=using_rocm_hipcc=true --repo_env TF_NEED_ROCM=1 --action_env TF_ROCM_AMDGPU_TARGETS=gfx900,gfx906,gfx908,gfx90a,gfx940,gfx941,gfx942,gfx1030,gfx1100
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:linux in file /media/win/storing/jax/.bazelrc: --config=posix --copt=-Wno-unknown-warning-option --copt=-Wno-stringop-truncation --copt=-Wno-array-parameter
2024-12-03 00:08:13,684 - INFO - INFO: Found applicable config definition build:posix in file /media/win/storing/jax/.bazelrc: --copt=-fvisibility=hidden --copt=-Wno-sign-compare --cxxopt=-std=c++17 --host_cxxopt=-std=c++17
2024-12-03 00:08:13,705 - INFO - Loading:
2024-12-03 00:08:13,716 - INFO - ERROR: /media/win/storing/jax/WORKSPACE:38:26: //external:python_dev: no such attribute 'version_variant' in 'custom_python_interpreter' rule
2024-12-03 00:08:13,738 - INFO - ERROR: Error computing the main repository mapping: at /home/win/.cache/bazel/_bazel_win/68b1c5f84d40bc8954a3200dd8f20909/external/xla/workspace3.bzl:4:6: Encountered error while reading extension file 'workspace3.bzl': no such package '@tsl//': error loading package 'external': Could not load //external package
Traceback (most recent call last):
File "/media/win/storing/jax/./build/build.py", line 610, in
asyncio.run(main())
File "/usr/lib/python3.10/asyncio/runners.py", line 44, in run
return loop.run_until_complete(main)
File "/usr/lib/python3.10/asyncio/base_events.py", line 649, in run_until_complete
return future.result()
File "/media/win/storing/jax/./build/build.py", line 603, in main
raise RuntimeError(f"Command failed with return code {result.return_code}")
RuntimeError: Command failed with return code 1
System info (python version, jaxlib version, accelerator, etc.)
Rocm: 6.2.4
attempting to build from source
https://github.com/jax-ml/jax
0.4.35
command python3 ./build/build.py build --wheels=jax-rocm-plugin --rocm_version=60 --rocm_path=/opt/rocm --local_xla_path=/media/win/storing/xla
failed
7900xtx AMD GPU
The text was updated successfully, but these errors were encountered: