| commit | 2bef72007dd975fa7469d2b1142378d037610981 | [log] [tgz] |
|---|---|---|
| author | Scott Todd <scott.todd0@gmail.com> | Tue Jul 30 08:14:38 2024 -0700 |
| committer | GitHub <noreply@github.com> | Tue Jul 30 08:14:38 2024 -0700 |
| tree | f102c7c7e886bd3b41d75b410734f68ea51a18f8 | |
| parent | d39c3c56682e006e842b32aa6f38c272f77c8f3c [diff] |
Migrate GPU test jobs to pkgci. (#18007) Progress on https://github.com/iree-org/iree/issues/16203. Depends on https://github.com/iree-org/iree/pull/18000. These jobs used to use the 3.2GB install directory produced by `cmake --build full-build-dir --target install` in the `build_all` job. Now they use the 73MB Python packages produced by `python -m pip wheel runtime/` and `python -m pip wheel compiler/` in the `build_packages` job. Python packages are what we expect users to consume, so test jobs should use them too. * Note that the Python packages will be larger once we enable asserts and/or debug symbols in them. These tests may also fail with less useful error messages and callstacks as a result of this change until that is fixed. I tried to keep changes to the workflow jobs minimal for now. Once the migrations are further along we can cut out some of the remaining layers of scripts / Dockerfiles. As before, these jobs are all opt-in on presubmit (always running on LLVM integrate PRs or PRs affecting NVGPU/AMDGPU code). Diffs between previous jobs and new jobs to confirm how similar they are: Job name | Logs before | Logs after | Notes -- | -- | -- | -- `test_nvidia_t4` | [workflow logs](https://github.com/iree-org/iree/actions/runs/10102664951/job/27939423899) | [workflow logs](https://github.com/iree-org/iree/actions/runs/10102841435/job/27939725841?pr=18007) | 433 tests -> 430 tests<br>skipping `samples/custom_dispatch/vulkan/shaders`<br>`IREE custom_dispatch/vulkan/shaders ignored -- glslc not found`<br>(no longer running under Docker) `test_amd_mi250` | [workflow logs](https://github.com/iree-org/iree/actions/runs/10102664951/job/27939423747) | [workflow logs](https://github.com/iree-org/iree/actions/runs/10102841435/job/27939725347?pr=18007) | 138 tests before/after `test_amd_mi300` | [workflow logs](https://github.com/iree-org/iree/actions/runs/10102664951/job/27939424223) | [workflow logs](https://github.com/iree-org/iree/actions/runs/10102841435/job/27939725525?pr=18007) | 141 tests before/after `test_amd_w7900` | [workflow logs](https://github.com/iree-org/iree/actions/runs/10102664951/job/27939424084) | [workflow logs](https://github.com/iree-org/iree/actions/runs/10102841435/job/27939725679?pr=18007) | 289 tests before/after Each job is now included from its own standalone workflow file, allowing for testing of individual workflows using `workflow_dispatch` triggers. I have some other ideas for further decoupling these optional jobs from the core workflow(s). ci-extra: test_amd_mi250, test_amd_mi300, test_amd_w7900, test_nvidia_t4
IREE (Intermediate Representation Execution Environment, pronounced as “eerie”) is an MLIR-based end-to-end compiler and runtime that lowers Machine Learning (ML) models to a unified IR that scales up to meet the needs of the datacenter and down to satisfy the constraints and special considerations of mobile and edge deployments.
See our website for project details, user guides, and instructions on building from source.
IREE is still in its early phase. We have settled down on the overarching infrastructure and are actively improving various software components as well as project logistics. It is still quite far from ready for everyday use and is made available without any support at the moment. With that said, we welcome any kind of feedback on any communication channels!
See our website for more information.
Community meeting recordings: IREE YouTube channel
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.