| commit | d6d64df5e70ae07b2e09d3f61feabcbbef5e8306 | [log] [tgz] |
|---|---|---|
| author | Scott Todd <scotttodd@google.com> | Tue Aug 22 09:17:16 2023 -0700 |
| committer | GitHub <noreply@github.com> | Tue Aug 22 09:17:16 2023 -0700 |
| tree | 29b3420cb8ccc88e5dd8c1afaa028a54de940a13 | |
| parent | 2a914c1e7fbfbdecafd506c29ba30df8ea251046 [diff] |
Include all GPU tests in test_a100 job. (#14765) See https://github.com/openxla/iree/issues/14169 for context. Tha `test_a100` job runs on a100 machines, which are in low supply relative to demand and cost more to run, so we only run them on postsubmit by default. It can take hours for a runner to be available, but then actual job time is usually around 10 minutes, only ~20 seconds of which is actually running tests ([sample logs](https://github.com/openxla/iree/actions/runs/5928273900/job/16074197115#step:7:265)). Previously, we only ran a100-specific tests (via the `requires-gpu-sm80` label). This PR changes to run all GPU tests (well, this should be all "a100-compatible GPU tests", but we don't have any AMD/ROCm/etc.-specific tests today). Differences between GPU architectures and driver versions are common, so it will help to have baseline feature test coverage on as many GPUs as we can get. configuration | total time | ctest time | number of tests | sample log link -- | -- | -- | -- | -- baseline | 7m 51s | 20 seconds | 13 | [logs](https://github.com/openxla/iree/actions/runs/5922237543/job/16056162777#step:7:272) all GPU tests, parallel 2 | 12m 32s | 292 seconds | 370 | [logs](https://github.com/openxla/iree/actions/runs/5931611100/job/16084038030?pr=14765#step:7:1008) all GPU tests, parallel 4 | 11m 40s | 248 seconds | 370 | [logs](https://github.com/openxla/iree/actions/runs/5932490304/job/16086531533?pr=14765#step:7:1010) all GPU tests, parallel 8 | 11m 42s | 245 seconds | 370 (1 timeout) | [logs](https://github.com/openxla/iree/actions/runs/5931784428/job/16084543385?pr=14765#step:7:1053) ci-exactly: build_all, test_a100
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.
IREE is licensed under the terms of the Apache 2.0 License with LLVM Exceptions. See LICENSE for more information.