Make precompilation timeout configurable via TORCHINDUCTOR_PRECOMPILATION_TIMEOUT_SECONDS environment variable. #153403
Conversation
…TION_TIMEOUT_SECONDS environment variable.
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/153403
Note: Links to docs will display an error until the docs builds have been completed. This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
This PR needs a
|
| @@ -8,6 +8,12 @@ | |||
| from torch.utils._config_module import Config, get_tristate_env, install_config_module | |||
|
|
|||
|
|
|||
| precompilation_timeout_seconds: int = Config( | |||
There was a problem hiding this comment.
lets not put it at the beginning. Also I think we don't need to set an env_name for it for now
| @@ -1821,7 +1821,7 @@ def __call__( | |||
| # arg, the function will be called instead of | |||
| # generating a random torch.Tensor for benchmarking. | |||
| input_gen_fns: Optional[dict[int, Callable[[ir.Buffer], torch.Tensor]]] = None, | |||
| precompilation_timeout_seconds: int = 60 * 60, | |||
| precompilation_timeout_seconds: int = config.precompilation_timeout_seconds, | |||
There was a problem hiding this comment.
this is not the only place you need to change it. Try to search for precompilation_timeout_seconds .
Also do not set put config.precompilation_timeout_seconds in the default value. Try to find where this function is called from, and pass config.precompilation_timeout_seconds there.
Fix issue with precompilation timeout handling
Fixes #153392
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @chenyang78 @kadeng @muchulee8 @amjames @chauhang @aakhundov