let hf trainer handle torch compile #516
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
tmm1
reviewed
Aug 31, 2023
tmm1
approved these changes
Aug 31, 2023
NanoCode012
reviewed
Sep 1, 2023
| @@ -526,6 +526,10 @@ wandb_log_model: # "checkpoint" to log model to wandb Artifacts every `save_step | |||
| # where to save the finished model to | |||
| output_dir: ./completed-model | |||
|
|
|||
| # whether to use torch.compile and which backend to use | |||
| torch_compile: # bool | |||
| torch_compile_backend: # Optional[str] | |||
There was a problem hiding this comment.
Is it necessary to open this "backend" config?
Refer to the PyTorch doc for possible values and note that they may change across PyTorch versions.
and reading the source of the torch compile, it seems to only have a condition for inductor.
There was a problem hiding this comment.
Not sure; there are other backends but maybe they're not useful?
torch._dynamo.list_backends()
['aot_ts_nvfuser', 'cudagraphs', 'inductor', 'ipex', 'nvprims_nvfuser', 'onnxrt', 'tvm']
NanoCode012
reviewed
Sep 1, 2023
|
It seems like there may be small benefits to leaving it to default on. Is there any case having it off is better? |
it cannot default on because it doesn't work. if you try to train it errors out. |
winglian
commented
Sep 7, 2023
| @@ -579,6 +580,21 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer, total_num_ | |||
| if cfg.bench_dataset: | |||
| training_arguments_kwargs["bench_dataset"] = cfg.bench_dataset | |||
|
|
|||
| if cfg.torch_compile: | |||
| if torch.__version__ < "2.1.0": # pylint: disable=protected-access | |||
There was a problem hiding this comment.
@tmm1 lmk if you think this is good enough for now
mkeoliya
pushed a commit
to mkeoliya/axolotl
that referenced
this pull request
Dec 15, 2023
* let hf trainer handle torch compile * remove torch compile checks, include option for backend * suppress torch errors to get further * require min torch version of 2.1.0 for torch compile to work --------- Co-authored-by: Aman Karmani <aman@tmm1.net>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.