Move get_world_size_and_rank to utils
#2155
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchtune/2155
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 38ff8e4 with merge base 9cfa288 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
joecummings
changed the title
get_world_size_and_rank to utils
ebsmothers
reviewed
Dec 12, 2024
| @@ -21,6 +21,19 @@ | |||
| BlockMask = torch.Tensor | |||
|
|
|||
|
|
|||
| def get_world_size_and_rank() -> Tuple[int, int]: | |||
There was a problem hiding this comment.
I understand the goal of this and agree with moving this function down into utils. But I also wonder whether device.py is really the best place for it (I know we already have _get_local_rank in here, and I know that having separate utils/distributed.py and training/distributed.py is probably confusing)
There was a problem hiding this comment.
I'm in favor of keeping it in here.
ebsmothers
approved these changes
Dec 13, 2024
13 tasks
rahul-sarvam
added a commit
to sarvamai/torchtune
that referenced
this pull request
Dec 18, 2024
* Llama 3.3 70B (pytorch#2124) * Llama 3.3 readme updates (pytorch#2125) * update configs (pytorch#2107) Co-authored-by: Felipe Mello <[email protected]> * Reduce logging output for distributed KD (pytorch#2120) * Support Early Exit Loss and/or Layer Dropout (pytorch#1076) Co-authored-by: ebsmothers <[email protected]> * Update checkpointing directory (pytorch#2074) Co-authored-by: Felipe Mello <[email protected]> Co-authored-by: vancoyendall <[email protected]> * pass correct arg (pytorch#2127) Co-authored-by: Felipe Mello <[email protected]> * update configs (pytorch#2128) Co-authored-by: Felipe Mello <[email protected]> * fix qat_lora_test (pytorch#2131) Co-authored-by: Felipe Mello <[email protected]> * guard ckpt imports (pytorch#2133) Co-authored-by: Felipe Mello <[email protected]> * [bug fix] add parents=True (pytorch#2136) Co-authored-by: Felipe Mello <[email protected]> * [bug fix] re-add model (pytorch#2135) Co-authored-by: Felipe Mello <[email protected]> * Update save sizes into GiB (pytorch#2143) * [bug fix] remove config download when source is kaggle (pytorch#2144) Co-authored-by: Felipe Mello <[email protected]> * [fix] remove "with_suffix" (pytorch#2146) Co-authored-by: Felipe Mello <[email protected]> * DoRA fixes (pytorch#2139) Co-authored-by: Mircea Mironenco <[email protected]> * [Fix] Llama 3.2 Vision decoder_trainable flag fixed (pytorch#2150) * Small readme, config updates (pytorch#2157) * Using `FormattedCheckpointFiles` in configs (pytorch#2147) * Move ``get_world_size_and_rank`` to utils (pytorch#2155) * Faster intermediate checkpoints with DCP async save in TorchTune (pytorch#2006) Co-authored-by: Saurabh Mishra <[email protected]> * torchdata integration - multi-dataset and streaming support (pytorch#1929) * Allow higher version of lm-eval (pytorch#2165) * Using `FormattedCheckpointFiles` in configs... round 2 (pytorch#2167) * [EZ] Fix set_torch_num_threads in multi-node. (pytorch#2164) --------- Co-authored-by: Philip Bontrager <[email protected]> Co-authored-by: ebsmothers <[email protected]> Co-authored-by: Felipe Mello <[email protected]> Co-authored-by: Felipe Mello <[email protected]> Co-authored-by: Joe Cummings <[email protected]> Co-authored-by: Mostafa Elhoushi <[email protected]> Co-authored-by: vancoyendall <[email protected]> Co-authored-by: Mircea Mironenco <[email protected]> Co-authored-by: salman <[email protected]> Co-authored-by: Saurabh Mishra <[email protected]> Co-authored-by: Saurabh Mishra <[email protected]> Co-authored-by: Andrew Ho <[email protected]> Co-authored-by: Eugen Hotaj <[email protected]>
rahul-sarvam
pushed a commit
to sarvamai/torchtune
that referenced
this pull request
Dec 23, 2024
rahul-sarvam
pushed a commit
to sarvamai/torchtune
that referenced
this pull request
Dec 23, 2024
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.
Context
What is the purpose of this PR? Is it to
Closes #2151
Changelog
What are the changes made in this PR?
get_world_size_and_rankfrom torchtune/training to torchtune/utils. Why does this make sense? B/c it's more of a utility anyways and can be used outside of the training loopget_world_size_and_rankTest plan
Please make sure to do each of the following if applicable to your PR. If you're unsure about any one of these just ask and we will happily help. We also have a contributing page for some guidance on contributing.
pre-commit install)pytest testspytest tests -m integration_testUX
If your function changed a public API, please add a dummy example of what the user experience will look like when calling it.
Here is a docstring example
and a tutorial example