[GraphTrainer][AutoDev] Add gradient accumulation integration test#2914
Draft
SherlockNoMad wants to merge 2 commits intomainfrom
Draft
[GraphTrainer][AutoDev] Add gradient accumulation integration test#2914SherlockNoMad wants to merge 2 commits intomainfrom
SherlockNoMad wants to merge 2 commits intomainfrom
Conversation
Port PR #2814: Add a JIT gradient accumulation integration test that verifies training with 2 gradient accumulation steps (local_batch_size=8, global_batch_size=64, 4 GPUs).
meta-cla
bot
added
the
CLA Signed
SherlockNoMad
commented
Apr 9, 2026
| # multiply by 2: 32 * 2 = 64. | ||
| "--module graph_trainer.llama3", | ||
| "--config graph_trainer_llama3_debugmodel", | ||
| "--compile.mode jit", |
Contributor
Author
There was a problem hiding this comment.
Good catch, but don't do this for jit mode, do this for aot_fx_trace, and test it with GraphTrainer.
we are planning to deprecate jit and aot, remember that.
Now try add the same gradient accumulation test for aot_fx_trace.
Contributor
Author
There was a problem hiding this comment.
AutoDev: Changed the test from JIT mode to aot_fx_trace mode and moved it to the aot_fx_trace test section. JIT and AOT are being deprecated in favor of aot_fx_trace.
…race mode Move the gradient accumulation integration test from the JIT section to the aot_fx_trace section, as JIT and AOT modes are being deprecated in favor of aot_fx_trace.
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
local_batch_size=8withglobal_batch_size=64on 4 GPUs, resulting in 2 gradient accumulation steps per optimizer stepjit_optional_checkpointand before AOT mode testsTest plan
jit_gradient_accumulationtest with 4 GPUs