fix(ray): name the missing flag instead of failing on int + None - #2519
Draft
Arist12 wants to merge 1 commit into
Draft
fix(ray): name the missing flag instead of failing on int + None#2519Arist12 wants to merge 1 commit into
int + None#2519Arist12 wants to merge 1 commit into
Conversation
Arist12
force-pushed
the
fix/require-rollout-num-gpus-when-disaggregated
branch
from
August 13, 2026 16:32
fd44e51 to
1035093
Compare
Without --colocate the rollout GPU count cannot be derived, and leaving --rollout-num-gpus unset surfaces as a TypeError on None inside placement-group sizing, several frames from the flag that is missing. --debug-rollout-only has the same gap one step earlier, in miles_validate_args, where the count feeds min(8, ...). Assert at each point of use.
Arist12
force-pushed
the
fix/require-rollout-num-gpus-when-disaggregated
branch
from
August 13, 2026 17:47
1035093 to
ed07904
Compare
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
Without
--colocatethe rollout GPU count cannot be derived, and leaving--rollout-num-gpusunset surfaces several frames later asfrom placement-group sizing, which never names the flag.
--debug-rollout-onlyhas the same gap one step earlier, inmiles_validate_args, where the count feedsmin(8, ...).Reproduced by dropping
--rollout-num-gpus 1fromexamples/lora/run-qwen2.5-3B-megatron-lora-disaggregated.sh.Fix
Assert at each point of use, naming the flag. Both places, because they are reached by different paths —
_get_placement_group_layoutfor the ordinary disaggregated run,miles_validate_argsfor--debug-rollout-only, which fails before the placement group is ever built.Not in
miles_validate_argsfor the general case: about 20 tests intests/fast/utils/test_arguments.pycall it with a minimal argument set (colocate off,rollout_num_gpusunset) and expect it to pass.Test
One case added to
tests/fast/ray/test_placement_group_shared_ppo.py, which already covers_get_placement_group_layoutfor colocate /debug_train_only/rollout_externalwith the same fixture.