Fix locked S3 dependencies for GPU Iris evals#47
Merged
Conversation
* fix(iris): persist external eval Harbor artifacts to S3 * style: format durable external eval launcher test --------- Co-authored-by: Benjamin Feuer <[email protected]>
Co-authored-by: Benjamin Feuer <[email protected]>
Contributor
|
Caution The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased. |
mrinal-bespoke
pushed a commit
to mrinal-bespoke/OpenThoughts-Agent
that referenced
this pull request
Jul 23, 2026
… -> MoE keep-1 token-salad (PR open-thoughts#47) The SKYRL_W13_RELOAD_BRACKET fix (2bb70a88) was wired only into fsdp_worker.broadcast_to_inference_engines; the megatron backend's broadcast_to_inference_engines had no bracket -> on H100/FlashInfer-CUTLASS the pre-generation weight sync served un-swapped [gate;up] FusedMoE w13 -> 100% token-salad on keep-1 (v13). Fixed by porting the bracket into megatron_worker (MarinSkyRL PR open-thoughts#47). Load-bearing gotcha for any MoE RL on the megatron backend. Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]> Claude-Session: https://claude.ai/code/session_0191M6UMWFw2eWr5y4GwC1Qg
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
GPU Iris evals install the
cloudextra in their frozen workspace closure, which supplies lock-pinneds3fsfor Harbors3://trace directories. The GPU image now uses the checked-inuv.lockdirectly, and the task bootstrap no longer force-reinstalls the environment on each run.CoreWeave eval launches use the normal Hugging Face/vLLM local-cache path instead of turning a regional cache hit into a RunAI S3 streaming URI. TPU offline mirroring is unchanged.