Repository navigation
Fix RSI training guest execution - #163
Merged
Merged
Conversation
… error torch's triton JIT needs a C compiler at forward time; without it QLoRA training fails in-guest. Also surface the guest stdout tail when a model job omits its result file, instead of reporting only the exit code.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Unblocks #9's live QLoRA run. Three guest-execution defects found while running the documented bounded command:
/opt/headlesscode-rsi-training/*.py, unreachable under the sandbox path policy (only /workspace + system paths are open). The scripts already ship in the job snapshot from the frozen base commit, so run them from/workspace/scripts/rsi-training/(stronger provenance: bound to the job digest). Image no longer copies the scripts.Verified: manual guest training completes 8/8 steps; fleet training completes with adapter artifact. Paired eval pending a run whose base commit includes the trainer fix.