Repository navigation
Conversation
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Adds a
mosaic_rlexample that generates grouped rollouts —kattempts per task, the shape GRPO-style trainers (SkyRL, Prime-RL) consume — on Mosaic Sandbox, plus a row in the root README's Optimization Examples table.uv run harbor_cookbook/mosaic_rl/rollout.py \ -t hello-world/hello-world --group-size 8 --env mosaic # hello-world: [1.0, 0.0, 1.0, 1.0, ...]Why it's a separate example rather than a flag on
harbor_rl: the sibling example targets the RL API from open PR #1291, whereas this one runs entirely on released Harbor 0.22Job/JobConfigvia a helper in the Mosaic SDK, so it works against a released Harbor today.--envacceptsmosaic,daytona, or anymodule:Classimport path, so the same script is the A/B harness.The point of the example is snapshot amortization: a group builds the task environment once into a Mosaic snapshot and restores an isolated sandbox per attempt, so only the first attempt pays the build.
It uses
import_pathtoday:and the README notes that
--env mosaicbecomes native once the provider PR lands upstream.The README's Mosaic-vs-Daytona section is methodology only — what to hold constant (task set,
k, concurrency, agent, model) and what to record (wall-clock, rollouts/sec, mean per-trial wall time, first-attempt vs. subsequent-attempt cost). No numbers are quoted, because I haven't run the comparison at a scale worth publishing.Requires a Mosaic credential (
mos loginorMOSAIC_API_TOKEN), a model-provider key, andDAYTONA_API_KEYonly for the comparison arm.