Repository navigation
Correct CPU importance weighting using the IMPALA residual - #36
Draft
m2kulkarni wants to merge 1 commit into
Draft
m2kulkarni wants to merge 1 commit into
m2kulkarni wants to merge 1 commit into
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Partially addresses #24 (CPU formula). Depends on #35. Author Codex root puffer-full-20261022; independent review pending.
The compiled CPU TD residual now weights the complete reward-plus-value error, following the IMPALA primary paper V-trace equations, cited in docs/development/cpu-vtrace.md. Previously only rewards were weighted. Trace recursion and terminal masks are unchanged.
Eight new compiled numerical cases compare a scalar recurrence with nonunit/zero/clipped rho, nonzero values, terminal/nonterminal and multistep traces. Before correction8 failed; after20 trainer tests passed plus9subtests. No old expected values changed or skips added. Existing training CI includes the new tests.
GPU source is not imported/supported; no GPU parity claim or complete closure of #24. Full simulator acceptance and packaging are separate. Exact-head independent review/CI pending. No merge without explicit human approval.