Files
foxhunt/infra
jgrusewski 3541ba7352 perf(diag): add --diag-every flag + fix PnL accumulation bug
- --diag-every N (default 100): skip device readback on non-diag steps.
  Eliminates ~10 stream.sync per step → ~2× throughput at large batch.
- Fix PnL bug: was accumulating raw_rewards (shaped, all batches) instead
  of rewards/scale on done steps only.
- Wire diag-every=100 in Argo template (diag JSONL every 100 steps).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-25 23:56:21 +02:00
..