lars
fcd77c2f4b
CI / Sync project version with tag (push) Has been skipped
CI / Lint (ruff check) (push) Successful in 1m29s
CI / Type check (ty) (push) Successful in 1m26s
CI / Format (ruff format) (push) Successful in 1m26s
CI / Sync project version with tag (pull_request) Has been skipped
CI / Lint (ruff check) (pull_request) Successful in 4m0s
CI / Format (ruff format) (pull_request) Successful in 3m59s
CI / Tests (push) Successful in 5m41s
CI / Type check (ty) (pull_request) Successful in 4m1s
CI / Bump version, tag, and update changelog on merge to master (push) Has been skipped
CI / Tests (pull_request) Successful in 4m15s
CI / Bump version, tag, and update changelog on merge to master (pull_request) Has been skipped
Taking argmax over the n_sec classifier logits collapses secondary multiplicity onto its conditional mode at fixed pre-step conditioning, under-dispersing n_sec in rollouts and biasing low wherever the true conditional count distribution is right-skewed (typical for multiplicity). Generalizes stage2_model.n_sec.stop_sampling (previously stop_token-only) into stage2_model.n_sec.sampling, covering both "head" (greedy: argmax; sample: categorical draw via torch.multinomial) and "stop_token" (unchanged: greedy threshold / Bernoulli draw) modes. stop_sampling is kept as a deprecated alias in NSecConfig.from_dict and migrate_config, since it appears in existing checkpoints' model_config. Default stays "greedy" so existing runs/checkpoints are unaffected.