Spaces:
Sleeping
Sleeping
Commit History
-
fix: improve GRPO logging and cache sampling behavior
garg-tejasAdd robust metric-key fallback logic so reward logging works across TRL versions, and include trajectory cache updates in the same training pipeline commit as requested.d140675 -
fix: convert dict actions to JSON strings for env.step() in cache building and replay
garg-tejas8271223 -
fix: random-step dataset + trajectory cache + experiment tracking for L40S
garg-tejas- Add server/greedy_policy.py: shared greedy dispatcher with graph travel times - Add server/trajectory_cache.py: parallel cache build, random-step sampling, replay - Refactor train_trl_grpo.py: - build_seed_dataset now samples random steps from greedy trajectories - make_reward_fn replays cache, injects model action, continues greedy - Fixes step-0-only training bug (model now learns all steps) - Add --num-generations, --cache-workers, --steps-per-seed, --cache-dir CLI args - Defaults optimized for L40S: 8 gens, 8 workers, 7 steps/seed, /data cache - Add wandb + TensorBoard experiment tracking - Update server/prompt_utils.py: explicit reward structure in SYSTEM_PROMPT - Update scripts/launch_hf_job.py: - Add --script grpo option - Add --wandb-project, --wandb-entity, --wandb-run-name, --tensorboard - Pass WANDB_API_KEY to job env - Default flavor: l40sx1, default num_generations: 8 - Update train_grpo.py: add HF Hub push, metrics.jsonl logging - Update working_note…6da8b7c