ViperEkura 9bcd696580 fix: token-level ratio and prompt masking in GRPO strategy ... - Mask prompt tokens to 0 so their logprobs excluded from ratio/KL - Switch to token-level ratio + PPO clipping via reduction='none' - Slice response token logprobs from full sequence output - Replace k3 KL estimator with non-negative k1 estimator - Fix epsilon from finifo.eps (~1e-38) to 1e-8 - Remove unused 'reduction' param from GRPOStrategy.__init__ - Clarify offline batch semantics in docstring - Add 11 unit tests for masking, advantage, KL, sync, clipping - Sync training.md and architecture.md docs		2026-07-12 21:24:09 +08:00
..
conftest.py	refactor : replace iteration with consumed_samples	2026-06-30 18:42:42 +08:00
test_callbacks.py	refactor : 清理工厂和配置系统中的死代码与冗余抽象	2026-06-07 11:39:50 +08:00
test_early_stopping.py	refactor : replace iteration with consumed_samples	2026-06-30 18:42:42 +08:00
test_grpo_strategy.py	fix: token-level ratio and prompt masking in GRPO strategy	2026-07-12 21:24:09 +08:00
test_train_strategy.py	refactor : 清理工厂和配置系统中的死代码与冗余抽象	2026-06-07 11:39:50 +08:00
test_trainer.py	refactor: 检查点加载重构，路径替代对象传递	2026-05-27 20:15:29 +08:00