main

GGitHubfeat(api): add unified RejectionSamplingConfig for async training (#1088)

bc9f0098创建于 4月20日历史提交

文件	最后提交记录	最后更新时间
tongyi_deepresearch	feat(api): add unified RejectionSamplingConfig for async training (#1088) Replace behave_imp_weight_cap/behave_imp_weight_mode with unified RejectionSamplingConfig supporting multiple metrics (ratio, kl_k1, kl_k2, kl_k3), levels (token/sequence), and actions (mask/clamp). Key changes: - Add RejectionSamplingConfig dataclass with comprehensive validation - Implement apply_rejection_sampling for 1D packed and 2D padded formats - Fix loss denominator scaling bug in mask mode (save count before filtering) - Use geometric mean for sequence-level ratio aggregation (matching GSPO) - Broadcast sequence-level geometric mean as uniform behave_imp_weight - Warn when use_decoupled_loss=True but rejection_sampling is None - Update ppo_actor_loss_fn and grpo_loss_fn to use new config - Migrate 40 example configs to new rejection_sampling field - Add 43 unit tests covering all modes, metrics, and edge cases Refs: #1052	5 个月前
local_1.5b_example.yaml	feat(api): add unified RejectionSamplingConfig for async training (#1088) Replace behave_imp_weight_cap/behave_imp_weight_mode with unified RejectionSamplingConfig supporting multiple metrics (ratio, kl_k1, kl_k2, kl_k3), levels (token/sequence), and actions (mask/clamp). Key changes: - Add RejectionSamplingConfig dataclass with comprehensive validation - Implement apply_rejection_sampling for 1D packed and 2D padded formats - Fix loss denominator scaling bug in mask mode (save count before filtering) - Use geometric mean for sequence-level ratio aggregation (matching GSPO) - Broadcast sequence-level geometric mean as uniform behave_imp_weight - Warn when use_decoupled_loss=True but rejection_sampling is None - Update ppo_actor_loss_fn and grpo_loss_fn to use new config - Migrate 40 example configs to new rejection_sampling field - Add 43 unit tests covering all modes, metrics, and edge cases Refs: #1052	5 个月前