ViperEkura 2c7a71a9c0 refactor: separate old policy and ref model in GRPO strategy ...

- Split single ref_model into old_model (importance sampling ratio) and ref_model (frozen KL regularizer)
- Move ref_model/old_model creation from strategy __init__ to TrainContextBuilder, pass as explicit parameters
- Remove periodic sync_ref_model + sync_interval; add sync_old_model for external rollout loop to call
- DPOStrategy also receives ref_model from builder
- Fix std to use unbiased=False (population std per GRPO paper)
- Remove redundant tests (test_grpo_kl_zero_at_init, test_grpo_no_sync_interval_param)
- Remove --grpo_sync_interval CLI arg

2026-07-14 20:03:45 +08:00

..

config

fix : SFT pipeline position_ids default & doc boundary preservation

2026-07-04 15:59:11 +08:00

dataset

feat: add JSONL dataset store with on-the-fly tokenization

2026-07-04 15:42:33 +08:00

extension

feat: wire up paged decode CUDA kernel to Python extension

2026-07-11 18:40:49 +08:00

inference

refactor: clean up inference design patterns

2026-07-05 11:41:54 +08:00

model

perf: replace paged KV cache with contiguous ContiguousCache, decode all groups

2026-07-05 11:34:36 +08:00

parallel

feat : add distributed checkpoint via executor checkpoint_context

2026-07-13 12:27:09 +08:00

preprocessing

refactor: split mask builder by single/multi output

2026-07-08 21:18:34 +08:00

serialization

feat : add distributed checkpoint via executor checkpoint_context

2026-07-13 12:27:09 +08:00

tokenize

refactor : Pipeline 去除去重，ids 重命名为 sequence，泛型透传

2026-05-31 15:14:27 +08:00

trainer

refactor: separate old policy and ref model in GRPO strategy

2026-07-14 20:03:45 +08:00

__init__.py

chore: bump version to 1.3.9

2026-07-12 21:04:20 +08:00

factory.py

fix: ForwardRef._evaluate Python 3.12 compatibility

2026-07-03 18:41:19 +08:00

protocols.py

refactor: 重构训练后端为 Executor 模式

2026-05-24 20:35:44 +08:00