ChasonJiang 0e868ada60 更改gpt并行推理时的mask策略为padding left，使batch_infer更接近于naive_infer ...

减少冗余操作并使用torch_sdpa，以提升推理速度

2025-03-04 16:27:18 +08:00

..

.gitignore

chores (#1528)

2024-08-23 17:47:27 +08:00

s1.yaml

more code refactor

2024-01-16 17:14:18 +01:00

s1big2.yaml

more code refactor

2024-01-16 17:14:18 +01:00

s1big.yaml

more code refactor

2024-01-16 17:14:18 +01:00

s1longer-v2.yaml

support gpt-sovits v2

2024-08-02 15:48:02 +08:00

s1longer.yaml

more code refactor

2024-01-16 17:14:18 +01:00

s1mq.yaml

more code refactor

2024-01-16 17:14:18 +01:00

s2.json

update Gradient Checkpointing to reduce VRAM usage (#2040)

2025-02-12 23:00:34 +08:00

train.yaml

more code refactor

2024-01-16 17:14:18 +01:00

tts_infer.yaml

更改gpt并行推理时的mask策略为padding left，使batch_infer更接近于naive_infer

2025-03-04 16:27:18 +08:00