YH ae86a29e23 Refact method of grad store (#2687)

2023-02-15 22:27:58 +08:00

..

bookkeeping

Refact method of grad store (#2687)

2023-02-15 22:27:58 +08:00

__init__.py

[zero] migrate zero1&2 (#1878)

2022-11-11 09:26:40 +08:00

_utils.py

[zero] fix gradient clipping in hybrid parallelism (#2521)

2023-01-29 15:09:57 +08:00

low_level_optim.py

Refact method of grad store (#2687)

2023-02-15 22:27:58 +08:00

sharded_optim_v2.py

fix move fp32 shards (#1604)

2022-09-16 17:33:16 +08:00