"git@developer.sourcefind.cn:sugon_wxj/megatron-lm.git" did not exist on "c53467942e26c5dbd12e8f1b3c517c75084a381e"
-
statelesshz authored
* deprecate fairscale's ShardedDDP * fix code style * roll back * deprecate the `sharded_ddp` training argument --------- Co-authored-by:jihuazhong <jihuazhong1@huawei.com>
8ba26c18
