X-LANCE / SLAM-LLM

Speech, Language, Audio, Music Processing with Large Language Model
MIT License
579 stars 52 forks source link

Fixes #92: FSDP training raise "KeyError: 'ShardingStrategy.NO_SHARD'" #145

Closed nuaalixu closed 1 month ago

nuaalixu commented 1 month ago

What does this PR do?

Fixes #92

Feature/Issue validation/testing

This KeyError occurs because the value of fsdp_config.sharding_strategy is mandatorily converted into a str obj rather than ShardingStrategy obj. This is the intrinsic feature of omegaconf.dictconfig.DictConfig. See this for more details.