Created by: suchenzang
Following https://github.com/facebookresearch/metaseq/pull/230 and continuing to break down https://github.com/facebookresearch/metaseq/pull/197.
Removed:
- unused
moe_disable_paddingarg - unused
from_pretrainedmethods, since we currently depend onload_model_ensemble_and_taskfromcheckpoint_utils(not great but saving that for another PR) - unused
hub_modelsmethod - unused
return_all_hiddensarg, and noencoder_statesbeing returned as a result -
EncoderDecoderModelclass