Wing Lian c67fb71583 Peft deepspeed resume (#1227) ...

* import deepspeed integration

* monkeypatch peft adapater with deepspeed for resume from checkpoint

* fix patch

* fix patches attempt 2

* make sure to set lora_model_dir

* skip pylint for deepspeed.utils

* pick up upstream fix in transformers

* remove monkeypatch for deepspeed/peft fix

* no need to set the lora_model_dir on resume

* unset load_in_*bit when using quant config

* guard before del

* better handling of load_in* kwargs

2024-01-31 18:13:29 -05:00

674 B

Raw Blame History

View Raw