recast loralayer, norm, lmhead + embed token weights per original qlora (#393)

* recast loralayer, norm, lmhead + embed token weights per original qlora

* try again for the fix

* refactor torch dtype picking

* linter fixes

* missing import for LoraLayer

* fix install for tests now that peft is involved
This commit is contained in:
Wing Lian
2023-08-21 18:41:12 -04:00
committed by GitHub
parent 50682a3c06
commit 96deb6bd67
4 changed files with 33 additions and 22 deletions

View File

@@ -32,5 +32,8 @@ setup(
"extras": [
"deepspeed",
],
"peft": [
"peft @ git+https://github.com/huggingface/peft.git",
],
},
)