423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-15 03:10:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

🐞 fix: typo, move MoE fix to patcher

Browse Source
This commit is contained in:
A-Cepheus
2024-01-22 16:01:58 +08:00
parent e1d5c98519
commit 319a72b48d
2 changed files with 5 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/llmtuner/model/patcher.py
View File

@@ -284,6 +284,11 @@ def patch_model(
if is_trainable:
_prepare_model_for_training(model, model_args)
if getattr(config, "model_type", None) == "mixtral" and is_deepspeed_zero3_enabled():
from deepspeed.utils import set_z3_leaf_modules
from transformers.models.mixtral.modeling_mixtral import MixtralSparseMoeBlock
set_z3_leaf_modules(model, [MixtralSparseMoeBlock])
def patch_valuehead_model(model: "AutoModelForCausalLMWithValueHead") -> None:
def tie_weights(self: "AutoModelForCausalLMWithValueHead") -> None: