423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-03 04:02:49 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #3276 from liu-zichen/fix_mixtral

Browse Source 
fix: turn on output_router_logits of mixtral
Former-commit-id: 0e0942d388bdb0122001a7f8e081315059d5d327
This commit is contained in:
hoshi-hiyouga 2024-04-15 15:38:16 +08:00 committed by GitHub
commit 1bdf7e4b9d
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/llmtuner/model/patcher.py
View File

@ -316,7 +316,7 @@ def patch_config(
if getattr(config, "model_type", None) == "qwen2" and is_trainable and model_args.flash_attn: if getattr(config, "model_type", None) == "qwen2" and is_trainable and model_args.flash_attn:
setattr(config, "use_cache", False) # qwen2 does not support use_cache when using flashattn setattr(config, "use_cache", False) # qwen2 does not support use_cache when using flashattn
if getattr(config, "model_type", None) == "qwen2_moe" and is_trainable: if getattr(config, "model_type", None) in ["mixtral", "qwen2_moe"] and is_trainable:
setattr(config, "output_router_logits", True) setattr(config, "output_router_logits", True)
init_kwargs["torch_dtype"] = model_args.compute_dtype init_kwargs["torch_dtype"] = model_args.compute_dtype