ros/LLaMA-Factory
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

support layerwise galore

Browse Source 
Former-commit-id: d43a4da0947897d0be3f62fad3107754d4c89f2b
This commit is contained in:
hiyouga
2024-03-10 00:24:11 +08:00
parent c635bbe465
commit 7ff8a064f3
14 changed files with 109 additions and 51 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/extras/galore/adamw_8bit_bf16.sh
View File

@@ -16,7 +16,7 @@ CUDA_VISIBLE_DEVICES=0 python ../../../src/train_bash.py \
--preprocessing_num_workers 16 \
--per_device_train_batch_size 1 \
--per_device_eval_batch_size 1 \
--gradient_accumulation_steps 2 \
--gradient_accumulation_steps 1 \
--lr_scheduler_type cosine \
--logging_steps 10 \
--warmup_steps 20 \