423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2026-07-07 09:35:27 +08:00
Code Issues Packages Projects Releases Wiki Activity

[v1] Fix device mesh, fix lora for reward model and fix sp (#10555)

Browse Source
This commit is contained in:
jiaqiw09
2026-06-25 20:05:56 +08:00
committed by GitHub
parent 666ee0ca78
commit b7615dbdc9
7 changed files with 58 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/v1/train_batching_strategy/train_full_fsdp2_dynamic_padding_free.yaml
View File

@@ -20,7 +20,7 @@ train_dataset: data/v1_sft_demo.yaml
output_dir: outputs/test_fsdp2
micro_batch_size: 4
batching_strategy: dynamic_padding_free
flash_attn: flash_attention2
flash_attn: flash_attention_2
cutoff_len: 2048
learning_rate: 1.0e-4
max_steps: 10

2
examples/v1/train_batching_strategy/train_full_fsdp2_padding_free.yaml
View File

@@ -20,7 +20,7 @@ train_dataset: data/v1_sft_demo.yaml
output_dir: outputs/test_fsdp2
micro_batch_size: 4
batching_strategy: padding_free
flash_attn: flash_attention2
flash_attn: flash_attention_2
cutoff_len: 2048
learning_rate: 1.0e-4
max_steps: 10