p04798526
/
LLaMA-Factory-Mirror
1
0
Fork 1
Code Issues Pull Requests Packages Projects Releases Wiki Activity

train: change yaml

This commit is contained in:
wql 2024-08-13 10:05:16 +08:00
parent 0841a0832f
commit f44393f413
1 changed files with 9 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
examples/train_lora/llama2_lora_sft.yaml
View File

@ -13,10 +13,9 @@ template: llama2
cutoff_len: 1024 cutoff_len: 1024
max_samples: 100000 max_samples: 100000
overwrite_cache: true overwrite_cache: true
preprocessing_num_workers: 16
### output ### output
output_dir: ./saves/LLaMA2-7B/lora/train_24_8_13_08_13 output_dir: ./saves/LLaMA2-7B/lora/train_24_8_13_10_02
logging_steps: 3 logging_steps: 3
save_steps: 100 save_steps: 100
plot_loss: true plot_loss: true
@ -25,16 +24,15 @@ overwrite_output_dir: true
### train ### train
per_device_train_batch_size: 2 per_device_train_batch_size: 2
gradient_accumulation_steps: 8 gradient_accumulation_steps: 8
learning_rate: 1.5e-4 learning_rate: 5.0e-05
num_train_epochs: 10 num_train_epochs: 10
lr_scheduler_type: cosine lr_scheduler_type: cosine
warmup_ratio: 0.1
fp16: true fp16: true
ddp_timeout: 180000000
### eval
val_size: 0.1
per_device_eval_batch_size: 1 # 1 最大
eval_strategy: steps
eval_steps: 500
max_grad_norm: 1.0
warmup_steps: 0
optim: adamw_torch
report_to: none
lora_rank: 8
lora_alpha: 16
lora_dropout: 0.1