forked from p04798526/LLaMA-Factory-Mirror
add: add predict yaml and chane old lora sft yaml
This commit is contained in:
parent
1190dfe2f2
commit
c97367ad0a
|
@ -0,0 +1,24 @@
|
||||||
|
### model
|
||||||
|
model_name_or_path: modelscope/Llama-2-7b-ms
|
||||||
|
|
||||||
|
### method
|
||||||
|
stage: sft
|
||||||
|
do_predict: true
|
||||||
|
finetuning_type: lora
|
||||||
|
|
||||||
|
### dataset
|
||||||
|
eval_dataset: alpaca_gpt4_zh
|
||||||
|
template: llama2
|
||||||
|
cutoff_len: 1024
|
||||||
|
max_samples: 50
|
||||||
|
overwrite_cache: true
|
||||||
|
preprocessing_num_workers: 16
|
||||||
|
|
||||||
|
### output
|
||||||
|
output_dir: ./results/inference/Llama2-7B/Llama2-7B_inference_08_15_11_01
|
||||||
|
overwrite_output_dir: true
|
||||||
|
|
||||||
|
### eval
|
||||||
|
per_device_eval_batch_size: 2
|
||||||
|
predict_with_generate: true
|
||||||
|
ddp_timeout: 180000000
|
|
@ -8,7 +8,7 @@ finetuning_type: lora
|
||||||
lora_target: all
|
lora_target: all
|
||||||
|
|
||||||
### dataset
|
### dataset
|
||||||
dataset: alpaca_gpt4_zh
|
dataset: belle_1m
|
||||||
template: llama2
|
template: llama2
|
||||||
cutoff_len: 1024
|
cutoff_len: 1024
|
||||||
max_samples: 10000
|
max_samples: 10000
|
||||||
|
@ -16,7 +16,7 @@ overwrite_cache: true
|
||||||
preprocessing_num_workers: 16
|
preprocessing_num_workers: 16
|
||||||
|
|
||||||
### output
|
### output
|
||||||
output_dir: ./saves/LLaMA2-7B/lora/train_24_8_13_13_16
|
output_dir: ./results/lora_sft/Llama2-7B/Llama2-7B_lora_sft_08_15_11_01
|
||||||
logging_steps: 3
|
logging_steps: 3
|
||||||
save_steps: 100
|
save_steps: 100
|
||||||
plot_loss: true
|
plot_loss: true
|
||||||
|
|
Loading…
Reference in New Issue