forked from PaddlePaddle/PaddleNLP
-
Notifications
You must be signed in to change notification settings - Fork 0
/
sft_argument.json
34 lines (34 loc) Β· 962 Bytes
/
sft_argument.json
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
{
"model_name_or_path": "IEITYuan/Yuan2-2B",
"dataset_name_or_path": "./data",
"output_dir": "./checkpoints/yuan_sft_ckpts",
"per_device_train_batch_size": 1,
"gradient_accumulation_steps": 2,
"per_device_eval_batch_size": 1,
"eval_accumulation_steps":1,
"num_train_epochs": 3,
"learning_rate": 3e-05,
"warmup_steps": 30,
"logging_steps": 1,
"evaluation_strategy": "epoch",
"save_strategy": "epoch",
"src_length": 2048,
"max_length": 4096,
"bf16": true,
"fp16_opt_level": "O2",
"do_train": true,
"do_eval": true,
"disable_tqdm": true,
"load_best_model_at_end": true,
"eval_with_do_generation": false,
"metric_for_best_model": "accuracy",
"recompute": true,
"save_total_limit": 1,
"tensor_parallel_degree": 1,
"pipeline_parallel_degree": 1,
"pipeline_parallel_config": "disable_p2p_cache_shape",
"sharding": "stage1",
"zero_padding": false,
"unified_checkpoint": true,
"use_flash_attention": false
}