参考地址：
https://github.com/hiyouga/LLaMA-Factory/blob/main/examples/README_zh.md#lora-微调
使用 Adam-mini 进行全参数训练
llamafactory-cli train examples/extras/adam_mini/qwen2_full_sft.yaml
qwen2_full_sft.yaml配置：
### model
model_name_or_path: /data/model/Qwen2.5-0.5B-Instruct
### method
stage: sft
do_train: true
finetuning_type: full
use_adam_mini: true
### dataset
dataset: alpaca_my_data
template: qwen
cutoff_len: 1024
max_samples: 1000
overwrite_cache: true
preprocessing_num_workers: 1
### output
output_dir: saves/qwen2_5-0_5b/full/sft
logging_steps: 10
save_steps: 500
plot_loss: true
overwrite_output_dir: true
### train
per_device_train_batch_size: 1
gradient_accumulation_steps: 8
learning_rate: 1.0e-5
num_train_epochs: 100.0
lr_scheduler_type: cosine
warmup_ratio: 0.1
bf16: true
ddp_timeout: 180000000
### eval
val_size: 0.1
per_device_eval_batch_size: 1
eval_strategy: steps
eval_steps: 500
opencompass模型评测工具
opencompass$ python run.py --models vllm_merged_sft --datasets demo_gsm8k_chat_gen demo_math_chat_gen --debug
| dataset |
version |
metric |
mode |
qwen2.5-0.5b-instruct-hf |
| demo_gsm8k |
1d7fe4 |
accuracy |
gen |
39.06 |
| demo_math |
393424 |
accuracy |
gen |
17.19 |
| dataset |
version |
metric |
mode |
qwen2.5-0.5b-instruct-vllm-local |
| demo_gsm8k |
1d7fe4 |
accuracy |
gen |
43.75 |
| demo_math |
393424 |
accuracy |
gen |
20.31 |
| dataset |
version |
metric |
mode |
qwen2.5-0.5b-instruct-vllm-sft-merged |
| demo_gsm8k |
1d7fe4 |
accuracy |
gen |
37.50 |
| demo_math |
393424 |
accuracy |
gen |
21.88 |
| dataset |
version |
metric |
mode |
internlm2-chat-1.8b-hf |
qwen2-1.5b-instruct-hf |
| demo_gsm8k |
1d7fe4 |
accuracy |
gen |
32.81 |
57.81 |
| demo_math |
393424 |
accuracy |
gen |
14.06 |
20.31 |
| dataset |
version |
metric |
mode |
qwen2.5-0.5b-instruct-vllm-sft |
| demo_gsm8k |
1d7fe4 |
accuracy |
gen |
4.69 |
| demo_math |
393424 |
accuracy |
gen |
1.56 |
