This commit is contained in:
@@ -0,0 +1,48 @@
|
||||
import os
|
||||
|
||||
os.environ['CUDA_VISIBLE_DEVICES'] = '0,1'
|
||||
os.environ['NPROC_PER_NODE'] = '2'
|
||||
|
||||
|
||||
def train():
|
||||
from swift import RLHFArguments, rlhf_main
|
||||
result = rlhf_main(
|
||||
RLHFArguments(
|
||||
rlhf_type='gkd',
|
||||
model='Qwen/Qwen3.5-4B',
|
||||
teacher_model='Qwen/Qwen3.5-4B',
|
||||
tuner_type='lora',
|
||||
lora_rank=64,
|
||||
lora_alpha=128,
|
||||
target_modules=['all-linear'],
|
||||
use_vllm=True,
|
||||
vllm_mode='colocate',
|
||||
vllm_gpu_memory_utilization=0.7,
|
||||
vllm_max_model_len=10240,
|
||||
sleep_level=1,
|
||||
external_plugins=['examples/train/rlhf/opsd/opsd_plugin.py'],
|
||||
dataset=['open-r1/OpenThoughts-114k-math'],
|
||||
lmbda=1.0,
|
||||
beta=0.5,
|
||||
temperature=1.2,
|
||||
sft_alpha=0,
|
||||
torch_dtype='bfloat16',
|
||||
max_steps=1000,
|
||||
per_device_train_batch_size=4,
|
||||
gradient_accumulation_steps=1,
|
||||
learning_rate=2e-5,
|
||||
save_steps=100,
|
||||
save_total_limit=10,
|
||||
logging_steps=1,
|
||||
max_length=8192,
|
||||
max_completion_length=2048,
|
||||
save_only_model=True,
|
||||
gradient_checkpointing=True,
|
||||
deepspeed='zero0',
|
||||
attn_impl='flash_attn',
|
||||
))
|
||||
return result
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
train()
|
||||
Reference in New Issue
Block a user