Nin8520/Qwen2.5-7B

Model Details

• Base Model: Qwen2.5-7B-Instruct

Training Details

Training Data

hon9kon9ize/yue-alpaca

Nin8520/words

Training Setting

args = dict(
 stage="sft",                                               
 do_train=True,
 model_name_or_path="Qwen/Qwen2.5-7B-Instruct", 
 dataset="yue_1,yue_2",          
 template="alpaca",                                         
 finetuning_type="lora",                                    
 lora_target="all",                                         
 output_dir="Qwen2.5_lora",                                 
 per_device_train_batch_size=2,                             
 gradient_accumulation_steps=4,                             
 lr_scheduler_type="cosine",                                
 logging_steps=5,                                           
 warmup_ratio=0.1,                                          
 save_steps=1000,                                           
 learning_rate=5e-5,                                        
 num_train_epochs=1.0,                                      
 max_samples=300,                                           
 max_grad_norm=1.0,                                         
 loraplus_lr_ratio=16.0,                                    
 fp16=True,                                                 
 report_to="none",                                          
) 

Evaluation

Not done