Views
No views yet
1r=8,
2lora_alpha=32,
3lora_dropout=0.1,
4task_type="CAUSAL_LM",
5target_modules="all-linear",
6bias="none",
7use_dora=True,
8use_rslora=True1per_device_train_batch_size=28,
2gradient_accumulation_steps=1,
3num_train_epochs=3,
4warmup_ratio=0.1,
5learning_rate=2e-5,
6bf16=True,
7save_strategy="steps",
8save_steps=500,
9save_total_limit=10,
10logging_steps=10,
11output_dir=log_output,
12optim="paged_adamw_8bit",
13gradient_checkpointing=True