Views
No views yet
.
├── scripts/
│ ├── train.py
│ ├── tokenization_qwen2.py
│ ├── requirements.txt
│ └── bootstrap.sh
├── sagemaker_train.py
└── README.mdQwen/Qwen2.5-7B-Instruct1{
2 'epochs': 3,
3 'per_device_train_batch_size': 4,
4 'gradient_accumulation_steps': 8,
5 'learning_rate': 1e-5,
6 'max_steps': 1000,
7 'bf16': True,
8 'max_length': 2048,
9 'gradient_checkpointing': True,
10 'optim': 'adamw_torch',
11 'lr_scheduler_type': 'cosine',
12 'warmup_ratio': 0.1,
13 'weight_decay': 0.01,
14 'max_grad_norm': 0.3
15}requirements.txt 생성bootstrap.sh 생성python sagemaker_train.pytokenization_qwen2.py)을 포함합니다: