Files
Llama3.2-3B-gsm8k-full-FT/finetune_config.json
ModelHub XC 1847568cd2 初始化项目,由ModelHub XC社区提供模型
Model: kmseong/Llama3.2-3B-gsm8k-full-FT
Source: Original Platform
2026-04-13 08:05:05 +08:00

19 lines
467 B
JSON

{
"base_model": "meta-llama/Llama-3.2-3B",
"fine_tuning_type": "Full Parameter Fine-tuning",
"dataset": "GSM8K",
"num_train_samples": 7473,
"batch_size": 2,
"grad_accum": 4,
"learning_rate": 1e-05,
"weight_decay": 0.01,
"warmup_ratio": 0.1,
"epochs": 3,
"max_length": 512,
"max_grad_norm": 1.0,
"lr_scheduler_type": "cosine",
"optimizer": "adamw_bnb_8bit",
"gradient_checkpointing": true,
"dtype": "bf16",
"trainer_type": "Trainer"
}