Llama3-SFT
Collection
A series of fine-tuned models based on the Llama model • 5 items • Updated
模型：
数据集：
- https://huggingface.co/datasets/TigerResearch/sft_zh
- https://huggingface.co/datasets/silk-road/alpaca-data-gpt4-chinese
- https://huggingface.co/datasets/REILX/extracted_tagengo_gpt4
- https://huggingface.co/datasets/LooksJuicy/ruozhiba
- https://huggingface.co/datasets/leo009/alpaca-cleaned-zh-cn
（使用langid清理以上数据集，删除其中非中文资料）
训练工具
https://github.com/hiyouga/LLaMA-Factory
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- total_eval_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 3.0