search

tammypi / llama-finetune-total

7 stars 1 forks source link

There's a small problem in the README #1

Open jxtse opened 8 months ago

jxtse commented 8 months ago

3.1.转换模型原始权重 将模型的原始权重，转换为PyTorch的格式。 cd /root/autodl-tmp/Finetune_LLAMA/LLAMA_Model python convert_llama_weights_to_hf.py --input_dir /root/autodl-tmp/llama/models --model_size 7B --output_dir /root/autodl-tmp/models cd /root/autodl-tmp/lit-lamma python scripts/convert_hf_checkpoint.py --checkpoint_dir /root/autodl-tmp/models --model_size 7B

lamma should be llama

tensorflowt commented 1 month ago

请问是单卡微调嘛？

tammypi commented 1 month ago

7B模型可以4090单卡微调的。现在推荐用llamafactory微调了，更方便。