yxli2123 / LoftQ

MIT License
180 stars 15 forks source link

[BUG]size mismatch for base_model.model.model.embed_tokens.weight #29

Closed Rain-yj closed 2 months ago

Rain-yj commented 2 months ago

Hello! I'm facing a similar issue as well. adapter can't use on base model after training, and the error always shows size mismatch info from PeftModelForCausalLM I am using LoftQ/Llama-2-7b-hf-4bit-64rank, Is there any solution to this? Thank you! 微信图片_20240424095327