kingoflolz / mesh-transformer-jax

Model parallel transformers in JAX and Haiku

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

How to stop model generating

jingrongchen opened this issue · comments

I have finetuned my model using my own dataset, but when i run the model, it generates a correct answer but the model keep on generating the prompt answer pair, is there any way to tell model to stop the genration after certain character

Hi, I am trying to finetune the model on my own dataset as well. But even with GPU memory of 24GB, I had issues loading the model weights into it. What would be the suggestion for the compute hardware for finetuning GPTJ 6B.

I am training with input data with an avg of 660 words.