search

conceptofmind / PaLM

An open-source implementation of Google's PaLM models
MIT License
813 stars 77 forks source link

how to train the model,can you explain more detail #6

Open qiuqingpo opened 1 year ago

qiuqingpo commented 1 year ago

i want to train my model，can you guide me?

conceptofmind commented 1 year ago

i want to train my model，can you guide me?

I will be releasing a fine-tuning script once the 3B model completes. Showing how the models can be trained on a new dataset.

zarandioon commented 1 year ago

@conceptofmind Any ETA for this? Have you tried to fine-tune the smaller models to make sure their quality is on-par with or better than the models with similar sizes?