Martin Laprise
Martin Laprise
Using a notebook is ok, but it get out of hand pretty quickly. I don't have access to weights yet, but I'll probably build a small Streamlit app. I'll share...
With Hectiq Lab (https://lab.hectiq.ai), It's possible to deploy Streamlit apps on GPU (Nvidia T4).
yeah sorry about the confusion. Deploying a GPU always cost money (except on Colab). There is a free tier CPU only but I'm not sure it's a viable solution for...
The be more specific, you need to reduce the batch size and INCREASE the gradient accumulation correspondingly.
I tested the latest version of the transformers repo and it's indeed merged. However, the name of model/tokenizer has changed LLaMA -> Llama)