Currently my preferred LLM: FLAN-T5. Watch my code optimization and examples.
Released Nov 2022 - it is an enhanced version of T5. Great for few-shot learning.
(By the way, I had to provide my mail and telephone number to OpenAI for a preview test of ChatGPT, only during the research preview, usage of ChatGPT is free. Plus: ChatGPT is currently not available on HuggingFace. And you know ... proprietary AI ...)
We directly use /code FLAN-T5 for inference from Huggingface, without further fine-tuning the models weights.
All rights with:
https://huggingface.co/docs/transform...
https://arxiv.org/pdf/2210.11416.pdf
#machinelearning
#ai
#datascience