Find answers from the community

Updated 2 years ago

Huggingface

At a glance

The community members are discussing the status of the HuggingFaceLLMPredictor class in the GPT-Index library. They note that the class is mentioned in the changelog, but the migration guide link is broken. The community members are unsure whether they should use HuggingFaceLLM instead, but the import doesn't seem to work. Additionally, they are having issues with the temperature parameter, which appears to be an unexpected parameter in the latest version of the library. The comments suggest that a big release is coming soon, and the community members provide an example of using a Hugging Face LLM from the pinned version of the documentation.

Useful resources
Is HuggingFaceLLMPredictor deprecated? I can see it in change log. But the migration guide link is broken

Changelog : https://gpt-index.readthedocs.io/en/latest/development/changelog.html#new-features
Migration Guide (Broken): https://gpt-index.readthedocs.io/how_to/customization/llms_migration_guide.html

Should we use HuggingFaceLLM instead of HuggingFaceLLMPredictor? But the import doesnt seem to work
from llama_index.llms import HuggingFaceLLM

Also, all examples of HuggingFaceLLM, and HuggingFaceLLMPredictor pass temperature as param. But both of them in most recent version seem to say temperature is an unexpected param.

Any working example for initializing a HF LLM with temperature?
L
3 comments
Yea there is a big release coming today soon, the docs are updated but not the pypi package (hence the import error)
The old way of doing it will work for now, here's an example from the pinned version of the docs
Add a reply
Sign up and join the conversation on Discord