Skip to content

Conversation

timBoML
Copy link

@timBoML timBoML commented Nov 15, 2024

Hello Byaldi Team!

Description

I added BitsAndBytes support for all us GPU-poor people. This enables 4-bit/8-bit quantization to run the models on smaller GPUs or, in my case, leave space for a bigger LLM.

Changes Made

  • Added BitsAndBytes quantization options to model loading
  • Updated dependencies to include bitsandbytes
  • Added quant_strategy in the example notebook

Testing

  • I am using Byaldi in a commercial setting and the 4-bit quantization didn't affect performance.

@timBoML timBoML closed this Nov 15, 2024
@timBoML
Copy link
Author

timBoML commented Nov 15, 2024

I found a typo

@timBoML timBoML reopened this Nov 15, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant