0.48.1

📅 Oct 2, 2025📦 bitsandbytesView on GitHub →

🐛 2 fixes🔧 1 symbols

Summary

Version 0.48.1 addresses a critical regression in LLM.int8() affecting inference with pre-quantized checkpoints and fixes an issue with 8bit parameter device movement.

🐛 Bug Fixes

Fixed a regression introduced in 0.48.0 that caused poor inference results with pre-quantized checkpoints in HF transformers when using LLM.int8().
Fixed regression in 8bit parameter device movement.

🔧 Affected Symbols

LLM.int8()