InfoCapability
PEFT v0.17.0: SHiRA, MiSS, LoRA for MoE
AI Impact Summary
New release v0.17.0 introduces SHiRA, MiSS, and LoRA support for Mixture of Experts models, expanding fine-tuning capabilities.
PEFT (Parameter-Efficient Fine-Tuning) library
Risk domains
775%665%
Source text
Release introducing SHiRA, MiSS, and LoRA support for Mixture of Experts (MoE) models. Extends fine-tuning capabilities for specialized model architectures.
- Date
- 1 Aug 2025
- Change type
- capability
- Severity
- info