Key Info
A retweeted thread from IlysMoutawwakil describes tooling to export, quantize, and run every transformers model end-to-end on any backend, with the same inputs and outputs as a drop-in replacement.
Highlights
- Covers the full pipeline: export, quantization, and execution
- Claims support for every transformers model
- Works across any backend
- Positioned as a drop-in replacement with identical inputs and outputs
- Links to documentation and release notes