Stories by tosh

100 tokens/s, 4-bit Mistral 7B in MLX on M2 Ultra (faster than llama.cpp)

Aligner: Achieving Efficient Alignment Through Answer Correction

Amazon Bedrock

TSMC Evacuates Factory Areas After Taiwan Earthquake, Tsunami Warning

Dolphin-2.8-mistral-7B-v02

Building an efficient zero-shot classifier

SWE-agent: autonomously solving GitHub issues

xz has shown how a dependence on unpaid volunteers can cause major problems

Pyodide

The Dragon in My Garage

Model Stock: All we need is just a few fine-tuned models

Breaking Silos

Transforming the Web: The End of Silos

Instructor: Structured Data Like JSON from Large Language Models

Model Stock: All we need is just a few fine-tuned models

Mini-Jamba

Editing Models with Task Arithmetic (2022)

Slerp

Orcid and GitHub Sign Memorandum of Understanding

Instruction-Following Evaluation for Large Language Models