Mozilla AI developers have released a new version of Llamafile, their solution for easy-to-use LLMs as a single file that work across hardware and operating systems. With Llamafile 0.10.4 is now Transcribefile, as a new piece built off their recently announced Transcribe.cpp project.
At the end of June, Mozilla AI announced the official release of Transcribe.cpp. This new Transcribe.cpp is a C/C++ speech-to-text inference library and aims to be similar to Llama.cpp for speech-to-text models. Transcribe.cpp builds on GGML and can be accelerated with the likes of Vulkan, NVIDIA CUDA, and Apple Metal. Now with Llamafile 0.10.4 it provides Transcribefile for Transcribe.cpp in single file builds and supporting 16+ TTS models.
Llamafile 0.10.4 also updates against its Llama.cpp upstream build, brings a few improvements to its Vulkan API and AMD ROCm acceleration handling, HTTPS download support, and pledge/SECCOMP sandboxing support.
Downloads and more details on the new Llamafile 0.10.4 release via GitHub.
Originally Appeared Here
