ANE optimised adaptation of Whisper for optimised inference
Project description
This project is an attempt to re-define certain architectures to be optimised to run on the ANE. More specifically, it aims to adopt this Apple's ml-ane-transformers
codebase to the Whisper architecture. I've been working with the commit hash d18e9ea
(there have been some changes to return signatures of internal functions -- that may not impact us, but I haven't tested that yet)
Install
pip install git+https://github.com/openai/whisper/.git@d18e9ea5dd2ca57c697e8e55f9e654f06ede25d0
pip install git+https://github.com/apple/ml-ane-transformers
pip install whisper_ane
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
whisper_ane-0.1.0.tar.gz
(1.7 kB
view hashes)
Built Distribution
Close
Hashes for whisper_ane-0.1.0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 758bccb6a53de2128652a66719be0d9e5249ed7df1e9717d554054fa6a9c5a5a |
|
MD5 | d455af73930fb80d593a241db2af9077 |
|
BLAKE2b-256 | 08f9cb2d56b0f4c005b640064e2dabd541b8cffdc652663fa28513507967cf04 |