A decentralized and distributed horizontally scalable model execution framework.
Project description
scattermind
Scattermind is a decentralized and distributed horizontally scalable model execution framework. Worried your model won't fit in memory? Scattermind splits the model into small independently executable sub-models to accomodate even small machines (*).
(*) Currently only available through plugins. Once the plugins have matured they will be included in the main distribution.
TODOs
A rough outline of planned features:
- node based caching
- record / replay inputs
- record / replay activity
- log granular activity
- split nodes
- benchmark
- bert model
- diffusion model
- llama2 model
- push notification on result
- computation timing estimates
- self configuration
- config hotloading
- model training
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
scattermind-0.2.0.tar.gz
(95.6 kB
view hashes)
Built Distribution
Close
Hashes for scattermind-0.2.0-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 7e76783c6cac2fde386a808bdb596d39be46f1e67e3cb9e9ad0055396e1b4399 |
|
MD5 | 2582c2fba1e9f261cbc799c811c1ecce |
|
BLAKE2b-256 | 086677207bd1518cde40482a13395fd48d366dc97e272bef5f3cb318a970252b |