Alex Cheema
Alex Cheema
Corrected a small error in the notebook **Description** - Describe your changes. Fixed a small grammar mistake. **Motivation and Context** - Why is this change required? What problem does it...
In the vits example (also used in other examples like conversation), the download is blocked for `ljs` and `vctk` because Google redirects to a page warning about virus scan. https://github.com/tinygrad/tinygrad/blob/df44a4e861ff25d9552f82d1164aaa36c3d4abd4/examples/vits.py#L662-L663...
- I want to keep exo 100% python if possible - Would like to compile swift or objc inference code in tinygrad - The deliverable here is a merged PR...
- Windows is largely untested. Don't know of anyone who has it working natively, only in WSL - Known issues on windows: https://github.com/exo-explore/exo/issues/184 , tinygrad not officially supported - Perhaps...
Why JAX? Read this: https://neel04.github.io/my-website/blog/pytorch_rant/ The deliverable is a `JAXInferenceEngine` that can run Llama.
Mistral just dropped a new vision multimodal model called Pixtral 12b. We already have support for Llava, another VLM. The task is to add support for Pixtral.
- it should automatically detect the best device to run on - We should require 0 manual configuration from the user, by default llama.cpp for example requires specifying the device
Currently you can only run one exo instance on each device. There are some design decisions here: - Should we support running multiple exo instances on the same device, with...
Some preliminary work here: https://github.com/exo-explore/exo/pull/195