Alexandre Marques

icon indicating an email [email protected]

icon company Neural Magic icon location Somerville, Massachusetts USA icon location Scientist specialized in numerical models and computational algorithms. Focus on Machine Learning.

Results 5 comments of Alexandre Marques

Let modifiers implement epoch adjustment when composing recipes

I like the idea of defining the advance_epochs method in the modifier class. However, note that the new method does not try to update all attributes w/ "epoch" in the...

Let modifiers implement epoch adjustment when composing recipes

I like the new approach, but shouldn't the standard advance_epochs in modifier.py reproduce the old behavior to be safe?

Add support to total batch size argument for transformers transfer learning

This PR is ready but needs to be merged by hand. Will create a new PR soon and delete this one when done.

Add support to total batch size argument for transformers transfer learning

Too much time has passed. Will re-implement if needed

Research: 4-bit quantization

Hi @truenorth8. Thanks for your interest in contributing to deepsparse. There's actually been plenty of research into quantizing LLMs to 4 bits. Probably some of the most notorious examples are...