Skip to main content
GitHub Repo starsGitHub License
Felafax Banner Felafax is a framework for continued-training and fine-tuning open source LLMs using XLA runtime. We take care of necessary runtime setup and provide a Jupyter notebook out-of-box to just get started.
  • Easy to use.
  • Easy to configure all aspects of training (designed for ML researchers and hackers).
  • Easy to scale training from a single TPU VM with 8 cores to entire TPU Pod containing 6000 TPU cores (1000X)!
Our goal at felafax is to build infra to make it easier to run AI workloads on non-NVIDIA hardware (TPU, AWS Trainium, AMD GPUs, and Intel GPUs).

​
✨ Finetune for Free

Add your dataset, click “Run All”, and you’ll run on free TPU resource on Google Colab!

Llama 3.1 (1B, 3B)

Start for free on Google Colab TPU

​
Currently Supported Models

​
LLaMa-3.1 JAX Implementation

Our flagship implementation offering maximum performance and scalability across hardware platforms.

Core Features

• Ported to JAX from PyTorch • Supported Models: 1B, 3B, 8B, 70B and 405B • Training Modes: Full-precision, LoRA

Key Benefits

• Run efficiently across wide range of hardware: TPUs, AWS Trainium, NVIDIA, and AMD • Scale seamlessly across multiple accelerators • Hardware-optimized XLA backend

​
LLaMa-3/3.1 PyTorch XLA

Meta’s PyTorch implementation with XLA support for TPU compatibility.

Features

• Architecture: PyTorch XLA• Training Modes: Full-precision, LoRA• Codepointer