Torch Stack

Important

Torch Stack is under active development

Torch Stack explores training, tuning, aligning, and quantizing Llama models with PyTorch native frameworks.

PyTorch Native Libraries and Frameworks

Note

PyTorch native libaries and frameworks support development on a variety of GPU types, including NVIDIA, Apple, and AMD GPUs.

• Training: Torch Titan
• Tuning: Torch Tune
• Aligning: Torch RL
• Quantization: Torch AO