TT-XLA Documentation
TT-XLA is a PJRT-based backend integration that enables JAX and PyTorch/XLA to run on Tenstorrent AI hardware.
Getting Started
Testing
- Test Infra
-
Model Auto-Discovery Tests
- Overview
- Prerequisites
- Quick start / commonly used commands
- How discovery and parametrization work
- Test IDs and filtering
- Parallelism modes
- Per-model requirements
- Test configuration and statuses
- Model status and bringup_status guidance
- Targeting architectures
- Placeholder models (report-only)
- CI setup
- Adding a new model to run in Nightly CI
- Troubleshooting
- Future enhancements
- Reference
Compiler Internals
Code Generation