This is the official developer portal for MN-Core. You'll find code examples and sample projects on GitHub, and the full API reference in the MLSDK Documentation.

We'll be posting release notes, updates, and announcements here going forward — stay tuned.

For questions or feedback, feel free to reach out via the Contact page.

What's New in v0.6

• Added a MN-Core 2 (GPFN3) emulator and assembler to MLSDK, with low-level debugging guidance in the technical notes
• Introduced a new fx2onnx.linter.lint API and documentation to detect Python constructs incompatible with MLSDK
• Added a new MLSDK advanced example stable_diffusion covering end-to-end fine-tuning and inference
• MNCL: integrated new MN-Core 2 device builtins
• Removed the inference-only run_stable_diffusion example, now replaced by the stable_diffusion advanced example
• Performance improvements and bug fixes across MN-Core SDK

What's New in v0.5

• New "Hardware Specification" section added to MLSDK documentation
• New advanced examples: detr_inference and ncf_training
• Base PyTorch version updated from 2.6.0 to 2.9.0
• Base CUDA version updated from 12.6.0 to 12.8.0
• Performance improvements and bug fixes

What's New

• Performance improvements and bug fixes in the compiler
• Dependency library version upgrades
• New MLSDK migration guide added to the documentation
• Sample MLSDK projects included at /opt/pfn/pfcomp/codegen/MLSDK/examples/