CUA
Collection
5 items • Updated
CUA Complex Function, epoch 1
Full Megatron training checkpoint at iter_0000737 from the October 2026 Slides01 action-window Function run. This is the end of epoch 1 of 3. It includes model and optimizer shards, scheduler and RNG state, training state, run configuration, and tokenizer assets.
The run initialized its model weights from the Atomic Function 3x-full iter_0000422 export via a verified DGX import. Optimizer, scheduler, RNG, and step count were reset when Complex training began. The saved run configuration contains DGX-specific paths that must be adapted before resuming on another system.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support