anlp-a2-part2-muon
Dense decoder-only transformer pretrained from scratch on browndw/human-ai-parallel-corpus
(1x the training split) with a from-scratch muon optimizer.
- params: 17,011,584
- trained on 41,680,896 tokens
- W&B run: https://wandb.ai/dunkronit-iiit-hyderabad/anlp-a2-part2/runs/muon-dd551d9d
Load with src.part2.model.Transformer.from_pretrained("DunkRonit/anlp-a2-part2-muon") from the assignment repo.
- Downloads last month
- 20
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support