aisingapore/Nemotron-SEA-LION-v4.8-30B-A3B-Base

Full fine-tune vs nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-Base-BF16. Fine-tune: 98% of the weights changed, across 52 of 52 layers.

Tensors changed: other 207/207, Embeddings 1/1, Norms 76/76, MLP 2689/2691, Attention 24/24, lm_head 0/1

1 derivatives checked