thoughtworks/Nemotron-3.5-30B-A3B-Antislop-FTPO

Full fine-tune vs nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16. Fine-tune: 100% of the weights changed, across 52 of 52 layers.

Tensors changed: other 210/210, Embeddings 1/1, Norms 81/81, MLP 2677/2679, Attention 28/28, lm_head 1/1

2 derivatives checked

• mradermacher/Nemotron-3.5-30B-A3B-Antislop-FTPO-GGUF: Quantization
• mradermacher/Nemotron-3.5-30B-A3B-Antislop-FTPO-i1-GGUF: Quantization

On the Hub · Downloads on Model Pulse