Edens-Gate/Hamanasu-QwQ-V1.5-Instruct-ckpts

Full fine-tune vs Delta-Vector/Hamanasu-32B-V1-QwQ. Fine-tune: 98% of the weights changed, across 64 of 64 layers.

Tensors changed: lm_head 1/1, Embeddings 0/1, Norms 126/129, MLP 192/192, Attention 448/448

1 derivatives checked

• mradermacher/Hamanasu-FFT-Instruct-GGUF: Quantization

On the Hub · Downloads on Model Pulse