grimjim/Daichi-Instructed-12B

Full fine-tune vs Delta-Vector/Daichi-12B. Fine-tune: 97% of the weights changed, across 48 of 48 layers.

Tensors changed: Norms 190/400, MLP 144/252, Attention 192/408, Embeddings 1/4, other 1/1

2 derivatives checked

• mradermacher/Daichi-Instructed-12B-GGUF: Quantization (fewer layers)
• mradermacher/Daichi-Instructed-12B-i1-GGUF: Quantization (fewer layers)

On the Hub · Downloads on Model Pulse