unalignment/Pixtral-12B-Captioner-Relaxed

Full fine-tune vs mistralai/Pixtral-12B-2409. Fine-tune: 95% of the weights changed, across 40 of 40 layers.

Tensors changed: Attention 160/160, Norms 0/81, MLP 120/120, lm_head 1/1, Embeddings 0/1

1 derivatives checked

• noctrex/Pixtral-12B-Captioner-Relaxed-GGUF: Quantization (fewer layers)

1 other repos hold byte-for-byte identical weight files.

On the Hub · Downloads on Model Pulse