btbtyler09/shrew-ocr-preview

Full fine-tune vs ibm-granite/granite-vision-4.1-4b. Fine-tune: 90% of the weights changed, across 40 of 40 layers.

Tensors changed: other 64/66, Norms 45/273, MLP 120/232, Attention 336/508, Embeddings 0/3

1 derivatives checked

• btbtyler09/shrew-ocr-preview-GGUF: Same weights, converted (fewer layers)

On the Hub · Downloads on Model Pulse