HopitAI/moda-ner-v-crop

Partial fine-tune (fewer layers) vs HopitAI/moda-fashion-distilled. HopitAI/moda-fashion-distilled matches part of HopitAI/moda-ner-v-crop (312 of its 479 tensors; the rest of HopitAI/moda-ner-v-crop has no counterpart). On the shared part: Partial change: 14% of the weights changed (attention, mlp, norms), in 4 of 12 layers.

Tensors changed: Norms 16/102, Embeddings 0/5, other 0/2, Attention 16/107, MLP 16/96

On the Hub · Downloads on Model Pulse