wren11ws/sunup

Partial fine-tune vs distilbert/distilgpt2. Partial change: 65% of the weights changed (attention, mlp, norms), in 4 of 6 layers.

Tensors changed: Attention 16/24, Norms 18/26, MLP 16/24, Embeddings 0/1

On the Hub · Downloads on Model Pulse