aariciah/gpt2-turkish-dutch-routed-seed123

Partial fine-tune vs aariciah/gpt2-turkish-20k-lc. Partial change: 50% of the weights changed (attention, mlp, norms), in 6 of 12 layers.

Tensors changed: Attention 12/48, Norms 12/50, MLP 12/48, Embeddings 0/1

On the Hub · Downloads on Model Pulse