necrasov-ilya/AliceAI-T5-35B-A0.6B-MPS-ExpertInt8

Full fine-tune vs yandex/AliceAI-T5-35B-A0.6B. Fine-tune: 98% of the weights changed, across 16 of 16 layers.

Tensors changed: MLP 84/168, Embeddings 0/1, Norms 0/70, Attention 0/160, lm_head 0/1

On the Hub · Downloads on Model Pulse