AGI-Eval/Auto-ATT

Partial fine-tune vs Qwen/Qwen2-Audio-7B-Instruct. Partial change: 77% of the weights changed (attention, mlp, other), in 32 of 32 layers.

Tensors changed: other 1/6, Embeddings 0/2, Norms 0/195, MLP 96/224, Attention 128/448, lm_head 0/1

On the Hub · Downloads on Model Pulse