cara-ai/ALM2Vec-FT
Partial fine-tune vs mispeech/midashenglm-7b-0804-fp32. Partial change: 5% of the weights changed (other), in 0 of 32 layers.
Tensors changed: Attention 0/128, MLP 0/128, Norms 0/130, Embeddings 0/2, other 4/4
Partial fine-tune vs mispeech/midashenglm-7b-0804-fp32. Partial change: 5% of the weights changed (other), in 0 of 32 layers.
Tensors changed: Attention 0/128, MLP 0/128, Norms 0/130, Embeddings 0/2, other 4/4