layonsan/google-t5-small
LoRA-style vs google-t5/t5-small. Looks like a merged LoRA (or a fine-tune with frozen norms): only the weight matrices of k, o, q, v changed, in 6 of 6 layers; norms, biases and embeddings are untouched.
Tensors changed: Attention 72/74, Norms 0/32, other 0/25