tejureddy01/ricegpt-v3

Partial fine-tune vs google/flan-t5-small. Partial change: 79% of the weights changed (attention, lm_head, norms, other), in 8 of 8 layers.

Tensors changed: Attention 98/98, Norms 42/42, other 48/49, lm_head 1/1

1 derivatives checked