aminLo/best-grade-model-2
Adapter vs google/flan-t5-base. aminLo/best-grade-model-2 is a LoRA adapter, not full weights: rank 64, on k, o, q, v, wi_0, wi_1, wo in 12 layers (216 adapted matrices). It modifies google/flan-t5-base at load time.
Adapter vs google/flan-t5-base. aminLo/best-grade-model-2 is a LoRA adapter, not full weights: rank 64, on k, o, q, v, wi_0, wi_1, wo in 12 layers (216 adapted matrices). It modifies google/flan-t5-base at load time.