why does this include the full t5 fp16 model alongside the svdq-t5.safetensors

#4
by tazztone - opened

this includes:
model-00001-of-00002.safetensors
model-00002-of-00002.safetensors
which is t5 fp16?

this is not the same file concept as with "transformer_blocks" and "unquantized_layers" found in "svdq-int4-flux.1"?

MIT HAN Lab org

Will exclude this.

Sign up or log in to comment