Unpruned W4A8 Diffusion Model Request
#27
by makisekurisu-jp - opened
Request to release an unpruned W4A8‑quantized version of the diffusion model.
https://huggingface.co/dummy9996/repro-temp/blob/main/minimax_music3_text_encoder_w4a8.safetensors
I'll delete the repo though because I will run out of storage someday, I'd like you to clone it
https://huggingface.co/dummy9996/repro-temp/blob/main/minimax_music3_text_encoder_w4a8.safetensors
I'll delete the repo though because I will run out of storage someday, I'd like you to clone it
Bro, you got it wrong. I’m talking about minimax_h3_fl2va_unpruned_w4a8_mixed.safetensors and minimax_h3_ref2va_unpruned_w4a8_mixed.safetensors. Kijai only provided the pruned W4A8‑mixed versions and didn’t release the unpruned ones. Really hope you can complete them. Thanks!