Flux Blockwise
I had to build several custom tools to allow for the mixed precision model, to my knowledge it is the first built like this.
Faster and more accurate then any other FP8 quantized model currently available
Works in Comfy and Forge but forge needs to be set to BF16 UNET
Comfy load as a diffuser model USE DEFAULT WEIGHT
Despite being larger the model works faster when used properly.
FP16 Upcasting should not be used unless absolutely necessary such as running CPU or IPEX
FORGE - set COMMANDLINE_ARGS= --unet-in-bf16 --vae-in-fp32
Other then the need to force forge into BF16, (FP32 VAE optionally) it should work the same as the DEV model with the added benefit of being 5GB smaller then the full BF16
Every FLUX model that is not BF16 should be mixed precision.
This is similar to the scaled FP8 E4M3 and E5M2 scaled T5_xxl only even more critical.
It turns out that every quantized model including my own up to this point to my knowledge has been built UN-optimally per blackforest.
Only the UNET blocks should be quantized in the diffuser model, also they should be upcast to BF16 and not FP16 (Comfy does this correctly)
I am currently trying to workout how to follow Blackforest recommendations but using GGUF
The blocks that should be BF16 and never quantized
All Text Blocks with (Txt)
'time_in.in_layer.bias',
'time_in.in_layer.weight',
'time_in.out_layer.bias',
'time_in.out_layer.weight',
'txt_in.bias',
'txt_in.weight',
'vector_in.in_layer.bias',
'vector_in.in_layer.weight',
'vector_in.out_layer.bias',
'vector_in.out_layer.weight'