Kijai/flux-fp8 · Discussions

#40 opened 8 months ago by

oddball516

dtype of flux1-dev-fp8-e5m2.safetensors is float8_e4m3fn ?

#39 opened 9 months ago by

pbouda-signature

Hyper-FLUX.1-dev-8steps-lora.safetensors

#37 opened 10 months ago by

LHJ0

FP8 for flux depth and flux canny.

#36 opened 11 months ago by

samurai-zero

HOW TO USE THE flux-fp8 IN PYTHON?

#35 opened 11 months ago by

KevinJMChen

Is your dev-fp8-e4m3fn version consistent with the official fp8 version released by Comfyui?

#33 opened about 1 year ago by

HUG-NAN

Which model is better for low vram/ram?

#32 opened about 1 year ago by

younyokel

Noisy Output When used in the Flux Inpainting Pipeline

#31 opened about 1 year ago by

VikramSingh178

cuda out of memory for flux1-dev-fp8.safetensors

#30 opened about 1 year ago by

pritam-tam

Silent exit when loading via FluxTransformer2DModel.from_single_file and no GPU use

#29 opened about 1 year ago by

nivgo

Inquiry on Latest Update for flux1-dev-fp8

5

#27 opened about 1 year ago by

torealise

fp8 inference

#26 opened about 1 year ago by

Melody32768

wrong model

#25 opened about 1 year ago by

sunhaha123

Update README.md

#24 opened about 1 year ago by

WBD8

"model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16" with ROCM6.0

5

#23 opened about 1 year ago by

12letter

Unet?

#22 opened about 1 year ago by

aiRabbit0

quite slow to load the fp8 model

👍 4

11

#21 opened about 1 year ago by

gpt3eth

RuntimeError: "addmm_cuda" not implemented for 'Float8_e4m3fn'

#20 opened about 1 year ago by

gradient-diffusion

How to load into VRAM?

#19 opened about 1 year ago by

MicahV

What setting to use for flux1-dev-fp8

#18 opened about 1 year ago by

fullsoftwares

'float8_e4m3fn' attribute error

6

#17 opened about 1 year ago by

Magenta6

Loading flux-fp8 with diffusers

#16 opened about 1 year ago by

8au

FP8 Checkpoint version size mismatch?

#15 opened about 1 year ago by

Thireus

Can this model be used on Apple Silicon?

👍 1

25

#14 opened about 1 year ago by

jsmidt

How to use fp8 models + original flux repo?

#13 opened over 1 year ago by

rolux

Quantization Method?

👍 2

10

#7 opened over 1 year ago by

vyralsurfer

ComfyUi Workflow

#6 opened over 1 year ago by

Jebari

Can you make FP8 version of schnell as well please?

#5 opened over 1 year ago by

MonsterMMORPG

Diffusers?

19

#4 opened over 1 year ago by

tintwotin

Minimum vram requirements?

#3 opened over 1 year ago by

joachimsallstrom

FP16

➕ 1