fireworks-ai
/

FLUX.1-dev-fp8-flumina

Model card Files Files and versions Community

FLUX.1-dev-fp8-flumina

7 contributors

History: 54 commits

flowpoint's picture

add benchmarks numbers for rtx4000ada (non-sff)

d316f04 4 months ago

configs
Adding device specific configs & more input image type options + small model spec from args change 5 months ago
modules
Improved precision / reduced frequency of nan outputs, allow bf16 t5, f32 rmsnorm, larger clamp 4 months ago
.gitignore

125 Bytes

Small fixes & clean up 5 months ago
LICENSE

11.3 kB

Create LICENSE 4 months ago
README.md

20.7 kB

add benchmarks numbers for rtx4000ada (non-sff) 4 months ago
api.py

907 Bytes

Remove unnecessary synchronize, add more universal seeding & limit if run on windows 5 months ago
float8_quantize.py

18.7 kB

remove torchao dependency, quantize entirely via linear 4 months ago
flux_emphasis.py

14.1 kB

Remove unnecessary tokenization (still needs work) 5 months ago
flux_pipeline.py

25.8 kB

Add lora loading 5 months ago
image_encoder.py

1.07 kB

Adding device specific configs & more input image type options + small model spec from args change 5 months ago
lora_loading.py

16 kB

Fix issue where lora alpha is not correct if lora from transformers checkpoint 4 months ago
main.py

6.29 kB

Add quantize embedders/modulation to argparse options 5 months ago
main_gr.py

4.28 kB

Small fixes & clean up 5 months ago
requirements.txt

202 Bytes

remove torchao dependency, quantize entirely via linear 4 months ago
util.py
10.6 kB

Improved precision / reduced frequency of nan outputs, allow bf16 t5, f32 rmsnorm, larger clamp 4 months ago