[Model] FLUX.1 Full Finetuned & Merged
Collection
This is *not* a collection of FLUX LoRA fine-tuned models, it collects FLUX.1-dev full fine-tuned models. • 5 items • Updated
import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("minpeter/FLUX-juaner-ghibli-v2-diffusers", dtype=torch.bfloat16, device_map="cuda")
prompt = "a elephant."
image = pipe(prompt).images[0]This model is a conversion of the original model from the fp8 single file format to the Diffusers standalone format.
The script used for the conversion, cast from fp8 to bf16 for convenience:
from diffusers import FluxPipeline, FluxTransformer2DModel
import torch
transformer = (
FluxTransformer2DModel.from_single_file(
"./jgs-v2-fp8.safetensors",
torch_dtype=torch.float8_e4m3fn,
)
.to("cuda")
.to(torch.bfloat16)
)
pipe = FluxPipeline.from_pretrained(
"black-forest-labs/FLUX.1-dev", transformer=transformer, torch_dtype=torch.bfloat16
)
pipe.save_pretrained("./output-model")
Base model
black-forest-labs/FLUX.1-dev