32 lines
954 B
Python
32 lines
954 B
Python
import torch
|
|
|
|
class CustomScheduler:
|
|
@classmethod
|
|
def INPUT_TYPES(s):
|
|
inputs = {
|
|
"required": {
|
|
"steps": ("INT", {"default": 4, "min": 1, "max": 25}),
|
|
},
|
|
"optional": {},
|
|
}
|
|
|
|
defaults = [4.12, 1.62, 0.7, 0.04]
|
|
for i in range(0, 26):
|
|
default = defaults[i] if i < len(defaults) else 0.0
|
|
inputs["optional"][f"sigma_{i}"] = ("FLOAT", {"default": default, "min": 0.0, "max": 100.0, "step": 0.01, "round": 0.0001})
|
|
|
|
return inputs
|
|
|
|
RETURN_TYPES = ("SIGMAS",)
|
|
CATEGORY = "sampling/custom_sampling/schedulers"
|
|
FUNCTION = "get_sigmas"
|
|
|
|
def get_sigmas(self, steps, **kwargs):
|
|
sigmas = []
|
|
for i in range(0, steps + 1):
|
|
sigma = kwargs[f"sigma_{i}"]
|
|
sigmas.append(sigma)
|
|
|
|
sigmas = torch.FloatTensor(sigmas).cpu()
|
|
return (sigmas,)
|