Incorrect brownian parameters

[gremlation]

Expected Behavior

No warnings.

Actual Behavior

It seems incorrect parameters are being passed into torchsde.BrownianTree here:

ComfyUI/comfy/k_diffusion/sampling.py

Line 110 in cba58ff

def __call__(self, t0, t1):

I receive a warning in the terminal stating:

UserWarning: Should have tb<=t1 but got tb=14.614644050598145 and t1=14.614643.

This is the relevant code:

def __call__(self, t0, t1):
    t0, t1, sign = self.sort(t0, t1)
    if self.cpu_tree:
        w = torch.stack([tree(t0.cpu().float(), t1.cpu().float()).to(t0.dtype).to(t0.device) for tree in self.trees]) * (self.sign * sign)
    else:
        w = torch.stack([tree(t0, t1) for tree in self.trees]) * (self.sign * sign)

    return w if self.batched else w[0]

At the point the warning is generated, these values are present:

t0 = tensor(8.3726, device='mps:0')
t1 = tensor(14.6146, device='mps:0')
sign = -1
self.trees = [BrownianTree(interval=BrownianInterval(t0=0.029, t1=14.615, size=(1, 4, 144, 112), dtype=torch.float32, device=device(type='cpu'), entropy=0, dt=None, tol=1e-06, pool_size=24, cache_size=45, levy_area_approximation='none'))]

Steps to Reproduce

Generate an image using dpmpp_sde or dpmpp_sde_gpu on macOS / Apple silicon.

Debug Logs

See above.

Other

This is with master, currently at cba58fff0bfebfc81fbe678bb80491890a3df14a.

[gremlation]

Also logged as an issue upstream:

google-research/torchsde#152