The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Fix some issues with sampling precision.
comfyanonymous111f1b525526
diff --git a/comfy/model_base.py b/comfy/model_base.pyindex 84cf9829d..37a52debf 100644--- a/comfy/model_base.py+++ b/comfy/model_base.py@@ -44,7 +44,7 @@ class ModelSamplingDiscrete(torch.nn.Module):else:betas = make_beta_schedule(beta_schedule, timesteps, linear_start=linear_start, linear_end=linear_end, cosine_s=cosine_s)alphas = 1. - betas- alphas_cumprod = np.cumprod(alphas, axis=0)+ alphas_cumprod = torch.tensor(np.cumprod(alphas, axis=0), dtype=torch.float32)# alphas_cumprod_prev = np.append(1., alphas_cumprod[:-1])timesteps, = betas.shape@@ -56,7 +56,7 @@ class ModelSamplingDiscrete(torch.nn.Module):# self.register_buffer('alphas_cumprod', torch.tensor(alphas_cumprod, dtype=torch.float32))# self.register_buffer('alphas_cumprod_prev', torch.tensor(alphas_cumprod_prev, dtype=torch.float32))- sigmas = torch.tensor(((1 - alphas_cumprod) / alphas_cumprod) ** 0.5, dtype=torch.float32)+ sigmas = ((1 - alphas_cumprod) / alphas_cumprod) ** 0.5self.register_buffer('sigmas', sigmas)self.register_buffer('log_sigmas', sigmas.log())diff --git a/comfy/samplers.py b/comfy/samplers.pyindex a74c8a1b8..518b666db 100644--- a/comfy/samplers.py+++ b/comfy/samplers.py@@ -137,10 +137,10 @@ def sampling_function(model_function, x, timestep, uncond, cond, cond_scale, moddef calc_cond_uncond_batch(model_function, cond, uncond, x_in, timestep, max_total_area, model_options):out_cond = torch.zeros_like(x_in)- out_count = torch.ones_like(x_in)/100000.0+ out_count = torch.zeros_like(x_in)out_uncond = torch.zeros_like(x_in)- out_uncond_count = torch.ones_like(x_in)/100000.0+ out_uncond_count = torch.zeros_like(x_in)COND = 0UNCOND = 1@@ -241,6 +241,8 @@ def sampling_function(model_function, x, timestep, uncond, cond, cond_scale, modout_uncond /= out_uncond_countdel out_uncond_count+ torch.nan_to_num(out_cond, nan=0.0, posinf=0.0, neginf=0.0, out=out_cond) #in case out_count or out_uncond_count had some zeros+ torch.nan_to_num(out_uncond, nan=0.0, posinf=0.0, neginf=0.0, out=out_uncond)return out_cond, out_uncond