The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Add cheap latent preview for flux 2. (#10907)
comfyanonymousf16219e3aadc
diff --git a/comfy/latent_formats.py b/comfy/latent_formats.pyindex e98c7d6d8..8e110f45d 100644--- a/comfy/latent_formats.py+++ b/comfy/latent_formats.py@@ -6,6 +6,7 @@ class LatentFormat:latent_dimensions = 2latent_rgb_factors = Nonelatent_rgb_factors_bias = None+ latent_rgb_factors_reshape = Nonetaesd_decoder_name = Nonedef process_in(self, latent):@@ -181,6 +182,45 @@ class Flux(SD3):class Flux2(LatentFormat):latent_channels = 128+ def __init__(self):+ self.latent_rgb_factors =[+ [0.0058, 0.0113, 0.0073],+ [0.0495, 0.0443, 0.0836],+ [-0.0099, 0.0096, 0.0644],+ [0.2144, 0.3009, 0.3652],+ [0.0166, -0.0039, -0.0054],+ [0.0157, 0.0103, -0.0160],+ [-0.0398, 0.0902, -0.0235],+ [-0.0052, 0.0095, 0.0109],+ [-0.3527, -0.2712, -0.1666],+ [-0.0301, -0.0356, -0.0180],+ [-0.0107, 0.0078, 0.0013],+ [0.0746, 0.0090, -0.0941],+ [0.0156, 0.0169, 0.0070],+ [-0.0034, -0.0040, -0.0114],+ [0.0032, 0.0181, 0.0080],+ [-0.0939, -0.0008, 0.0186],+ [0.0018, 0.0043, 0.0104],+ [0.0284, 0.0056, -0.0127],+ [-0.0024, -0.0022, -0.0030],+ [0.1207, -0.0026, 0.0065],+ [0.0128, 0.0101, 0.0142],+ [0.0137, -0.0072, -0.0007],+ [0.0095, 0.0092, -0.0059],+ [0.0000, -0.0077, -0.0049],+ [-0.0465, -0.0204, -0.0312],+ [0.0095, 0.0012, -0.0066],+ [0.0290, -0.0034, 0.0025],+ [0.0220, 0.0169, -0.0048],+ [-0.0332, -0.0457, -0.0468],+ [-0.0085, 0.0389, 0.0609],+ [-0.0076, 0.0003, -0.0043],+ [-0.0111, -0.0460, -0.0614],+ ]++ self.latent_rgb_factors_bias = [-0.0329, -0.0718, -0.0851]+ self.latent_rgb_factors_reshape = lambda t: t.reshape(t.shape[0], 32, 2, 2, t.shape[-2], t.shape[-1]).permute(0, 1, 4, 2, 5, 3).reshape(t.shape[0], 32, t.shape[-2] * 2, t.shape[-1] * 2)+def process_in(self, latent):return latentdiff --git a/latent_preview.py b/latent_preview.pyindex 95d3cb733..ddf6dcf49 100644--- a/latent_preview.py+++ b/latent_preview.py@@ -37,13 +37,16 @@ class TAESDPreviewerImpl(LatentPreviewer):class Latent2RGBPreviewer(LatentPreviewer):- def __init__(self, latent_rgb_factors, latent_rgb_factors_bias=None):+ def __init__(self, latent_rgb_factors, latent_rgb_factors_bias=None, latent_rgb_factors_reshape=None):self.latent_rgb_factors = torch.tensor(latent_rgb_factors, device="cpu").transpose(0, 1)self.latent_rgb_factors_bias = Noneif latent_rgb_factors_bias is not None:self.latent_rgb_factors_bias = torch.tensor(latent_rgb_factors_bias, device="cpu")+ self.latent_rgb_factors_reshape = latent_rgb_factors_reshapedef decode_latent_to_preview(self, x0):+ if self.latent_rgb_factors_reshape is not None:+ x0 = self.latent_rgb_factors_reshape(x0)self.latent_rgb_factors = self.latent_rgb_factors.to(dtype=x0.dtype, device=x0.device)if self.latent_rgb_factors_bias is not None:self.latent_rgb_factors_bias = self.latent_rgb_factors_bias.to(dtype=x0.dtype, device=x0.device)@@ -85,7 +88,7 @@ def get_previewer(device, latent_format):if previewer is None:if latent_format.latent_rgb_factors is not None:- previewer = Latent2RGBPreviewer(latent_format.latent_rgb_factors, latent_format.latent_rgb_factors_bias)+ previewer = Latent2RGBPreviewer(latent_format.latent_rgb_factors, latent_format.latent_rgb_factors_bias, latent_format.latent_rgb_factors_reshape)return previewerdef prepare_callback(model, steps, x0_output_dict=None):