qnloft
/
ComfyUI
mirror of https://github.com/comfyanonymous/ComfyUI.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Make VAEDecodeTiled node work with video VAEs.

This commit is contained in:
comfyanonymous 2024-11-07 03:47:12 -05:00
parent 5e29e7a488
commit b49616f951
2 changed files with 27 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
comfy/sd.py
View File

@ -361,9 +361,24 @@ class VAE:
pixel_samples = pixel_samples.to(self.output_device).movedim(1,-1)
return pixel_samples
def decode_tiled(self, samples, tile_x=64, tile_y=64, overlap = 16):
def decode_tiled(self, samples, tile_x=None, tile_y=None, overlap=None):
model_management.load_model_gpu(self.patcher)
output = self.decode_tiled_(samples, tile_x, tile_y, overlap)
dims = samples.ndim - 2
args = {}
if tile_x is not None:
args["tile_x"] = tile_x
if tile_y is not None:
args["tile_y"] = tile_y
if overlap is not None:
args["overlap"] = overlap
if dims == 1:
args.pop("tile_y")
output = self.decode_tiled_1d(samples, **args)
elif dims == 2:
output = self.decode_tiled_(samples, **args)
elif dims == 3:
output = self.decode_tiled_3d(samples, **args)
return output.movedim(1, -1)
def encode(self, pixel_samples):

12
nodes.py
View File

@ -290,15 +290,21 @@ class VAEDecodeTiled:
@classmethod
def INPUT_TYPES(s):
return {"required": {"samples": ("LATENT", ), "vae": ("VAE", ),
"tile_size": ("INT", {"default": 512, "min": 320, "max": 4096, "step": 64})
"tile_size": ("INT", {"default": 512, "min": 128, "max": 4096, "step": 32}),
"overlap": ("INT", {"default": 64, "min": 0, "max": 4096, "step": 32}),
}}
RETURN_TYPES = ("IMAGE",)
FUNCTION = "decode"
CATEGORY = "_for_testing"
def decode(self, vae, samples, tile_size):
return (vae.decode_tiled(samples["samples"], tile_x=tile_size // 8, tile_y=tile_size // 8, ), )
def decode(self, vae, samples, tile_size, overlap):
if tile_size < overlap * 4:
overlap = tile_size // 4
images = vae.decode_tiled(samples["samples"], tile_x=tile_size // 8, tile_y=tile_size // 8, overlap=overlap // 8)
if len(images.shape) == 5: #Combine batches
images = images.reshape(-1, images.shape[-3], images.shape[-2], images.shape[-1])
return (images, )
class VAEEncode:
@classmethod