MCPcopy Create free account
hub / github.com/OpenImagingLab/FlashVSR / single_decode

Method single_decode

diffsynth/models/stepvideo_vae.py:1025–1030  ·  view source on GitHub ↗
(self, hidden_states, device)

Source from the content-addressed store, hash-verified

1023 return x
1024
1025 def single_decode(self, hidden_states, device):
1026 chunks = list(hidden_states.split(self.latent_len, dim=1))
1027 for i in range(len(chunks)):
1028 chunks[i] = self.decode_naive(chunks[i].to(device), True).permute(0,2,1,3,4).cpu()
1029 x = torch.cat(chunks, dim=1)
1030 return x
1031
1032 def build_1d_mask(self, length, left_bound, right_bound, border_width):
1033 x = torch.ones((length,))

Callers 1

decodeMethod · 0.95

Calls 3

decode_naiveMethod · 0.95
cpuMethod · 0.80
toMethod · 0.45

Tested by

no test coverage detected