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

Method forward

diffsynth/models/svd_unet.py:92–100  ·  view source on GitHub ↗
(self, timesteps)

Source from the content-addressed store, hash-verified

90 self.computation_device = computation_device
91
92 def forward(self, timesteps):
93 t_emb = get_timestep_embedding(
94 timesteps,
95 self.num_channels,
96 flip_sin_to_cos=self.flip_sin_to_cos,
97 downscale_freq_shift=self.downscale_freq_shift,
98 computation_device=self.computation_device,
99 )
100 return t_emb
101
102
103class TrainableTemporalTimesteps(torch.nn.Module):

Callers

nothing calls this directly

Calls 1

get_timestep_embeddingFunction · 0.70

Tested by

no test coverage detected