MCPcopy Create free account
hub / github.com/MotrixLab/ViMoGen / WanMotionTensorDatasetWiRefMotionM2M

Class WanMotionTensorDatasetWiRefMotionM2M

datasets/video_datasets.py:207–218  ·  view source on GitHub ↗

Dataset variant for motion-to-motion refinement where text embeddings are unused. Reuses all logic from WanMotionTensorDatasetWiRefMotion but always feeds the null text context instead of loading precomputed embeddings.

Source from the content-addressed store, hash-verified

205
206
207class WanMotionTensorDatasetWiRefMotionM2M(WanMotionTensorDatasetWiRefMotion):
208 """
209 Dataset variant for motion-to-motion refinement where text embeddings are unused.
210 Reuses all logic from WanMotionTensorDatasetWiRefMotion but always feeds the
211 null text context instead of loading precomputed embeddings.
212 """
213
214 def __init__(self, *args, **kwargs):
215 super().__init__(*args, **kwargs)
216
217 def _resolve_prompt_embedding(self, data):
218 return self.null_context.clone()
219
220
221class MBenchWiRefMotion(torch.utils.data.Dataset):

Callers

nothing calls this directly

Calls

no outgoing calls

Tested by

no test coverage detected