MCPcopy Create free account
hub / github.com/MiniMax-AI/VTP / forward_clip

Method forward_clip

vtp/models/vtp.py:340–360  ·  view source on GitHub ↗
(
        self,
        image: Optional[torch.Tensor],
        text: Optional[torch.Tensor]
    )

Source from the content-addressed store, hash-verified

338 return self.forward_reconstruction(reconstruction_image)
339
340 def forward_clip(
341 self,
342 image: Optional[torch.Tensor],
343 text: Optional[torch.Tensor]
344 ):
345 image_features = self.encode_image(image, normalize=True) if image is not None else None
346 text_features = self.encode_text(text, normalize=True) if text is not None else None
347
348 if self.output_dict:
349 out_dict = {
350 "image_features": image_features,
351 "text_features": text_features,
352 "logit_scale": self.logit_scale.exp()
353 }
354 if self.logit_bias is not None:
355 out_dict['logit_bias'] = self.logit_bias
356 return out_dict
357
358 if self.logit_bias is not None:
359 return image_features, text_features, self.logit_scale.exp(), self.logit_bias
360 return image_features, text_features, self.logit_scale.exp()
361
362 def forward_reconstruction(self, reconstruction_image: torch.Tensor):
363 return self.get_reconstruction_outputs(reconstruction_image)

Callers 1

forwardMethod · 0.95

Calls 2

encode_imageMethod · 0.95
encode_textMethod · 0.95

Tested by

no test coverage detected