facebookresearch / facebookresearch/large_concept_model
training procedure of base-lcm
Open
- Dominant language
- Python
- Stars
- 2.4k
- Forks
- 213
- PR merge metrics
- No merged PRs in 30d
Description
hi,
really thanks for your impressive work !
just curious about if the training of base-lcm will contain all the steps below ?
def normalize(self, embeddings: Tensor) -> Tensor:
if self.config.with_fft:
embeddings = self.fft_transform(embeddings)
embeddings = (embeddings - self.center) / self.scale
if self.config.clip_proba is not None:
embeddings = torch.clamp(embeddings, min=self.clip_min, max=self.clip_max)
return embeddings
Contributor guide
Assessment
This issue has not been assessed yet.