We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
add ability to freeze image transformer during training, as suggested… … by LiT paper (use a pretrained image net, unlock text encoder only)
allow for learned downsampling when projecting image embeddings to la… …tents, for FILIP
bug
image to text logits tensor should be transposed
fix FILIP variable text length masking, add ability to compose InfoLO… …OB, cite
first commit, with first attempt at FILIP