Closed ethancohen123 closed 4 months ago
need it too
We will not be releasing the large model (ViT-L/8+) weights at this time. However, we have partnered with Nvidia to host a smaller but more flexible version of the foundation model, which we call Phenom-Beta. It achieves about 0.56 recall on CORUM (better than ViT-L imagenet). Users can hit their API to get embeddings directly, which handles the auto-scaling and inference logic:
Are you planning to release the model (the big one ) weights with some inference code ?