Add OpenAI / 'import clip' variant

A CLIP ViT-L/14, resilient to typographic attacks - thanks to k_proj orthogonalization and attention head dropout during fine-tuning (+ some adversarial training)!

ViT-L-14-KO-LITE-FULL-model-OpenAI-format.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d13af6ca385ad86bcbb09bb3e91dd1ac9cb919b696ceb5144f683f60aab9cac
+size 1710517724