Uploaded the training and test splits.

Note, they only contain the Clip embeddings as we train our model in the clip space. They also contain the metainfo as 'content_idx' and 'content_description'. We are working on a solution to host the original images.

test.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6d6b1338a2382adc31f6a19f289a5305de0f51dcd597b0d1c514b5f0bb5c0c8
+size 337096482

train.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:888068e06ec07ceae45de7a5363edac10508ac9dd6668b506a734b950353be96
+size 770775315