Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
h
uint64
v
list
85,990,485,491,744
[ -0.00011920928955078125, -0.0191650390625, -0.021728515625, -0.01220703125, -0.000728607177734375, 0.057373046875, 0.049560546875, 0.027099609375, 0.0191650390625, 0.04833984375, -0.0216064453125, -0.0189208984375, -0.000335693359375, -0.035888671875, 0.060302734375, 0.0220947265625, ...
147,383,184,109,856
[ -0.00014400482177734375, -0.0091552734375, 0.040771484375, -0.0130615234375, -0.000690460205078125, 0.080078125, 0.060302734375, -0.00177001953125, 0.0201416015625, 0.0322265625, 0.01434326171875, 0.003021240234375, 0.00048065185546875, -0.033935546875, 0.037353515625, -0.0233154296875...
155,116,804,796,304
[ -0.00007104873657226562, 0.003082275390625, 0.068359375, 0.015380859375, 0, 0.043701171875, 0.051513671875, -0.0189208984375, 0.0255126953125, -0.00390625, 0.0172119140625, 0.045166015625, -0.00023365020751953125, -0.00457763671875, 0.00970458984375, -0.0517578125, -0.01275634765625,...
316,163,029,699,248
[ 0.00001055002212524414, 0.0322265625, 0.034423828125, -0.00543212890625, 0, 0.0283203125, 0.01190185546875, 0.0155029296875, 0.01416015625, -0.05078125, 0.016357421875, 0.029296875, -0.00567626953125, -0.03076171875, 0.026123046875, -0.041015625, 0.00518798828125, -0.04296875, 0....
458,302,078,578,848
[0.00017547607421875,0.0137939453125,0.046630859375,-0.0166015625,0.001129150390625,0.04443359375,0.(...TRUNCATED)
541,885,496,381,344
[-0.000125885009765625,-0.01153564453125,0.051513671875,0.003326416015625,-0.000286102294921875,0.05(...TRUNCATED)
724,350,665,651,648
[-0.0003223419189453125,0.03759765625,0.034423828125,0.0277099609375,-0.00171661376953125,0.08007812(...TRUNCATED)
751,295,181,797,808
[-0.00020122528076171875,-0.0216064453125,0.025146484375,-0.0216064453125,-0.0002574920654296875,0.0(...TRUNCATED)
792,320,047,963,360
[-0.0002593994140625,-0.00946044921875,-0.0196533203125,0.00026702880859375,-0.00136566162109375,0.0(...TRUNCATED)
810,894,516,574,928
[-0.0002841949462890625,-0.0096435546875,-0.007537841796875,-0.02734375,-0.00107574462890625,0.05102(...TRUNCATED)
End of preview. Expand in Data Studio

TabLake-Close-Emb

Frozen text-encoder vectors of every table string of TabLake-Close (WTQ, NQ-Tables, WikiSQL, FeTaQA, TabFact; train, dev and test tables together, each distinct string once). For research use only. This repository holds no source text.

  • <model>/bucket-BBBB.parquet: h (uint64, ascending) and v (fixed-size list of float16), one row per distinct string. A string's hash is blake2b(UTF-8 bytes, digest_size=8) read little-endian (0 is stored as 1); its bucket is hash % 16. Vectors have the model's own size, are L2-normalised, and come from whole strings (no truncation below the model's length limit). Files are zstd-compressed with byte-stream-split, so they read back bit for bit.

Strings are a table's cells as "header: value", its rows and columns as text, and its title. Sources: WikiTableQuestions, NQ-Tables, WikiSQL, FeTaQA and TabFact, each under its own licence.

Downloads last month
48

Collection including mtybilly/TabLake-Close-Emb