Semasia
Collection
Large-scale dataset of latent embeddings from ~1700 pretrained vision models across 8 image benchmarks. Enables studying latent geometry, benchmarking • 9 items • Updated
Search is not available for this dataset
id uint32 | fine_label int64 | coarse_label int64 | model_name large_string | embedding list |
|---|---|---|---|---|
0 | 19 | 11 | aimv2_1b_patch14_224.apple_pt | [
-0.5018783807754517,
-0.07303106784820557,
0.5435730218887329,
-1.1702991724014282,
2.587554454803467,
0.7994844913482666,
-0.41242659091949463,
0.792163610458374,
-0.5119707584381104,
0.5177141427993774,
0.017276767641305923,
0.06964855641126633,
0.21180777251720428,
-0.45131662487983704,... |
1 | 29 | 15 | aimv2_1b_patch14_224.apple_pt | [
0.26695969700813293,
-0.6941990852355957,
1.59690260887146,
1.614107370376587,
1.1038377285003662,
2.2037172317504883,
-0.9704791307449341,
0.20185983180999756,
1.1171530485153198,
-0.23079699277877808,
-0.688133716583252,
-0.28796589374542236,
0.4706859886646271,
-0.7571161389350891,
0.... |
2 | 0 | 4 | aimv2_1b_patch14_224.apple_pt | [
1.923365592956543,
-1.0337409973144531,
-1.2523202896118164,
-0.3857957720756531,
1.2818078994750977,
0.7779320478439331,
-0.24341921508312225,
0.46973833441734314,
-0.44776394963264465,
0.33595147728919983,
-0.4142840504646301,
-0.2633814513683319,
0.36376553773880005,
0.2019135057926178,... |
3 | 11 | 14 | aimv2_1b_patch14_224.apple_pt | [
-1.1503233909606934,
-0.014015592634677887,
-0.8086898326873779,
0.3589625358581543,
2.8555240631103516,
2.1034114360809326,
0.5507277250289917,
1.2752879858016968,
0.7300180792808533,
0.16706523299217224,
-0.5981718897819519,
-1.1127309799194336,
0.35179024934768677,
-0.17244172096252441,... |
4 | 1 | 1 | aimv2_1b_patch14_224.apple_pt | [-1.9100857973098755,0.49201443791389465,1.059056282043457,-0.7058097720146179,2.5206727981567383,1.(...TRUNCATED) |
5 | 86 | 5 | aimv2_1b_patch14_224.apple_pt | [-1.4613734483718872,-0.2938721776008606,-0.019463688135147095,-0.03339255601167679,0.98457741737365(...TRUNCATED) |
6 | 90 | 18 | aimv2_1b_patch14_224.apple_pt | [-1.0961846113204956,-0.2855706214904785,-0.8166818022727966,0.4526106119155884,-0.15303364396095276(...TRUNCATED) |
7 | 28 | 3 | aimv2_1b_patch14_224.apple_pt | [2.216897964477539,-0.9013237953186035,-2.7437620162963867,-0.04815726354718208,-1.2815349102020264,(...TRUNCATED) |
8 | 23 | 10 | aimv2_1b_patch14_224.apple_pt | [-0.702642023563385,-0.45184558629989624,1.6008177995681763,0.45919546484947205,-1.3706309795379639,(...TRUNCATED) |
9 | 31 | 11 | aimv2_1b_patch14_224.apple_pt | [1.8118222951889038,-0.7348673343658447,2.269331216812134,0.12038184702396393,0.2440721094608307,1.5(...TRUNCATED) |
End of preview. Expand in Data Studio
Latents for cifar100 (timm)
This repository hosts precomputed embeddings for cifar100 across many timm models.
Each config corresponds to a single model;
only that model's Parquet files are read on load_dataset.
Usage
from datasets import load_dataset
ds_test = load_dataset("spaicom-lab/semasia-cifar100", "aimv2_1b_patch14_224.apple_pt", split="test")
ds_train = load_dataset("spaicom-lab/semasia-cifar100", "aimv2_1b_patch14_224.apple_pt", split="train")
Notes
- Configs are generated from what is actually uploaded on the Hub (parquet presence).
- Based on uoft-cs/cifar100
- Downloads last month
- 4,718