Dataset Viewer
Auto-converted to Parquet Duplicate
text
stringlengths
9
31
old new 1.58
smart intelligent 9.2
hard difficult 8.77
happy cheerful 9.55
hard easy 0.95
fast rapid 8.75
happy glad 9.17
short long 1.23
stupid dumb 9.58
weird strange 8.93
wide narrow 1.03
bad awful 8.42
easy difficult 0.58
bad terrible 7.78
hard simple 1.38
smart dumb 0.55
insane crazy 9.57
happy mad 0.95
large huge 9.47
hard tough 8.05
new fresh 6.83
sharp dull 0.6
quick rapid 9.7
dumb foolish 6.67
wonderful terrific 8.63
strange odd 9.02
happy angry 1.28
narrow broad 1.18
simple easy 9.4
old fresh 0.87
apparent obvious 8.47
inexpensive cheap 8.72
nice generous 5
weird normal 0.72
weird odd 9.2
bad immoral 7.62
sad funny 0.95
wonderful great 8.05
guilty ashamed 6.38
beautiful wonderful 6.5
confident sure 8.27
dumb dense 7.27
large big 9.55
nice cruel 0.67
impatient anxious 6.03
big broad 6.73
strong proud 3.17
unnecessary necessary 0.63
restless young 1.6
dumb intelligent 0.75
bad great 0.35
difficult simple 0.87
necessary important 7.37
bad terrific 0.65
mad glad 1.45
honest guilty 1.18
easy tough 0.52
easy flexible 4.1
certain sure 8.42
essential necessary 8.97
different normal 1.08
sly clever 7.25
crucial important 8.82
harsh cruel 8.18
childish foolish 5.5
scarce rare 9.17
friendly generous 5.9
fragile frigid 2.38
long narrow 3.57
big heavy 6.18
rough frigid 2.47
bizarre strange 9.37
illegal immoral 4.28
bad guilty 4.2
modern ancient 0.73
new ancient 0.23
dull funny 0.55
happy young 2
easy big 1.12
great awful 1.17
tiny huge 0.6
polite proper 7.63
modest ashamed 2.65
exotic rare 8.05
dumb clever 1.17
delightful wonderful 8.65
noticeable obvious 8.48
afraid anxious 5.07
formal proper 8.02
dreary dull 8.25
delightful cheerful 6.58
unhappy mad 5.95
sad terrible 5.4
sick crazy 3.57
violent angry 6.98
laden heavy 5.9
dirty cheap 1.6
elastic flexible 7.78
hard dense 5.9
recent new 7.05
End of preview. Expand in Data Studio

AI Residency — blog toy data

Small slices of the datasets used by the nine systems in RoshBeed/ai-residency, cut down so the toy models in the posts on roshbeed.com train in seconds on a GitHub Actions runner.

Every post pins a commit revision of this dataset rather than tracking main, so a figure on the site cannot change because something here did.

path what it is source
text8/text8-2m.txt first 2,000,000 characters of text8 roshbeed/ai-residency-text8
mnist/mnist-small.npz 6,000 train / 1,500 test MNIST digits, uint8 28×28 MNIST
hn/hn-sample.json.gz 20,000 train / 5,000 test Hacker News posts, split by time roshbeed/ai-residency-hn-posts
audio/clip-16k.npz one 1.06 s clip, 16 kHz mono float32 the whisper-fine-tuning demo clip

These are for illustration. The real training runs use the full corpora; see each service's README in the residency repo for the numbers that came out of them.

Downloads last month
247