Dataset Viewer
Auto-converted to Parquet Duplicate
name
string
images
int64
expected
int64
duplicates_across_parts
int64
id_sha256_matches_manifest
bool
parts
dict
shards
int64
bytes
int64
sidecar_fingerprint
string
contents
string
problems
list
certified
bool
image 10 full training cache
80,000
80,000
0
true
{ "0": { "images": 9688, "shards": 19, "bytes": 5049394798, "ledger_sha256": "9ae223b70347b830d1d2e2924a68f477a1d5b72572322bd5707d558a3cedee29", "complete": true, "set_aside_mismatch": 0 }, "1": { "images": 10001, "shards": 20, "bytes": 5212034821, "ledger_sha256": "8f40860...
161
41,709,084,724
bfef99625a8c6437c669baf37126801ee082daf759ee0637b8934e58bb425d54
per image: sidecar feature 1280, embedding 256, logits 33, dense 1280x12x12 (fp16) + image10 physics tile features 1048, availability, coordinates, tile offsets, 11,528-d image summary, 12-d label vector, per-tile truth fraction, 12x12 recipe mask
[]
true

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

image 10 full training cache: image10 sidecar (with 12x12 recipe masks) x 80K physics cache, joined per image. 80,000 images, certified. See p*/ledger.jsonl.

Downloads last month
158