Datasets:
medotter batch 10 ops (FF-C section 1): add an image_id column (the file stem) to the COph100 shard
Browse filesThe MedOtter loader reads a row's identity off the image_id column and, with it absent, falls back to the row's position in its hub split, while local mode names the same sample by its .jpg stem -- so the two modes could not be aligned row for row. image_id holds the value the mirror's own filename column already carries: the file stem, which is also what each row's mask.path (<stem>_mask.png) is built from.
Additive only. The eight existing columns keep their names, types, order and values; row order is unchanged; row groups, per-column compression, dictionary encoding and the datasets feature metadata are reproduced from the original shard.
data/test-00000-of-00001.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b8bfcc6fa9b5fcb455342c9f5b5452f451390e2c64eeec35b8dc32cf652987df
|
| 3 |
+
size 112485002
|