Datasets:
Configure SAVE train, validation and test viewer splits
Browse filesMap each split explicitly to its existing scene manifest, excluding READY.json from dataset-row loading. Dataset payloads and split membership remain unchanged.
README.md
CHANGED
|
@@ -6,6 +6,15 @@ tags:
|
|
| 6 |
- speech-separation
|
| 7 |
- sound-source-localization
|
| 8 |
- multichannel
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
---
|
| 10 |
# SAVE: paper split
|
| 11 |
|
|
|
|
| 6 |
- speech-separation
|
| 7 |
- sound-source-localization
|
| 8 |
- multichannel
|
| 9 |
+
configs:
|
| 10 |
+
- config_name: default
|
| 11 |
+
data_files:
|
| 12 |
+
- split: train
|
| 13 |
+
path: manifests/tr.jsonl
|
| 14 |
+
- split: validation
|
| 15 |
+
path: manifests/cv.jsonl
|
| 16 |
+
- split: test
|
| 17 |
+
path: manifests/tt.jsonl
|
| 18 |
---
|
| 19 |
# SAVE: paper split
|
| 20 |
|