arsaporta commited on
Commit
e12caad
1 Parent(s): 08fb740

Upload dataset (part 00003-of-00004)

Browse files
README.md CHANGED
@@ -38,6 +38,14 @@ configs:
38
  path: symile-m3-5-s/val-*
39
  - split: test
40
  path: symile-m3-5-s/test-*
 
 
 
 
 
 
 
 
41
  dataset_info:
42
  - config_name: symile-m3-5-m
43
  features:
@@ -95,6 +103,34 @@ dataset_info:
95
  num_examples: 50000
96
  download_size: 159628727029
97
  dataset_size: 156989775197.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
98
  ---
99
  # Dataset Card for Symile-M3
100
  Symile-M3 is a multilingual dataset of (audio, image, text) samples. The dataset is specifically designed to test a model's ability to capture higher-order information between three distinct high-dimensional data types: by incorporating multiple languages, we construct a task where text and audio are both needed to predict the image, and where, importantly, neither text nor audio alone would suffice.
 
38
  path: symile-m3-5-s/val-*
39
  - split: test
40
  path: symile-m3-5-s/test-*
41
+ - config_name: symile-m3-5-xs
42
+ data_files:
43
+ - split: train
44
+ path: symile-m3-5-xs/train-*
45
+ - split: val
46
+ path: symile-m3-5-xs/val-*
47
+ - split: test
48
+ path: symile-m3-5-xs/test-*
49
  dataset_info:
50
  - config_name: symile-m3-5-m
51
  features:
 
103
  num_examples: 50000
104
  download_size: 159628727029
105
  dataset_size: 156989775197.0
106
+ - config_name: symile-m3-5-xs
107
+ features:
108
+ - name: lang
109
+ dtype: string
110
+ - name: audio
111
+ dtype: audio
112
+ - name: image
113
+ dtype: image
114
+ - name: text
115
+ dtype: string
116
+ - name: cls
117
+ dtype: string
118
+ - name: cls_id
119
+ dtype: int64
120
+ - name: target_text
121
+ dtype: string
122
+ splits:
123
+ - name: train
124
+ num_bytes: 70410563197.0
125
+ num_examples: 500000
126
+ - name: val
127
+ num_bytes: 3607295872.0
128
+ num_examples: 25000
129
+ - name: test
130
+ num_bytes: 3624041386.0
131
+ num_examples: 25000
132
+ download_size: 80003029310
133
+ dataset_size: 77641900455.0
134
  ---
135
  # Dataset Card for Symile-M3
136
  Symile-M3 is a multilingual dataset of (audio, image, text) samples. The dataset is specifically designed to test a model's ability to capture higher-order information between three distinct high-dimensional data types: by incorporating multiple languages, we construct a task where text and audio are both needed to predict the image, and where, importantly, neither text nor audio alone would suffice.
symile-m3-5-xs/test-00001-of-00008.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70f5472710c4036e33ef4157725760381a3e43ded72382c6fe2b1376d2f6d263
3
+ size 444336449
symile-m3-5-xs/test-00002-of-00008.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ab16ac9147ceb99e75bde858eae31d9e64ce7202f8219a1a2f85349597a69c3
3
+ size 446590727
symile-m3-5-xs/test-00003-of-00008.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c0d912ded54a1a21487e5dc2cadfaa729e89e22656d7eaae69355bf1f0413e1
3
+ size 452114483
symile-m3-5-xs/test-00004-of-00008.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c29f2f66fa0ee7cfd8bd70bf4b06ba5dbf99150ea69728a531e9bb074c07bcb
3
+ size 474134401
symile-m3-5-xs/test-00005-of-00008.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a19baf0caae5a5dbd7fa602a041239734573a60f7022103b37f5b2a267d0635
3
+ size 461031005
symile-m3-5-xs/test-00006-of-00008.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b139516f448cc01047bfcd693982613a9971f5987a34fd5b19af9c30224a074
3
+ size 455049613
symile-m3-5-xs/test-00007-of-00008.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0675cffc2c5be74f21fe06c3642c3b6aaef648cf2e6a1c974a1c69f19f8eaeb
3
+ size 460316618