Leyo commited on
Commit
08e31d6
1 Parent(s): 66a38fb

Fix table layout

Browse files
Files changed (1) hide show
  1. README.md +6 -6
README.md CHANGED
@@ -113,12 +113,12 @@ We closel follow the training procedure layed out in [Flamingo](https://arxiv.or
113
 
114
  The model is trained on the following data mixture of openly accessible data:
115
 
116
- | Data Source | Type of Data | Number of Tokens in Source | Number of Images in Source | Epochs | Effective Proportion in Number of Tokens |
117
- |-------------|----------------------------|------------|----------------------------|--------|
118
- | PMD | Image-Text Pairs | TODO | TODO | 3 | 73.85% |
119
- | LAION | Image-Text Pairs | TODO | TODO | 1 | 6.15% |
120
- | OBELISC | Unstructured Multimodal Web Documents | TODO | TODO | 3 | 2.82% |
121
- | Wikipedia | Unstructured Multimodal Web Documents | TODO | TODO | 1 | 17.18% |
122
 
123
  For multimodal web documents, we feed the model sequences corresponding to the succession of text paragraphs and images.
124
  For image-text pairs, we form the training sequences by packing images with their captions.
 
113
 
114
  The model is trained on the following data mixture of openly accessible data:
115
 
116
+ | Data Source | Type of Data | Number of Tokens in Source | Number of Images in Source | Epochs | Effective Proportion in Number of Tokens |
117
+ |-------------|-----------------------------------------|---------------------------|---------------------------|--------|-----------------------------------------|
118
+ | PMD | Image-Text Pairs | TODO | TODO | 3 | 73.85% |
119
+ | LAION | Image-Text Pairs | TODO | TODO | 1 | 6.15% |
120
+ | OBELISC | Unstructured Multimodal Web Documents | TODO | TODO | 3 | 2.82% |
121
+ | Wikipedia | Unstructured Multimodal Web Documents | TODO | TODO | 1 | 17.18% |
122
 
123
  For multimodal web documents, we feed the model sequences corresponding to the succession of text paragraphs and images.
124
  For image-text pairs, we form the training sequences by packing images with their captions.