Update README.md
Browse files
README.md
CHANGED
@@ -16,8 +16,10 @@ The Text-to-Text Transfer Transformer (T5) leveraged a unified text-to-text form
|
|
16 |
|
17 |
| | Link |
|
18 |
| -------- | :-----------------------: |
|
19 |
-
| **Small** | [**
|
20 |
-
| **Base** | [**
|
|
|
|
|
21 |
|
22 |
## How to use
|
23 |
|
@@ -101,6 +103,8 @@ python3 scripts/convert_t5_from_uer_to_huggingface.py --input_model_path cluecor
|
|
101 |
--type t5
|
102 |
```
|
103 |
|
|
|
|
|
104 |
### BibTeX entry and citation info
|
105 |
|
106 |
```
|
@@ -113,3 +117,5 @@ python3 scripts/convert_t5_from_uer_to_huggingface.py --input_model_path cluecor
|
|
113 |
}
|
114 |
```
|
115 |
|
|
|
|
|
|
16 |
|
17 |
| | Link |
|
18 |
| -------- | :-----------------------: |
|
19 |
+
| **Small** | [**Small**][small] |
|
20 |
+
| **Base** | [**Base**][base] |
|
21 |
+
|
22 |
+
In T5, spans of the input sequence are masked by so-called sentinel token. Each sentinel token represents a unique mask token for the input sequence and should start with <extra_id_0>, <extra_id_1>, … up to <extra_id_199>. However, <extra_id_xxx> is separated into multiple parts in Huggingface's Hosted inference API. Therefore, we replace <extra_id_xxx> with extraxxx in vocabulary and BertTokenizer regards extraxxx as one sentinel token.
|
23 |
|
24 |
## How to use
|
25 |
|
|
|
103 |
--type t5
|
104 |
```
|
105 |
|
106 |
+
Notice that
|
107 |
+
|
108 |
### BibTeX entry and citation info
|
109 |
|
110 |
```
|
|
|
117 |
}
|
118 |
```
|
119 |
|
120 |
+
[small]:https://huggingface.co/uer/t5-small-chinese-cluecorpussmall
|
121 |
+
[base]:https://huggingface.co/uer/t5-base-chinese-cluecorpussmall
|