Merge pull request #93 from ConvLab/readme

update nlu model link

convlab/base_models/t5/README.md

@@ -46,6 +46,10 @@ Trained models and their performance are available in [Hugging Face Hub](https:/
 | [t5-small-nlu-sgd](https://huggingface.co/ConvLab/t5-small-nlu-sgd) | NLU           | SGD                          |
 | [t5-small-nlu-tm1_tm2_tm3](https://huggingface.co/ConvLab/t5-small-nlu-tm1_tm2_tm3) | NLU           | TM1+TM2+TM3                  |
 | [t5-small-nlu-multiwoz21_sgd_tm1_tm2_tm3](https://huggingface.co/ConvLab/t5-small-nlu-multiwoz21_sgd_tm1_tm2_tm3) | NLU           | MultiWOZ 2.1+SGD+TM1+TM2+TM3 |
+| [t5-small-nlu-multiwoz21-context3](https://huggingface.co/ConvLab/t5-small-nlu-multiwoz21-context3) | NLU (context=3)          | MultiWOZ 2.1 |
+| [t5-small-nlu-tm1-context3](https://huggingface.co/ConvLab/t5-small-nlu-tm1-context3) | NLU (context=3)          | TM1 |
+| [t5-small-nlu-tm2-context3](https://huggingface.co/ConvLab/t5-small-nlu-tm2-context3) | NLU (context=3)          | TM2 |
+| [t5-small-nlu-tm3-context3](https://huggingface.co/ConvLab/t5-small-nlu-tm3-context3) | NLU (context=3)          | TM3 |
 | [t5-small-dst-multiwoz21](https://huggingface.co/ConvLab/t5-small-dst-multiwoz21) | DST           | MultiWOZ 2.1                 |
 | [t5-small-dst-sgd](https://huggingface.co/ConvLab/t5-small-dst-sgd) | DST           | SGD                          |
 | [t5-small-dst-tm1_tm2_tm3](https://huggingface.co/ConvLab/t5-small-dst-tm1_tm2_tm3) | DST           | TM1+TM2+TM3                  |

convlab/nlu/jointBERT/README.md

@@ -33,7 +33,7 @@ The result (`output.json`) will be saved under the `output_dir` of the config fi
 
 ## Performance on unified format datasets
 
-To illustrate that it is easy to use the model for any dataset that in our unified format, we report the performance on several datasets in our unified format. We follow `README.md` and config files in `unified_datasets/` to generate `predictions.json`, then evaluate it using `../evaluate_unified_datasets.py`. Note that we use almost the same hyper-parameters for different datasets, which may not be optimal.
+To illustrate that it is easy to use the model for any dataset that in our unified format, we report the performance on several datasets in our unified format. We follow `README.md` and config files in `unified_datasets/` to generate `predictions.json`, then evaluate it using `../evaluate_unified_datasets.py`. Note that we use almost the same hyper-parameters for different datasets, which may not be optimal. Trained models are available at [Hugging Face Hub](https://huggingface.co/ConvLab/bert-base-nlu).
 
 <table>
 <thead>

convlab/nlu/milu/README.md

@@ -45,7 +45,7 @@ See `nlu.py` under `multiwoz` and `unified_datasets` directories.
 
 ## Performance on unified format datasets
 
-To illustrate that it is easy to use the model for any dataset that in our unified format, we report the performance on several datasets in our unified format. We follow `README.md` and config files in `unified_datasets/` to generate `predictions.json`, then evaluate it using `../evaluate_unified_datasets.py`. Note that we use almost the same hyper-parameters for different datasets, which may not be optimal.
+To illustrate that it is easy to use the model for any dataset that in our unified format, we report the performance on several datasets in our unified format. We follow `README.md` and config files in `unified_datasets/` to generate `predictions.json`, then evaluate it using `../evaluate_unified_datasets.py`. Note that we use almost the same hyper-parameters for different datasets, which may not be optimal.  Trained models are available at [Hugging Face Hub](https://huggingface.co/ConvLab/milu).
 
 <table>
 <thead>