Note: This recipe is trained with the codes from this PR https://github.com/k2-fsa/icefall/pull/428
# Pre-trained Transducer-Stateless5 models for the TAL_CSASR dataset with icefall.
The model was trained on the far data of [TAL_CSASR](https://ai.100tal.com/dataset) with the scripts in [icefall](https://github.com/k2-fsa/icefall) based on the latest version k2.  
## Training procedure
The main repositories are list below, we will update the training and decoding scripts with the update of version.  
k2: https://github.com/k2-fsa/k2 
icefall: https://github.com/k2-fsa/icefall 
lhotse: https://github.com/lhotse-speech/lhotse
* Install k2 and lhotse, k2 installation guide refers to https://k2.readthedocs.io/en/latest/installation/index.html, lhotse refers to https://lhotse.readthedocs.io/en/latest/getting-started.html#installation. I think the latest version would be ok. And please also install the requirements listed in icefall.
* Clone icefall(https://github.com/k2-fsa/icefall) and check to the commit showed above.
```
git clone https://github.com/k2-fsa/icefall
cd icefall
```
* Preparing data.
```
cd egs/tal_csasr/ASR
bash ./prepare.sh
```
* Training
```
export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5"
./pruned_transducer_stateless5/train.py \
                  --world-size 6 \
                  --num-epochs 30 \
                  --start-epoch 1 \
                  --exp-dir pruned_transducer_stateless5/exp \
                  --lang-dir data/lang_char \
                  --max-duration 90
```
## Evaluation results
The decoding results (CER%) on TAL_CSASR(dev and test) are listed below:

|decoding-method | epoch(iter) | avg | dev | test |
|--|--|--|--|--|
|greedy_search | 30 | 24 | 7.49 | 7.58|
|modified_beam_search | 30 | 24 | 7.33 | 7.38|
|fast_beam_search | 30 | 24 | 7.32 | 7.42|
|greedy_search(use-averaged-model=True) | 30 | 24 | 7.30 | 7.39|
|modified_beam_search(use-averaged-model=True) | 30 | 24 | 7.15 | 7.22|
|fast_beam_search(use-averaged-model=True) | 30 | 24 | 7.18 | 7.27|
|greedy_search | 348000 | 30 | 7.46 | 7.54|
|modified_beam_search | 348000 | 30 | 7.24 | 7.36|
|fast_beam_search | 348000 | 30 | 7.25 | 7.39 |

The results (CER(%) and WER(%)) for Chinese CER and English WER respectivly (zh: Chinese, en: English):

|decoding-method | epoch(iter) | avg | dev | dev_zh | dev_en | test | test_zh | test_en |
|--|--|--|--|--|--|--|--|--|
|greedy_search(use-averaged-model=True) | 30 | 24 | 7.30 | 6.48 | 19.19 |7.39| 6.66 | 19.13|
|modified_beam_search(use-averaged-model=True) | 30 | 24 | 7.15 | 6.35 | 18.95 | 7.22| 6.50 | 18.70 |
|fast_beam_search(use-averaged-model=True) | 30 | 24 | 7.18 | 6.39| 18.90 |  7.27| 6.55 | 18.77|