diff --git a/examples/contextnet/README.md b/examples/contextnet/README.md index f788a67dbb..3619707a5d 100644 --- a/examples/contextnet/README.md +++ b/examples/contextnet/README.md @@ -17,3 +17,31 @@ Training, see `python examples/contextnet/train_*.py --help` Testing, see `python examples/contextnet/test_*.py --help` TFLite Conversion, see `python examples/contextnet/tflite_*.py --help` + +## RNN Transducer Subwords - Results on LibriSpeech + +**Summary** + +- Number of subwords: 1008 +- Maximum length of a subword: 10 +- Subwords corpus: all training sets +- Number of parameters: 12,075,320 +- Number of epochs: 86 +- Train on: 8 Google Colab TPUs +- Train hours: 8.375 days uncontinuous (each day I trained 12 epoch because colab only allows 12 hours/day and 1 epoch required 1 hour) => 86 hours continuous (3.58333333 days) + +**Pretrained and Config**, go to [drive](https://drive.google.com/drive/folders/1fzOkwKaOcMUMD9BAjcLLmSG2Tfpeabbq?usp=sharing) + +**Epoch Transducer Loss** + +subword_contextnet_loss + +**Epoch Learning Rate** + +epoch_learning_rate + +**Error Rates** + +| **Test-clean** | Test batch size | Epoch | WER (%) | CER (%) | +| :------------: | :-------------: | :---: | :----------------: | :----------------: | +| _Greedy_ | 1 | 86 | 10.356436669826508 | 5.8370333164930344 | \ No newline at end of file diff --git a/examples/contextnet/figs/1008_epoch_learning_rate.svg b/examples/contextnet/figs/1008_epoch_learning_rate.svg new file mode 100644 index 0000000000..f1c16b7273 --- /dev/null +++ b/examples/contextnet/figs/1008_epoch_learning_rate.svg @@ -0,0 +1 @@ +3e-54e-55e-56e-57e-58e-59e-51e-42e-4-20020406080100120 \ No newline at end of file diff --git a/examples/contextnet/figs/1008_subword_contextnet_loss.svg b/examples/contextnet/figs/1008_subword_contextnet_loss.svg new file mode 100644 index 0000000000..f1c5ca2799 --- /dev/null +++ b/examples/contextnet/figs/1008_subword_contextnet_loss.svg @@ -0,0 +1 @@ +-202468101214161820-100102030405060708090 \ No newline at end of file