-
We use RNNT to train librispeech dataset, results are as follows：
| encoder type | test-clean wer |
| ---- | ----|
| 4 layer blstm | 5.0 |
| 8 layer lstm | 12.0 |
The config of 4 layer b…
-
##### System information (version)
- OpenCV => 4.2
- Operating System / Platform => Windows 64 Bit
- Compiler =>Visual Studio 2015
##### Detailed description
##### Steps to reprod…
-
I have a trained e2e ASR transducer model that trained on fbank features (Librispeech). I'm trying to concatenate auxiliary vector to the fbank input during testing. The idea is to feed the model cont…
-
- [x] ag-news-lstm
- [x] ag-news-single-lstm
- [x] ag-news
- [x] conll-bert
- [x] conll-cnn
- [x] ~~conll-elmo-hub~~
- [x] conll-elmo
- [x] conll-no-crf
- [x] conll-single-tf
- [x] conll-sing…
-
Hi,
How would you build a bidirectional RNN with Flux? Drawing from a few examples
1. [Bidirectional LSTM example A](https://github.com/FluxML/Flux.jl/issues/866)
2. [Bidirectional LSTM example…
-
In `slicemap(f, A; dims)` `f` is restricted to `Function` type. This prevents from passing to it `Flux` layers. Can we fix this?
-
为什么用双层blstm，单层blstm不会收敛？
-
请问一下，BLSTM_CRF的参数设置是什么？
-
-
Hi, I am trying to make a Dockerfile and refactor directory structures.
In my branch [docker](https://github.com/alex-ht/CAT/tree/docker), I have finished a [dockerfile](https://github.com/alex-ht…