Nagendra Goel

C++ Inference capability

2

C++ inference capability will allow for easy backward integration with other Kaldi-2009-2019 code, and create a softer transition pathway.

With Threaded decoder, sometimes word alignment is not created

6

Sometimes (not always) when using threaded decoder, and creating both word and phone alignments, word alignment is not created although phone alignment is created. The plugin does not crash or...

It appears that a new connection is allowed in the time when client closes connection, but worker has not received EOS from decoder

13

It appears that the worker is make available and a new connection from a client is allowed when a previous client session has closed connection, but the worker is still...

continuous_decoding mode seems to miss alternate segments.

17

**Describe the bug** I decoded several files in continuous decoding mode and did a formal WER calculation. To my surprise, the pattern that I see is that half the segments...

lhotse kaldi import fails silently

2

Lhotse Kaldi import used to throw an error if there was something wrong in the data directory. Now, when I updated to the latest version, it seems to be failing...

Memory Hungry

67

scripts such as compute_fbank_musan.py consume 3.6GB of memory. This limits how many jobs can be run in parallel, despite the number of available CPUs. On another note - lhotse kaldi...

ElasticSearch consumes too much disk space

1

### Issues > *If you're having trouble with Articulate we definitely want to help you. If you can provide a little bit of information it will make our troubleshooting job...

sherpa/sherpa/bin/pruned_transducer_statelessX/offline_server.py goes OOM

8

offline_server.py goes OOM if the audio file is 30+ minutes on a 16GB GPU. No messge is printed on the screen, but when I hit a ^C on the server,...

sherpa/bin/streaming_pruned_transducer_statelessX/streaming_server.py throws errors under heavy load

3

I have seen the following types of errors: File "/mnt/dsk4/sherpa/./sherpa/bin/streaming_pruned_transducer_statelessX/streaming_server.py", line 499, in handle_connection_impl await socket.send(result) File "/home/ngoel/anaconda3/envs/k2/lib/python3.9/site-packages/websockets/legacy/protocol.py", line 620, in send await self.ensure_open() File "/home/ngoel/anaconda3/envs/k2/lib/python3.9/site-packages/websockets/legacy/protocol.py", line 930, in ensure_open...

Need word timestamps

1

Information on word timestamps or at least the length of audio processed "till now" in the streaming server, will be helpful, both for downstream audio segmenting / alignment tasks and...