# Run the demo First login and move to the directory: - `ssh ssh rrkarim@52.152.171.105` pass: Sevil@@@@2664 - `sudo su` - `docker exec -it 39c6a98a79a1 bash` - `cd egs/sevil/demo` There is a directory `./dir`, to run the demo: - Move your audio file `.ogg` or `.wav` to `./dir` - Move your content file to `./dir`. Be sure there is no punctuation, and unknown characters. - Run `bash prep_demo_ogg.sh` or `prep_demo.sh` depending on the audio extension. (TODO: just add condition to check it) - Then `cd ../` - And run `bash inference.sh` - See any errors that can occur, usually it is when the conent file badly encoded.
Last changed by  
 
Rasul Kerimov
89

Read more

To run the demo bot:

cd /opt/kaldi/egs/sevil/ conda activate base python ./demo/main.py There are 2 directories that are required to preprocess the data for training/inference: cd /opt/kaldi/egs/sevil /prepare contains the code to download the data from s3 and preprocess it. There is users.txt file which is the id of users that are extracted from the s3. ./data/local. After the preprocess is done we need to have corpus.txt file, which is the collection of all possible inputs of ASR. Also we need to update ./data/local/dict/lexicon.txt file with all the word tokens in corpus.txt.

1/16/2021
Final evaluation

Work that is merged PR: pymc4/#306 Issues: pymc4/#187, pymc4/#171 The main goal of the work during the summer was to provide the support for various samplers and sampling of discrete variables. We have written and pushed and an inteface that could easily expanded for various number of sampling algorithms. First, we have provided the base class for all the pymc4 sampling algorithms: class _BaseSampler(metaclass=abc.ABCMeta): def _sample(self, ...):

10/13/2020
Read more from Rasul Kerimov
Published on HackMD