[ ] Understand how ctranslate can optimize inference process
Whisper dependencies
- ffmpeg CLI
sudo apt update && sudo apt install ffmpeg
- OpenAI's tiktoken => Use Rust if tiktoken does not work.
- ffmpeg-python
<message>
<body>
</body>
<langexbot xmlns="langex:chatbot">
<feature-element>
</langexbot>
</message>
- Onboard
<onboard xmlns="langex:chatbot:onboard"> </onboard>
- Pronunciation Assessment
<pronunc_assess xmlns="langex:chatbot:pronunc_assess"> </pronunc_assess>
To effectively build the image by not reinstalling all packages whenever requirements.txt
changed. You should execute sudo make config-dockerd
to config daemon.json
for docker daemon. Be careful that command will override all current configurations daemon.json
. If it is in case, you should manually append configuration in daemon.json
file in the current working directory to that system configuration daemon.json
file.
Model is installed at /root/.cache/huggingface
and /usr/local/src/lec-chatbot/chatbot/models/whisper/model
. Please mount it to avoid downloading model whenever run the container.
docker run -d --rm narutosimaha/lec-chatbot
[ ] Set up certificate for file service.