-
Notifications
You must be signed in to change notification settings - Fork 27
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add vocab feature #124
Add vocab feature #124
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Did you test this with a few keywords? How were the results?
For example, on this YouTube video: https://youtu.be/v2X51AVgl3o I added some vocab: Here are the results with the vocab on ( + Right now, open-source contributions are being used as the new resume.
- Right now, open source contributions are being used as the new resume.
+ In this video, we will be discussing what is open-source contributions and how do you actually do that.
- In this video, we will be discussing what is open source contributions and how do you actually do that.
+ The next place where you can find these projects is GitHub.
- The next place where you can find these projects is GitHub.
+ For example, if you're really good at Python programming language and want to contribute.
- For example, if you are really good at Python programming language and want to contribute.
+ Now open this folder in your visual studio code and open the readme.md file.
- Now open this folder in your Visual Studio code and open the readme.md file. It's not perfect, for example We could add an extra |
And do fuzzy match? That could go wrong in unforseen ways. What about if you try prepending the terms with "Make sure these words are spelled correctly: " |
It strictly doesn't change anything on the sample test I use. |
Ok, let's stick to the initial method of splitting by comma, since the prompt is limited to a certain number of tokens I believe. |
There could be a simpler way than fuzzy match to post-process maybe - just look for the exact words after lowercasing (and then replacing symbols with spaces) in the custom vocab. For example:
wdyt? |
…hub.com/Wordcab/wordcab-transcribe into 123-use-prompting-for-custom-vocabulary
* add multi gpus handling for transcription * add model index for transcription and diarization * add gpu_index for alignment models * fix diarization gpu indexing * multi gpu setup, with transcription errors * Updated error payload for svix in cortex endpoint * Extra languages performed poorly, commenting out tests that require "he" for this param * update the download_audio function to avoid extension * add audio_duration key in repsonse + fix dual_channel bug * fix tests and endpoint returns * Add a catch for empty audio (#128) * add a catch for empty audio file * rename utterances -> response for coherence * fix quality * Add vocab feature (#124) * add vocab feature * fix youtube endpoint * update the prompt sentence * add vocab feature * fix youtube endpoint * update the prompt sentence * Upgraded base docker image to nvidia/cuda:11.7.1-cudnn8-runtime-ubuntu20.04 * add multi gpus handling for transcription * add model index for transcription and diarization * add gpu_index for alignment models * fix diarization gpu indexing * multi gpu setup, with transcription errors * lower batch_size * fix alignment device index * revert transcribe service to no mapping * update gpu service queue manager * fix Exception returns for endpoints * fixed dual_channel * fix flake and darglint * run black linter * fix nemo config tests * fix typo --------- Co-authored-by: Aleks <aleks@wordcab.com> Co-authored-by: Aleksandr Smechov <35517862+aleksandr-smechov@users.noreply.github.com>
Add a simple way for the user to add some extra vocab in the payload. These words will be concatenated into a single string and provided to the model during inference.