ghananlp / kasa Goto Github PK

The goal of this is to migrate the current translation app using an appropriate Azure service. (Functions for example)

Acceptance Criteria

1. Automatically scalable to serve a large number of requests
2. Feature Rate-limiting to prevent over-loading of the service
3. Quick response time
4. Input Validation and Error-handling
5. Authentication (Optional for now)

You will need to request for access to the current github repository to start working on this. Please contact me.

The preprocessing.py is functioning nd in the pypi.org repo. Please check and move it to the master branch. You can check by using

pip install Kasa

We are now able to create products and subscriptions via the Azure API management service which allows access to the API using a subscription key.

We now need to expand on this to create a portal where users can sign-up and access the various subscription packages to our translation API and make payments/ access a free trial.

Acceptance Criteria:

• The user should be able to sign up and login on to their account.
• User should be able to select a product and may payment through the Stripe payment provider.

Below are links to resources on how to implement this:

https://docs.microsoft.com/en-us/azure/api-management/monetization-support

https://azure.microsoft.com/en-gb/blog/how-to-monetize-apis-with-azure-api-management/

https://endjin.com/what-we-think/talks/monetize-apis-with-azure-api-management-using-stripe

https://github.com/microsoft/azure-api-management-monetization

https://github.com/microsoft/azure-api-management-monetization/blob/main/documentation/stripe-deploy.md

• create one-pager for how to contribute to library
• Prepare basic structure for kasa library taking into all possible sub classes
• Improvements to Kasa API - usability etc.

Hi,
Thank you for building and sharing this translation system.
I am trying out the eng_twi_transformer.ipynb using jupyter notebook with the provided data, jw300.en and jw300.tw.
Everything seems to work until I tried to execute the predict function (line 544).

Here are the results on the console:

Epoch 1 Batch 0 Loss 1.9547 Elapsed time 3.26s
Incompatible shapes: [1,74,128] vs. [73,128] [Op:AddV2]
Epoch 2 Batch 0 Loss 1.9929 Elapsed time 10.66s
Incompatible shapes: [1,74,128] vs. [73,128] [Op:AddV2]
...

I am not sure why this happens. Any help or advice is greatly appreciated.
Thank you.

Add the possibility for user to in put special characters for local languages, if they are not available on the keyboard.
The first objective is to implement this on the web app.

An example of this feature is shown on the following website:
https://www.webonary.org/buli/

we are using the test server for now to try out the changes on the streamlit app. once all confirmed, we will port it to the public application.

I could not open the following link: https://www.kaggle.com/azunre/jw300entw
Where can I find the JW300 en-twi data set?

Thanks!