0dust's Blog

Weekly checkin #5

0dust
Published: 06/29/2020

What did i do this week? 

I finished adding QA models and the PR got merged.  With this i wrapped up adding models to DFFML which was the first part of my project.

What's next? 

I already started reading for my second part of the project which is adding NLP operations. I will be discussing with my mentors regarding different approaches to add these operations.

Did i get stuck somewhere?

No, it was smooth. There was a bug that creeped in and we realized that after PR was merged. So I made a small PR to fix that. 

View Blog Post

Weekly Check-in #4

0dust
Published: 06/22/2020

What did I do this week?

I added few more tests in classification model as per mentor's suggestion and got the PR merged, Yayy! I also finished writing the accuracy and predict methods for QA model and added the unit tests.

What's next?

I will be adding integration tests and example usage of QA model and thus will finish the first part of the project by this week!

Did I get stuck somewhere?

No, this week was pretty smooth. I had few confusions which were resolved in weekly meetings.

View Blog Post

Weekly Check-in #3

0dust
Published: 06/15/2020

What did I do this week?

I finished adding the classification models and opened the PR for review. I also worked on Question Answering model but soon realised the complexity of implementing it in Tensorflow. So I switched to Pytorch and made a good progress so far.

What's next?

I will be finishing the Accuracy and Predict methods for QA model and will open the PR for review by mentors. I will also be making changes, if any, in classification model as per the instruction of mentors.

Did I get stuck somewhere?

I got stuck while implementing accuracy method for QA model. Due to the complexity of SQuAD metric it was getting quite complex to implement it in tensorflow. Luckily its lot easier to implement in Pytorch so the problem was resolved once I switched to Pytorch.

 

 

View Blog Post

Weekly Check-in #2

0dust
Published: 06/08/2020

What did I do this week?

I worked on question answering and classification models using Hugging Face Transformers. Now, both question answering and classification models can be trained in DFFML. 

What's next?

I will be adding the Accuracy and Prediction method for question answering model and will wrap up the classification model to open pull request for review by mentors.

Did I get stuck somewhere?

Yes, I got stuck in reading the json file of SQuAD dataset for QA model. As you already know my mentor is super helpful so it got resolved quickly. 

 

View Blog Post

Weekly Check-in #1

0dust
Published: 06/02/2020

Hello there, I am Himanshu Tripathi, a final year CS undergrad from NIT Silchar. I have been contributing to DFFML from past few months and got selected to spend this summer (and lockdown :P) working on Adding Machine Learning models and Operations in DFFML

What did I do this week?

I worked on DataFlow Diagrams. Diagrams show the input and output flow of data among operations in DFFML using Mermaidjs. I added the feature to show the conditional dependence of one operation on the others if there exist any.

Ok, What's next?

I wil be working on adding Question Answering model in DFFML.

Did I get stuck somewhere?

Yes, I got stuck in few places, but we have meeting twice a week and my mentor is very active (Yayy!) so everything got resolved quickly.

View Blog Post