Search found 105 matches

by lazyprogrammer
Thu Jan 26, 2023 9:13 pm
Forum: Data Science: Transformers for Natural Language Processing
Topic: Lecture 33: Parameters befor and after training
Replies: 1
Views: 130

Re: Lecture 33: Parameters befor and after training

Thanks for your inquiry. The variable 'newmodel' is a pipeline object. It's not the same object type as the previous model. To understand how the parameters stored in the 'model' variable could have changed, you should study "object-oriented programming" to give you some idea of what's hap...
by lazyprogrammer
Mon Jan 23, 2023 8:41 pm
Forum: Time Series Analysis, Forecasting, and Machine Learning
Topic: Improving GARCH using deep neural network.
Replies: 1
Views: 127

Re: Improving GARCH using deep neural network.

Thanks for your inquiry.

This is related to one of the VIP projects in the PyTorch course. We use a neural network for uncertainty estimation (uncertainty being the variance, hence exactly what GARCH is doing).
by lazyprogrammer
Sat Dec 17, 2022 2:58 am
Forum: General Discussion
Topic: How can I become a great Deep Learning programmer?
Replies: 2
Views: 299

Re: How can I become a great Deep Learning programmer?

Thanks for your question. I would recommend the following 2 things: 1) Look at the "course order" list (https://deeplearningcourses.com/course_order). If you can make it up to the advanced courses (for example, the advanced sections of the transformers and NLP courses), then you can consid...
by lazyprogrammer
Tue Dec 13, 2022 2:58 am
Forum: Data Science: Transformers for Natural Language Processing
Topic: loss calculation in Decoder
Replies: 2
Views: 252

Re: loss calculation in Decoder

Thanks for your inquiry.

As mentioned in the video this is how the PyTorch library was implemented.

In general this is their convention as you may have seen with images as well.

You can always check the documentation and source code to confirm.
by lazyprogrammer
Tue Dec 13, 2022 2:56 am
Forum: Data Science: Transformers for Natural Language Processing
Topic: 2 points about "Question - Answering"
Replies: 2
Views: 247

Re: 2 points about "Question - Answering"

> I can understand the former "None" because question is replaced by "None" and only context is shown as tuples after tokenize_fn_validation However I am not clear in latter "None", because context is shown tuples in the result but after the tuples "None" is s...
by lazyprogrammer
Sun Dec 11, 2022 6:52 pm
Forum: Time Series Analysis, Forecasting, and Machine Learning
Topic: wget command not working
Replies: 1
Views: 261

Re: wget command not working

Thanks for your inquiry.

I'd recommend Google searching the problem for tool-specific issues.

I've never used Visual Studio Code so I can't comment on how it works.

Also note that Google Colab is the official environment for the course so you can always do that instead.
by lazyprogrammer
Tue Dec 06, 2022 8:46 pm
Forum: Data Science: Transformers for Natural Language Processing
Topic: Positional Encoding
Replies: 2
Views: 280

Re: Positional Encoding

1) Sure you may have to modify the code if you change that. I'd recommend trying it as an exercise. 2) As mentioned in the lecture the plot was created by first performing PCA. Therefore, the 2 axes correspond to the first 2 principal components. If you don't know about PCA check it out in "Uns...
by lazyprogrammer
Fri Nov 25, 2022 6:52 am
Forum: Data Science: Transformers for Natural Language Processing
Topic: the recources
Replies: 14
Views: 949

Re: the recources

Not willing to followup?
by lazyprogrammer
Thu Nov 17, 2022 1:09 am
Forum: Data Science: Transformers for Natural Language Processing
Topic: the recources
Replies: 14
Views: 949

Re: the recources

But wouldn't this lead to EXACTLY THE SAME PROCESS as I describe above, albeit with a far less convenient user experience? What pedagogical benefit is there to this? That's where we disagree, and you haven't convinced me that this is a real pain point for you, although I am ready and willing to lis...
by lazyprogrammer
Tue Nov 15, 2022 8:17 pm
Forum: Data Science: Transformers for Natural Language Processing
Topic: the recources
Replies: 14
Views: 949

Re: the recources

You have still not answered why your links were working opening Colab notebooks 2 days ago, but are not now. Correction: you didn't ask why. If you want to know "why", I obviously can't share my full reasoning with you as business decisions involve more than simply what students feel enti...

Go to advanced search