PinnedPublished inTDS ArchiveGeneralized Attention Mechanism: BigBird’s Theoretical Foundation and General Transformers ModelsAs a little background, BigBird is a model released from Google Research relatively recently that can handle much lengthier sequences of…Dec 19, 2021Dec 19, 2021
Published inTDS Archive[New Hugging Face Feature] Constrained Beam Search with 🤗 TransformersA new Hugging Face feature allows you customize and guide your language model outputs (like forcing a certain sequence within the output).Mar 22, 2022Mar 22, 2022
Published inExplained Relentlessly: Deep Learning & Natural Language ProcessingTutorial: Storing Large A.I. Models with ‘gdrive’ (don’t use Git LFS)Git-LFS is a short-term solution that shouldn’t be used for your AI projects. Learn how to use gdrive to save & export your models.Jul 3, 20212Jul 3, 20212
Published inExplained Relentlessly: Deep Learning & Natural Language ProcessingANCE Contrastive Learning for Dense Retrieval: Sampling Negative Examples & The Variation…Why negative examples are so important in training Neu-IR models and how ANCE obtains the mathematically optimized distribution.Jun 25, 2021Jun 25, 2021
How to Set up code-server daemon on Ubuntu: Root Access, Docker, systemd, NGINX Reverse-Proxy, and…systemctl code-server + nginx publicly expose. Or use Docker to set it up. How to use Nginx as reverse-proxy to IP white-list.Jun 23, 2021Jun 23, 2021
Published inExplained Relentlessly: Deep Learning & Natural Language ProcessingRetrieval-Augmented Generation (RAG): Control Your Model’s Knowledge and… Hallucinations!Large LMs are really good today. But they’re known to hallucinate. Maybe we should let it cheat a little with Wikipedia.Jun 20, 20211Jun 20, 20211
Published inExplained Relentlessly: Deep Learning & Natural Language Processing[Part I] Predicting on Text Pairs with Transformers: Cross-Encoding with BERT[Code Implementation] Make SOTA sentence pair regression using BERT’s Cross-Encoding with SequenceClassification.Jun 18, 2021Jun 18, 2021
Published inExplained Relentlessly: Deep Learning & Natural Language ProcessingAutoencoders & Power of Mathematical OptimizationsI don’t know if many people agree with this sentiment but I was pretty shocked when I learned how A.I. worked the first time. Growing up, INov 23, 20183Nov 23, 20183
Published inExplained Relentlessly: Deep Learning & Natural Language ProcessingOverfitting (What They Are & Train, Validation, Test & Regularization)I’ve taken a bit of a hiatus because I’ve been busy with school and other things and I’ve been spending time learning more ML/DL stuff…Apr 15, 2018Apr 15, 2018
Published inExplained Relentlessly: Deep Learning & Natural Language ProcessingDocument Classification Part 4: Variations of This Approach (Malware Detection Using Document…In this article I will explain an application of the concepts so far that goes beyond just categorizing text documents.Mar 22, 2018Mar 22, 2018
![[New Hugging Face Feature] Constrained Beam Search with 🤗 Transformers](https://miro.medium.com/v2/resize:fill:160:106/1*1F4C5ZJMaxrHskH3ra4vdw.jpeg)
![[New Hugging Face Feature] Constrained Beam Search with 🤗 Transformers](https://miro.medium.com/v2/resize:fill:320:214/1*1F4C5ZJMaxrHskH3ra4vdw.jpeg)
![[Part I] Predicting on Text Pairs with Transformers: Cross-Encoding with BERT](https://miro.medium.com/v2/resize:fill:160:106/1*6-U1IkR2BCfIqWfLRJqAEw.jpeg)
![[Part I] Predicting on Text Pairs with Transformers: Cross-Encoding with BERT](https://miro.medium.com/v2/resize:fill:320:214/1*6-U1IkR2BCfIqWfLRJqAEw.jpeg)
