erlemar/wheat

Wheat Detection challenge on Kaggle

Hello, I'm Andrey 👋

Important

I have a personal website now! You can find it here

I'm an economist by education and usually try to understand why things happen.

For a long time, I was fascinated by foreign languages and studied them: English (my native is Russian), German, Japanese, and Spanish.

Currently, I'm Data Scientist and an active Kaggler.

Also, I have a portfolio listing some of my projects, which include Kaggle kernels, pet-projects, and other things.

Latest Blog Posts

Medium

• Paper Review: Unveiling Encoder-Free Vision-Language Models
• Paper Review: Husky: A Unified, Open-Source Language Agent for Multi-Step Reasoning
• Paper Review: Samba: Simple Hybrid State Space Models for Efficient Unlimited Context Language…
• Book Review: Accelerate Model Training with PyTorch 2.X
• Paper Review: σ-GPTs: A New Approach to Autoregressive Models

Habr (russian language)

• [Перевод] Мой путь изучения иностранных языков: прошлое, настоящее и будущее
• Третья жизнь пет-проекта по распознаванию рукописных цифр
• Как мы не смогли создать медицинского чат-бота. История проекта, который так и не увидел свет
• Из экономистов в дата-сайнтисты или как стать лидером рейтинга Kaggle Notebooks
• Обзор на статью Visual Transformers — новый подход к тренировке моделей компьютерного зрения на основе visual tokens

My talks/interviews

2019:

• ML Training talks about gold winning solution in Kaggle Predicting Molecular Properties Competition
• An overview of 2019 year in Kaggle, Talk in Russian at ODS.AI event
• Data Halloween: predicting chaotic actions of clients. Talk in Russian at ODS.AI event
• Storytelling with data visualization: case of kaggle kernels. Talk in Russian at Data Fest 2019
• Interview with Kaggle Kernels Grandmaster #1: Artgor | Andrew Lukyanenko. Chai Time Data Science | player.fm link
• Kaggle: People, We’ve Gamified Progress

2020:

• Andrey Lukyanenko - Handwritten digit recognition w/ a twist & topic modelling over time
• Pair Programming: Deep Learning Model For Drug Classification With Andrey Lukyanenko
• DataFest 2020 talk: My Kaggle Story
• Overview of approaches to "Mechanisms of Action (MoA) Prediction" competition on Kaggle (3 weeks before the end of the competition)
• Metro Analysis and Visualization
• Writing reusable pipelines in Deep Learning | Mindhack! Summit
• December Lightning Talks. Training pipeline with Pytorch Lightning and Hydra

2021:

• Podcast talk about Kaggle, modern ML technologies, frameworks, hardware, and many other things
• Building your Data Science Career: how to become a data scientist

My Github and StackOverflow stats

wheat's Issues

Pip install -r requirements.txt fails

Hi there I've tried to pip install the dependencies but I get the following error both on Ubuntu 18 and macOS ERROR: Could not find a version that satisfies the requirement pytorch_lightning==0.8.0. Any idea? Did you use python2?

Why is there no losses.backward and optimizer.step on training_steps?

I am using this as a baseline on my custom dataset. While I have used Pytorch for FasterRCNN before (not with Pytorch Lightning), I am curious as to why it does not have the typical ending of each training step like the ff:

...
losses = sum(loss for loss in loss_dict.values())
optimizer.zero_grad()
losses.backward()
optimizer.step()
...

I am not seeing this being implemented explicitly in the code and the utils. Am I missing something why isnt this included?