https://medium.com/swlh/fun-with-apache-lucene-and-bert-embeddings-c2c496baa559 #embeddings #searchengines #lucene
Medium
Fun With Apache Lucene and BERT Embeddings
After publishing the blog post on neural search with BERT and Solr (6.6.0), I got a few questions on how to run this with version 8.6.x of…
https://www.programmersought.com/article/60001511310/ Focal loss implementations #tutorial
Programmersought
Several implementation versions of focal loss (Keras/Tensorflow) - Programmer Sought
Several implementation versions of focal loss (Keras/Tensorflow), Programmer Sought, the best programmer technical posts sharing site.
https://github.com/aliabd/history-of-interpretation
Saliency methods link a deep neural network's (DNN) prediction to the input features that most
influence that prediction. #Library
Saliency methods link a deep neural network's (DNN) prediction to the input features that most
influence that prediction. #Library
GitHub
aliabd/history-of-interpretation
TensorFlow implementation for SmoothGrad, Grad-CAM, Guided backprop, Integrated Gradients and other saliency techniques - aliabd/history-of-interpretation
Forwarded from Denis Sexy IT 🤖
GPT-3 моделька на английском языке, на 2.7 миллиардов параметров стала доступна онлайн.
Это реплика и обучили ее энтузиасты eleuther.ai а не люди из OpenAI, оригинальная GPT-3 на ~175 миллиардов параметров продается как коммерческий продукт и отдается по эксклюзивной лицензии через Microsoft.
В общем, опенсорс сообщество как всегда великолепно – также выложили модельку поменьше, все по ссылкам ниже.
Код | Коллаб
Количество сгенерированных текстов в интернете увеличилось вдвое 🌚
Это реплика и обучили ее энтузиасты eleuther.ai а не люди из OpenAI, оригинальная GPT-3 на ~175 миллиардов параметров продается как коммерческий продукт и отдается по эксклюзивной лицензии через Microsoft.
В общем, опенсорс сообщество как всегда великолепно – также выложили модельку поменьше, все по ссылкам ниже.
Код | Коллаб
Количество сгенерированных текстов в интернете увеличилось вдвое 🌚
https://arxiv.org/pdf/1912.13318.pdf
LayoutLM: Pre-training of Text and Layout for Document Image Understanding #paper
LayoutLM: Pre-training of Text and Layout for Document Image Understanding #paper
Russian corpus collection
https://natasha.github.io/corus/
https://github.com/natasha/corus
https://nbviewer.jupyter.org/github/natasha/corus/blob/master/docs.ipynb
#Corpus #Dataset
https://natasha.github.io/corus/
https://github.com/natasha/corus
https://nbviewer.jupyter.org/github/natasha/corus/blob/master/docs.ipynb
#Corpus #Dataset
natasha.github.io
Corus — коллекция русскоязычных NLP-датасетов
Ссылки на публичные русскоязычные датасеты, Python-пакет с функциями-загрузчиками
https://github.com/fbdesignpro/sweetviz In-depth EDA (target analysis, comparison, feature analysis, correlation) #Library #Tools
example of use: http://cooltiming.com/SV/SWEETVIZ_REPORT_COMPARED.html
example of use: http://cooltiming.com/SV/SWEETVIZ_REPORT_COMPARED.html
GitHub
GitHub - fbdesignpro/sweetviz: Visualize and compare datasets, target values and associations, with one line of code.
Visualize and compare datasets, target values and associations, with one line of code. - fbdesignpro/sweetviz
https://github.com/jessevig/bertviz
BertViz is a tool for visualizing attention in the Transformer model, supporting most models from the transformers library (BERT, GPT-2, XLNet, RoBERTa, XLM, CTRL, MarianMT, etc.). #Library #Tools
BertViz is a tool for visualizing attention in the Transformer model, supporting most models from the transformers library (BERT, GPT-2, XLNet, RoBERTa, XLM, CTRL, MarianMT, etc.). #Library #Tools
GitHub
GitHub - jessevig/bertviz: BertViz: Visualize Attention in NLP Models (BERT, GPT2, BART, etc.)
BertViz: Visualize Attention in NLP Models (BERT, GPT2, BART, etc.) - GitHub - jessevig/bertviz: BertViz: Visualize Attention in NLP Models (BERT, GPT2, BART, etc.)
https://github.com/calculatedcontent/weightwatcher
WeightWatcher (WW): is an open-source, diagnostic tool for analyzing Deep Neural Networks (DNN), without needing access to training or even test data.
#Diagnostic #Tool
WeightWatcher (WW): is an open-source, diagnostic tool for analyzing Deep Neural Networks (DNN), without needing access to training or even test data.
#Diagnostic #Tool
GitHub
GitHub - CalculatedContent/WeightWatcher: The WeightWatcher tool for predicting the accuracy of Deep Neural Networks
The WeightWatcher tool for predicting the accuracy of Deep Neural Networks - GitHub - CalculatedContent/WeightWatcher: The WeightWatcher tool for predicting the accuracy of Deep Neural Networks