andras/great-ai
1
0
Fork 0
Code Issues Pull requests 1 Projects Releases 4 Packages Wiki Activity Actions

Deployed 4f97884 with MkDocs version: 1.3.0

Browse source
This commit is contained in:
2022-07-17 15:16:02 +00:00
parent 4f61718172
commit 58876d64ab
48 changed files with 6685 additions and 2550 deletions
Download patch file Download diff file Expand all files Collapse all files

295
examples/scibert/index.html
View file

@ -17,7 +17,7 @@
<title>Explainable SciBERT - GreatAI documentation</title>
<title>Summarising scientific publications from a tech-transfer perspective - GreatAI documentation</title>
@ -116,7 +116,7 @@
<div class="md-header__topic" data-md-component="header-topic">
<span class="md-ellipsis">
Explainable SciBERT
Summarising scientific publications from a tech-transfer perspective
</span>
</div>
@ -665,6 +665,34 @@
<li class="md-nav__item md-nav__item--active md-nav__item--nested">
<input class="md-nav__toggle md-toggle" data-md-toggle="__nav_5_2" type="checkbox" id="__nav_5_2" checked>
<label class="md-nav__link" for="__nav_5_2">
Explainable SciBERT
<span class="md-nav__icon md-icon"></span>
</label>
<nav class="md-nav" aria-label="Explainable SciBERT" data-md-level="2">
<label class="md-nav__title" for="__nav_5_2">
<span class="md-nav__icon md-icon"></span>
Explainable SciBERT
</label>
<ul class="md-nav__list" data-md-scrollfix>
<li class="md-nav__item md-nav__item--active">
<input class="md-nav__toggle md-toggle" data-md-toggle="toc" type="checkbox" id="__toc">
@ -673,10 +701,158 @@
<label class="md-nav__link md-nav__link--active" for="__toc">
Summarising scientific publications from a tech-transfer perspective
<span class="md-nav__icon md-icon"></span>
</label>
<a href="./" class="md-nav__link md-nav__link--active">
Explainable SciBERT
Summarising scientific publications from a tech-transfer perspective
</a>
<nav class="md-nav md-nav--secondary" aria-label="Table of contents">
<label class="md-nav__title" for="__toc">
<span class="md-nav__icon md-icon"></span>
Table of contents
</label>
<ul class="md-nav__list" data-md-component="toc" data-md-scrollfix>
<li class="md-nav__item">
<a href="#objectives" class="md-nav__link">
Objectives
</a>
</li>
<li class="md-nav__item">
<a href="#overview" class="md-nav__link">
Overview
</a>
</li>
<li class="md-nav__item">
<a href="#summary" class="md-nav__link">
Summary
</a>
<nav class="md-nav" aria-label="Summary">
<ul class="md-nav__list">
<li class="md-nav__item">
<a href="#data-notebook" class="md-nav__link">
Data notebook
</a>
</li>
<li class="md-nav__item">
<a href="#training-notebook" class="md-nav__link">
Training notebook
</a>
</li>
<li class="md-nav__item">
<a href="#deployment-notebook" class="md-nav__link">
Deployment notebook
</a>
<nav class="md-nav" aria-label="Deployment notebook">
<ul class="md-nav__list">
<li class="md-nav__item">
<a href="#additional-files" class="md-nav__link">
Additional files
</a>
</li>
</ul>
</nav>
</li>
</ul>
</nav>
</li>
</ul>
</nav>
</li>
<li class="md-nav__item">
<a href="data/" class="md-nav__link">
Explore data and feasibility of approach
</a>
</li>
<li class="md-nav__item">
<a href="train/" class="md-nav__link">
Fine-tune SciBERT
</a>
</li>
<li class="md-nav__item">
<a href="deploy/" class="md-nav__link">
Create an inference function
</a>
</li>
<li class="md-nav__item">
<a href="additional-files/" class="md-nav__link">
Additional files in the repository
</a>
</li>
</ul>
</nav>
</li>
@ -723,6 +899,75 @@
<label class="md-nav__title" for="__toc">
<span class="md-nav__icon md-icon"></span>
Table of contents
</label>
<ul class="md-nav__list" data-md-component="toc" data-md-scrollfix>
<li class="md-nav__item">
<a href="#objectives" class="md-nav__link">
Objectives
</a>
</li>
<li class="md-nav__item">
<a href="#overview" class="md-nav__link">
Overview
</a>
</li>
<li class="md-nav__item">
<a href="#summary" class="md-nav__link">
Summary
</a>
<nav class="md-nav" aria-label="Summary">
<ul class="md-nav__list">
<li class="md-nav__item">
<a href="#data-notebook" class="md-nav__link">
Data notebook
</a>
</li>
<li class="md-nav__item">
<a href="#training-notebook" class="md-nav__link">
Training notebook
</a>
</li>
<li class="md-nav__item">
<a href="#deployment-notebook" class="md-nav__link">
Deployment notebook
</a>
<nav class="md-nav" aria-label="Deployment notebook">
<ul class="md-nav__list">
<li class="md-nav__item">
<a href="#additional-files" class="md-nav__link">
Additional files
</a>
</li>
</ul>
</nav>
</li>
</ul>
</nav>
</li>
</ul>
</nav>
</div>
</div>
@ -737,24 +982,54 @@
<a href="https://github.com/schmelczer/great-ai/edit/main/docs/examples/scibert.md" title="Edit this page" class="md-content__button md-icon">
<a href="https://github.com/schmelczer/great-ai/edit/main/docs/examples/scibert/index.md" title="Edit this page" class="md-content__button md-icon">
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M20.71 7.04c.39-.39.39-1.04 0-1.41l-2.34-2.34c-.37-.39-1.02-.39-1.41 0l-1.84 1.83 3.75 3.75M3 17.25V21h3.75L17.81 9.93l-3.75-3.75L3 17.25Z"/></svg>
</a>
<h1 id="summarising-scientific-publications-from-a-tech-transfer-perspective">Summarising scientific publications from a tech-transfer perspective<a class="headerlink" href="#summarising-scientific-publications-from-a-tech-transfer-perspective" title="Permanent link">#</a></h1>
<p>This example shows how <code>great-ai</code> is used in practice at ScoutinScience. </p>
<div style="display: flex; justify-content: center;">
<p><a class="md-button md-button--primary" href="https://github.com/schmelczer/great-ai/tree/main/docs/examples/scibert"><span class="twemoji"><svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M7 2v2h1v14a4 4 0 0 0 4 4 4 4 0 0 0 4-4V4h1V2H7m4 14c-.6 0-1-.4-1-1s.4-1 1-1 1 .4 1 1-.4 1-1 1m2-4c-.6 0-1-.4-1-1s.4-1 1-1 1 .4 1 1-.4 1-1 1m1-5h-4V4h4v3Z"/></svg></span> Check out the code</a></p>
<p>This is a simplified example illustrating how <code>great-ai</code> is used in practice at <a href="https://www.scoutinscience.com/" target="_blank">ScoutinScience</a>. The subpages show <code>great-ai</code> in action by going over the life-cycle of finetuning and deploying a BERT-based software service.</p>
<details class="note">
<summary>Propriety data</summary>
<p>The purpose of this example is to show you different ways in which <code>great-ai</code> can assist you. The exact NLP task being solved is not central. Stemming from this, and from the difficult nature of obtaining appropriate training data, the propriety dataset used for the experiments is not shared.</p>
</details>
<h2 id="objectives">Objectives<a class="headerlink" href="#objectives" title="Permanent link">#</a></h2>
<ol>
<li>You will see how the <a href="/reference/utilities">great_ai.utilities</a> can integrate into your Data Science workflow.</li>
<li>You will see how <a href="/reference/large-file">great_ai.large_file</a> can be used to version and store your trained model.</li>
<li>You will see how <a class="autorefs autorefs-internal" href="../../reference/#great_ai.GreatAI">GreatAI</a> should be used prepare your model for a robust and responsible deployment.</li>
<li>You will see multiple ways of customising your deployment.</li>
</ol>
<h2 id="overview">Overview<a class="headerlink" href="#overview" title="Permanent link">#</a></h2>
<p>One of the core features of the ScoutinScience platform is summarising research papers form a tech-transfer perspective. In short, extractive summarisation is preferred using a binary classifier trained on clients' judgement of sentence interestingness. Thus, documents are sentences and the expected output is a binary label showing whether a sentence is "worthy" of being in the tech-transfer summary. Providing an explanation for each decision is imperative since ScoutinScience embraces applying only explainable AI (XAI) methods wherever feasible.</p>
<div class="admonition success">
<p class="admonition-title">Success</p>
<p>You are ready to start the tutorial. Feel free to come back to the <a href="#summary">summary</a> section once you're finished.</p>
</div>
<div style="display: flex; justify-content: space-evenly;">
<p><a class="md-button md-button--primary" href="data/"><span class="twemoji"><svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M12 3C7.58 3 4 4.79 4 7s3.58 4 8 4 8-1.79 8-4-3.58-4-8-4M4 9v3c0 2.21 3.58 4 8 4s8-1.79 8-4V9c0 2.21-3.58 4-8 4s-8-1.79-8-4m0 5v3c0 2.21 3.58 4 8 4s8-1.79 8-4v-3c0 2.21-3.58 4-8 4s-8-1.79-8-4Z"/></svg></span> Examine data</a></p>
<p><a class="md-button md-button--primary" href="train/"><span class="twemoji"><svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 448 512"><!--! Font Awesome Free 6.1.1 by @fontawesome - https://fontawesome.com License - https://fontawesome.com/license/free (Icons: CC BY 4.0, Fonts: SIL OFL 1.1, Code: MIT License) Copyright 2022 Fonticons, Inc.--><path d="M160 80c0-26.51 21.5-48 48-48h32c26.5 0 48 21.49 48 48v352c0 26.5-21.5 48-48 48h-32c-26.5 0-48-21.5-48-48V80zM0 272c0-26.5 21.49-48 48-48h32c26.5 0 48 21.5 48 48v160c0 26.5-21.5 48-48 48H48c-26.51 0-48-21.5-48-48V272zM400 96c26.5 0 48 21.5 48 48v288c0 26.5-21.5 48-48 48h-32c-26.5 0-48-21.5-48-48V144c0-26.5 21.5-48 48-48h32z"/></svg></span> Train model</a></p>
<p><a class="md-button md-button--primary" href="deploy/"><span class="twemoji"><svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path d="M6 20a6 6 0 0 1-6-6c0-3.09 2.34-5.64 5.35-5.96A7.496 7.496 0 0 1 12 4a7.5 7.5 0 0 1 7.35 6A5.02 5.02 0 0 1 24 15a5 5 0 0 1-5 5H6M9.09 8.4 4.5 13l4.59 4.6 1.41-1.42L7.32 13l3.18-3.18L9.09 8.4m5.82 0L13.5 9.82 16.68 13l-3.18 3.18 1.41 1.42L19.5 13l-4.59-4.6Z"/></svg></span> Deploy service</a></p>
</div>
<h2 id="summary">Summary<a class="headerlink" href="#summary" title="Permanent link">#</a></h2>
<h3 id="data-notebook"><a href="data/">Data notebook</a><a class="headerlink" href="#data-notebook" title="Permanent link">#</a></h3>
<p>We load and analyse the data by calculating inter-rater reliability and checking the feasibility of using an AI-based approach by testing the accuracy of a trivial baseline method.</p>
<h3 id="training-notebook"><a href="train/">Training notebook</a><a class="headerlink" href="#training-notebook" title="Permanent link">#</a></h3>
<p>We simply fine-tune SciBERT.</p>
<p>After training and evaluating a model, it is exported using <a class="autorefs autorefs-internal" href="../../reference/#great_ai.save_model">great_ai.save_model</a>. For more info, checkout <a href="/how-to-guides/configure-service">the configuration how-to page</a>.</p>
<h3 id="deployment-notebook"><a href="deploy/">Deployment notebook</a><a class="headerlink" href="#deployment-notebook" title="Permanent link">#</a></h3>
<p>We customise the GreatAI configuration, create custom cahcing for the model and implement an inference function that can be hardened by wrapping it in a <a class="autorefs autorefs-internal" href="../../reference/#great_ai.GreatAI">GreatAI</a> instance. We also extract the attention weights as a quasi-explanation.</p>
<p>Finally, we test the model's inference function through the GreatAI dashboard. <a href="/how-to-guides/use-service">The only thing left is to deploy the hardened service properly.</a></p>
<h4 id="additional-files"><a href="additional-files/">Additional files</a><a class="headerlink" href="#additional-files" title="Permanent link">#</a></h4>
<p>There are some other files required for deploying the notebook. For example, the config file for S3 and MongoDB or a Dockerfile for building a custom image. These are compiled and shown on a separate page.</p>
<hr>
<div class="md-source-file">
<small>
Last update:
<span class="git-revision-date-localized-plugin git-revision-date-localized-plugin-date">July 13, 2022</span>
<span class="git-revision-date-localized-plugin git-revision-date-localized-plugin-date">July 17, 2022</span>
</small>
@ -791,13 +1066,13 @@
<a href="../../explanation/" class="md-footer__link md-footer__link--next" aria-label="Next: Explanation" rel="next">
<a href="data/" class="md-footer__link md-footer__link--next" aria-label="Next: Explore data and feasibility of approach" rel="next">
<div class="md-footer__title">
<div class="md-ellipsis">
<span class="md-footer__direction">
Next
</span>
Explanation
Explore data and feasibility of approach
</div>
</div>
<div class="md-footer__button md-icon">