<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[ML Systems Flashcards - Medium]]></title>
        <description><![CDATA[Master machine learning engineering concepts with interactive flashcards. Perfect for quick revision, interview prep, and practical insights. Learn efficiently and effectively. - Medium]]></description>
        <link>https://medium.com/ml-systems-flashcards?source=rss----170b7e7abd22---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>ML Systems Flashcards - Medium</title>
            <link>https://medium.com/ml-systems-flashcards?source=rss----170b7e7abd22---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 02:36:56 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/ml-systems-flashcards" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Flashcard: ML Model Deployment in GCP — Production]]></title>
            <link>https://medium.com/ml-systems-flashcards/flashcard-ml-model-deployment-in-gcp-production-f6cd785cd67f?source=rss----170b7e7abd22---4</link>
            <guid isPermaLink="false">https://medium.com/p/f6cd785cd67f</guid>
            <category><![CDATA[ml-model-deployment]]></category>
            <category><![CDATA[mlops]]></category>
            <category><![CDATA[flashcards]]></category>
            <category><![CDATA[gcp]]></category>
            <category><![CDATA[vertex-ai]]></category>
            <dc:creator><![CDATA[HARI Y.L.N]]></dc:creator>
            <pubDate>Fri, 13 Sep 2024 11:35:59 GMT</pubDate>
            <atom:updated>2024-09-13T11:37:16.637Z</atom:updated>
            <content:encoded><![CDATA[<h3>Flashcard: ML Model Deployment in GCP — Production</h3><p>To deploy a machine learning model on Google Cloud Platform’s Vertex AI. It involves containerizing the model, creating a prediction script, and registering the model in Vertex AI. This process ensures portability and consistency for the model deployment.</p><h4><strong>Components Needed:</strong></h4><ul><li><strong>Model Artifacts:</strong> Trained model, scalers, encoders, etc.</li><li><strong>requirements.txt:</strong> Python dependencies.</li><li><strong>Prediction Script:</strong> Handles requests, uses the model for predictions, and has /health and /predict routes.</li><li><strong>Docker file:</strong> Defines the image’s build instructions.</li></ul><h4><strong>Model deployment/registration:</strong></h4><ol><li><strong>Containerize your model: </strong>Create a Docker file that sets up a FastAPI app environment with dependencies, model &amp; scaler artifacts, and runs your prediction script as a Http server.</li><li><strong>Build and push the image: </strong>Authenticate with GCP using gcloud auth login. Set your project ID and build the image with docker build -t .Push the image to Artifact Registry with docker push &lt;container_link&gt;.</li><li><strong>Register the model (choose one):</strong></li></ol><ul><li><strong>Command Line:</strong> Use gcloud ai models create with your model name, region, version, and container image URI.</li><li><strong>Vertex AI UI:</strong> Navigate to Model Registry, select “Import Model,” and provide your model details, including the container image URI.</li></ul><p>check out my full blog post <a href="https://medium.com/applied-ml-mlops/register-deploy-ml-models-in-production-using-gcp-943c04254f1b">here</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f6cd785cd67f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/ml-systems-flashcards/flashcard-ml-model-deployment-in-gcp-production-f6cd785cd67f">Flashcard: ML Model Deployment in GCP — Production</a> was originally published in <a href="https://medium.com/ml-systems-flashcards">ML Systems Flashcards</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Flashcard : ML Model Compression]]></title>
            <link>https://medium.com/ml-systems-flashcards/flashcard-ml-model-compression-8487ccf1f14e?source=rss----170b7e7abd22---4</link>
            <guid isPermaLink="false">https://medium.com/p/8487ccf1f14e</guid>
            <category><![CDATA[model-optimization]]></category>
            <category><![CDATA[latency-optimization]]></category>
            <category><![CDATA[model-compression]]></category>
            <dc:creator><![CDATA[HARI Y.L.N]]></dc:creator>
            <pubDate>Sat, 31 Aug 2024 06:43:01 GMT</pubDate>
            <atom:updated>2024-08-31T06:43:01.109Z</atom:updated>
            <content:encoded><![CDATA[<h3>Flashcard: ML Model Compression</h3><p>A technique to optimize machine learning models by reducing their size, making them more efficient without losing accuracy.</p><blockquote><strong>Ways to reduce inference latency : </strong>Make the model smaller (<strong><em>model compression</em></strong>), Make it do inference faster (<strong>inference optimization</strong>), Improve the hardware it runs on.</blockquote><blockquote><strong>Why do we need model compression: </strong>to fit models on edge devices, for critical, highly scalable, low latency applications.</blockquote><h4><strong>Techniques for Model Compression:</strong></h4><p><strong>Pruning:</strong> Originally used for decision trees, pruning removes uncritical sections of a neural network. It can either remove entire nodes or set unimportant weights to zero, making the network more sparse and efficient. Try it with TensorFlow’s Model Optimization Toolkit.</p><p><strong>Quantization: </strong>This method reduces a model’s size by using fewer bits to represent parameters. For example, using 16-bit or 8-bit integers instead of 32-bit floats. Explore it with PyTorch’s Quantization API.</p><p><strong>Low-Rank Factorization:</strong> This technique replaces high-dimensional tensors with lower-dimensional ones, reducing parameters and increasing speed. Models like SqueezeNets and MobileNets use this approach to achieve high accuracy with fewer parameters.</p><p><strong>Knowledge Distillation:</strong> A smaller model (student) is trained to mimic a larger model (teacher), retaining most of the performance with fewer parameters. For instance, DistilBERT reduces parameters by 40% while maintaining 97% of its capabilities.</p><p>check out my full blog post <a href="https://medium.com/applied-ml-mlops/model-compression-enhancing-operational-efficiency-13aa51ff4152">here</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8487ccf1f14e" width="1" height="1" alt=""><hr><p><a href="https://medium.com/ml-systems-flashcards/flashcard-ml-model-compression-8487ccf1f14e">Flashcard : ML Model Compression</a> was originally published in <a href="https://medium.com/ml-systems-flashcards">ML Systems Flashcards</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>