Experiment Tracking Using MLflow in Machine Learning | Model Versioning & Model Registry

No video

Experiment Tracking Using MLflow in Machine Learning | Model Versioning & Model Registry | Part 1

Рет қаралды 35,444

Күн бұрын

MLflow Live Demo | Experiment Tracking and Model Versioning
Topics Covered:
1. Train a Basic classifier using Random Forest
2. Create Experiment-Basic classifier
3. Log metrics, model, and other artifacts
4. Tune model using hyperparameter tuning using Randomized Search CV
5. Create another experiment in MLFlow- Optimised classifier
6. Use SQLite as the backend database for model registry
7. Explore MLflow UI, interpret experiments and runs
If you find this video helpful, don't forget to like share and subscribe.
notebook link: github.com/Tri...
LLM and Langchain live demo and code explanation:
• Step by Step guide to ...
Connect with me:
LinkedIn: / ashutoshtripathiai
Instagram: / ashutoshtripathi_ai
Twitter: / ashutosh_ai
Website: ashutoshtripat...
Video Edited Using filmora.wonder...

Пікірлер: 160

@newsxreactions 20 сағат бұрын

This is GoldMine for MLFlow deployement . Thanks for sharing such Gold Mine content.

@AshutoshTripathi_AI 19 сағат бұрын

Thank you

@taukeerahmad9328 7 ай бұрын

after watching tons of lectures finally I got first real MLOPS video ...going to share it with my whole class.....I subscribed it sir thamks alot.

@AshutoshTripathi_AI 7 ай бұрын

Thank you very much for sharing it. I am glad you liked it and found it useful.

@praveenkuthuru7439 22 күн бұрын

This is a wonderful explanation of the MLflow implementation....thanks for the efforts and making the learner's understanding very simple yet effective...good work...

@AshutoshTripathi_AI 20 күн бұрын

Thanks a lot Praveen

@RISWANAHAMEDMbtech_ai_b2022 Ай бұрын

the best mlflow i first seen

@AshutoshTripathi_AI Ай бұрын

Thank you

@AutomateByte 6 ай бұрын

Finally found the best channel to learn MLFLOW with practical examples

@AshutoshTripathi_AI 6 ай бұрын

Thank you Friend 🙏

@dev09able 6 ай бұрын

Thanks for creating this video, after watching many videos this gave me the clarity what MLFlow is all about.

@AshutoshTripathi_AI 6 ай бұрын

Glad that it helped you

@skyrayzor3693 11 ай бұрын

This is one of the best MLFlow tutorial, thank you sir. 👍

@AshutoshTripathi_AI 10 ай бұрын

Most welcome

@shubhampatil4176 Жыл бұрын

I was finding for such explanation.. Got it here.. Excited for next part

@AshutoshTripathi_AI Жыл бұрын

Thank you. Will upload the next part soon. Came onsite for office work. So it got delayed.

@AshutoshTripathi_AI Жыл бұрын

Hi, just uploaded the part 2 which explains the model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@user-kg1tt8pw4t 6 ай бұрын

Very good job, Ashutosh! Thank you so much for this playlist!

@AshutoshTripathi_AI 6 ай бұрын

Good to hear. Thank you for watching!

@plusminuschirag Жыл бұрын

Great explanation man, resources like yours are very rare in MLOps domain. Great work, please make a video on curated list of resources or path to follow if possible.

@AshutoshTripathi_AI Жыл бұрын

Hi, just uploaded the video on model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@flipthecointwice 6 ай бұрын

You are great ! This is how tutorial should be

@AshutoshTripathi_AI 5 ай бұрын

Thank you.

@DeepeshSinghAndroid Жыл бұрын

Great job Ashutosh. Quite easy to follow, and brilliantly explained.. Thank you!

@AshutoshTripathi_AI Жыл бұрын

Glad you liked it. Here is the part 2 of this video that explains the model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@DeepeshSinghAndroid Жыл бұрын

@@AshutoshTripathi_AI thank you. BTW I have been trying to enable SQLite tracking on my MacBook but it was constantly giving errors. After many attempts, I found these two things which worked for me. Sharing my note with your viewers. drive.google.com/file/d/1nnWFKx5bUw_kU27SgUfzW24O3tRocCZ2/view

@AshutoshTripathi_AI Жыл бұрын

@@DeepeshSinghAndroid thanks for sharing

@geen160 2 ай бұрын

After watching tons of lectures, I finally found the best channel to learn MLOps. Thank you and keep going

@AshutoshTripathi_AI Ай бұрын

Happy to hear that!

@md.towfiqurrahman3316 Жыл бұрын

Thanks for the explanations. ZenML is a good tool for CI/CD with ML models.

@AshutoshTripathi_AI Жыл бұрын

I will check ZenML. Thank you.

@basi6621 Жыл бұрын

good one.! thank you.. can't wait for the part 2

@AshutoshTripathi_AI Жыл бұрын

Coming soon.

@AshutoshTripathi_AI Жыл бұрын

Hi, just uploaded the part 2 which explains the model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@atulanand4824 11 ай бұрын

One of the best explanation sir🎉was looking for Mlflow, keep going sir and if possible please come up with new playlist of PySpark for Data Science much needed

@AshutoshTripathi_AI 11 ай бұрын

Sure, will try to create a playlist on pyspark

@Girishrrao Жыл бұрын

Good Job Ashutosh. Awaiting for more videos on MLFlow and the part2 of this video.😊

@AshutoshTripathi_AI Жыл бұрын

Thank you. Very soon will publish

@AshutoshTripathi_AI Жыл бұрын

Hi Girish, just uploaded the part 2 which explains the model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@lavanyasagunthala3801 Жыл бұрын

Thank you sir for the detailed and clear explanation.

@AshutoshTripathi_AI Жыл бұрын

Thanks

@mohittewari6796 Жыл бұрын

Loved this video. very nicely explained.

@AshutoshTripathi_AI Жыл бұрын

Thank you

@ZavierBanerjea 2 ай бұрын

Great work, Ashutosh 👍

@AshutoshTripathi_AI 2 ай бұрын

Thank you.

@mrityunjay4750 Жыл бұрын

This one is really nice ! Thanks for you time and effort for this awesome work.

@AshutoshTripathi_AI 8 ай бұрын

You are welcome 🤗

@krishj8011 Ай бұрын

Great tutorial

@AshutoshTripathi_AI 3 күн бұрын

Thank you

@chinmaykulkarni8559 9 ай бұрын

Very useful Ashutosh Ji

@AshutoshTripathi_AI 8 ай бұрын

Thank you.

@abusufyanvu Жыл бұрын

Great , Rare tutorial

@AshutoshTripathi_AI Жыл бұрын

Thank you

@vinayakdumbre2828 Жыл бұрын

Great explanation

@AshutoshTripathi_AI Жыл бұрын

Thank you 🙏

@Gabbar_viratfan18 Жыл бұрын

Hi @AshutoshTripathi_AI , this is the best explanation of mlflow working...but where is the next part of this vedio sir. Please reply...

@AshutoshTripathi_AI Жыл бұрын

Coming soon. Sorry for the delay.

@AshutoshTripathi_AI Жыл бұрын

Hi, just uploaded the part 2 which explains the model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@sohildoshi2655 6 ай бұрын

thankyou nice explanation really helpful

@AshutoshTripathi_AI 6 ай бұрын

Welcome 🤗

@shekharrawat5703 Ай бұрын

Hello Ashutosh , Thank for explaining everything in details its really Helpful , I have a question , - Can we acheive CI/CD in Machine learning with MLflow ? - Why do we use kubeflow is it different from MLflow ? - Which solution MLflow or Kubeflow is widly in Industry Practise ?

@AshutoshTripathi_AI 7 күн бұрын

Hi Shekhar, Mlflow or kubeflow are not meant for CI/CD. For CI/CD you can use either GitHub actions or Jenkins or anyother tool. There is overlap between kubeflow and mlflow in terms of functionalities. Kubeflow provides ml pipeline creation which was not there in mlflow earlier but now mlflow also provides pipelines named as recipes. Mlflow mainly used for experiment tracking and model registry. So in industry for experiment tracking and model registry mlflow is frequently used. Kubeflow is used by some industry but not so widely as mlflow. Because for pipeline creation people use airflow more frequently

@shekharrawat5703 6 күн бұрын

@@AshutoshTripathi_AI thanks for detailed reply , much appreciated.

@chrisogonas Жыл бұрын

That was pretty useful. Thanks

@AshutoshTripathi_AI Жыл бұрын

Glad it was helpful!🙏

@chrisogonas Жыл бұрын

@@AshutoshTripathi_AI Certainly!

@AshutoshTripathi_AI Жыл бұрын

Part 2 Model Serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@ngocanhphantich Жыл бұрын

Tks so much for this video

@AshutoshTripathi_AI 8 ай бұрын

Welcome 😊

@aditya_01 Жыл бұрын

great video thanks a lot

@AshutoshTripathi_AI Жыл бұрын

Glad you found it good.

@evegarcianz Жыл бұрын

Great video! Gracias!

@AshutoshTripathi_AI Жыл бұрын

Thank you.

@suhasnr6053 Жыл бұрын

Hello Ashutosh, thanks for this video. Could you please come up with a video by Deploying the same in production as you said? I am waiting for that video. Good work!!!

@AshutoshTripathi_AI Жыл бұрын

Sure. You will get it soon. Thank you

@suhasnr6053 Жыл бұрын

Thank you for replying

@AshutoshTripathi_AI Жыл бұрын

Hi, just uploaded the part 2 which explains the model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@ranjithkumarkarthic8696 2 жыл бұрын

Very good tutorial. Thanks for that. But, why don't you share the notebook. It is very helpful.

@AshutoshTripathi_AI 2 жыл бұрын

Please check the description of the video. I have shared the GitHub link where you can find the notebook. Thanks for watching

@AshutoshTripathi_AI 2 жыл бұрын

You can also find it here: github.com/TripathiAshutosh/mlflow

@ranjithkumarkarthic8696 2 жыл бұрын

@@AshutoshTripathi_AI Thank you so much sir. Thanks for your replay. You are great :)

@msaclodev Жыл бұрын

Great

@AshutoshTripathi_AI Жыл бұрын

Thank you

@vishalwaghmare3130 Жыл бұрын

Very helpful 👍

@AshutoshTripathi_AI Жыл бұрын

Thanks

@geetatripathi9335 Жыл бұрын

Very Helpful

@BIZSURESH Жыл бұрын

Excellent and your explanation is makes easy understanding....👌👌👌👌👍👍🙏🙏🙏🙏🙏 Time stamp 28:24 ; request your help for get_metrics() is belongs to which class....?

@AshutoshTripathi_AI Жыл бұрын

I have created a function named get_metrics() and that is being called here. You just scroll up and you will see all the functions. If this answers then fine, if you meant something else then please let me know.

@chetanmundhe7899 Жыл бұрын

Very Nice..Thanks for sharing & keep posting such videos, have you created next part of this video? please share the link since I have not found the next part .

@AshutoshTripathi_AI Жыл бұрын

Thank Chetan. Actually I am travelling due to office work. So I did not get a chance to create. But will resume it next week. So soon it will be available.

@chetanmundhe7899 Жыл бұрын

Also please try to add one thing in video that you can show how the production model gets updated in one click, that's you CICD pipeline, or you can make a seperate video on that... Just sharing my thought

@AshutoshTripathi_AI Жыл бұрын

@@chetanmundhe7899 actually good suggestion to create a new video on that. Thanks I will try to make one on this.

@AshutoshTripathi_AI Жыл бұрын

Hi, just uploaded the part 2 which explains the model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@sagarbhagwani7193 4 ай бұрын

is this a complete ml ops playlist covering each and everything? @Ashutosh Tripathi

@mandilquioxtenlp1202 Жыл бұрын

Thank you so much.

@AshutoshTripathi_AI Жыл бұрын

You are welcome

@zahrafarhadi1338 2 жыл бұрын

Very good Tutorial, thanks! I just have a questions and thought you might have the answer to it. After we register multiple versions of a model, is it possible to change the version name? For example, I want to be able to change Version 1 to "V1-xfeatures"

@AshutoshTripathi_AI 2 жыл бұрын

I don't think it is possible. As each new model registered under the same model name increments the version number. So if we have some random names to new version it would be difficult to identify the latest one or give it an order. You can rename the model and also update the description. I will check again and if I find anything I will share. Thanks for watching

@tarunsingh5615 2 ай бұрын

Please create a mlops playlist (basic to advance) covering imp topics and also please divide your video into multiple smaller parts. love your explanation!!

@AshutoshTripathi_AI 2 ай бұрын

Sure, noted, I will try to do the above.

@srinivasat2691 Жыл бұрын

Very good tutorial !!! Unable to find next part

@AshutoshTripathi_AI Жыл бұрын

You will get it soon. Came to onsite travel. So I did not get time to record.

@AshutoshTripathi_AI Жыл бұрын

Hi, just uploaded the part 2 which explains the model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@ratnakarkadipikonda9798 Жыл бұрын

Nice one..What is the difference between MLflow and MLops?? Please can you do on MLops video.Thank you.

@AshutoshTripathi_AI Жыл бұрын

MLOps is the set of practices to be performed to deploy machine learning models in production where model should run in more efficient and reliable manner. To achieve this for every steps there are certain tools and techniques. One of the tool is MLflow which does mainly the experiment tracking and provides model registry which helps us to achieve model Versioning.

@AshutoshTripathi_AI Жыл бұрын

Have a look into this short video on MLOps vs MLflow kzfaq.infopU6pJvFxoU8?feature=share

@ratnakarkadipikonda9798 Жыл бұрын

Thank you

@ratnakarkadipikonda9798 Жыл бұрын

Can you try to explain MLflow in Databricks….Thank you

@AshutoshTripathi_AI Жыл бұрын

@@ratnakarkadipikonda9798 ok.

@deepaktripathi4094 Жыл бұрын

Hi Ashutosh, it was a great video. I was looking for video on productionization using mlflow, but could not find one. Did you create that yet?

@AshutoshTripathi_AI Жыл бұрын

Thank you Deepak. In productionization not created yet. You will find it soon in the channel.

@mehrozalam94 2 жыл бұрын

Great Demo brother, do you know the difference between mlfow 1.0 and mlflow 2.0. because mlflow is an older project as I know. Thank you

@AshutoshTripathi_AI 2 жыл бұрын

I think MLFlow2.0 is not released yet. Databrick is planning to release it soon. What I know the main addition is pipeline creation which helps in automating the ml in production deployment.

@AshutoshTripathi_AI Жыл бұрын

Hi, just uploaded the part 2 which explains the model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@sakarsalunke3245 Жыл бұрын

Video was quite insightful and awesome video for beginners. But while implementing tensorflow model. I am not able to log tensorflow model on mlflow ui. Please help me in this issue!

@AshutoshTripathi_AI Жыл бұрын

Hi, sorry for the late reply if you are still having the issue, Send me the screenshot of the error, i will have a look and see if i can suggest something.

@pzende 2 жыл бұрын

Hello Ashutosh, it is a great tutorial , perhaps the best on the MLFlow in entire universe :). But may I know if i had installed jupyter already, do I need to open the jupyter within the enviornment?

@AshutoshTripathi_AI 2 жыл бұрын

Not necessary to open the jupyter within the environment. But then if you want to use the conda environment then from the jupyter notebook you have to select the kernel of that environment which has all dependencies installed. Else it will try to fetch the dependencies from the base environment. And to make kernel available inside jupyter notebook you have create first and add to notebook. I did open jupyter within the environment so that whatever ipykernel I have created is pre-selected in the notebook. The bottom line is to use the specific conda environment which has all dependencies installed. Hope it is clear. And thank you very much for your kind words.

@pzende 2 жыл бұрын

@@AshutoshTripathi_AI Hello thanks for the video once again, it saved my time. when you are uploading the next video?

@AshutoshTripathi_AI 2 жыл бұрын

@@pzende very soon. In progress.

@AshutoshTripathi_AI Жыл бұрын

Tensorflow based model experiment tracking live demo: kzfaq.info/get/bejne/n7eWZJx0qa7MmJc.html

@bonitaparangjaya7891 Жыл бұрын

Hi Ashutosh, plot_roc_curve is not available in sklearn.metrics. i am using scikit-learn 1.2.2 and python 3.10.10 please advise

@AshutoshTripathi_AI Жыл бұрын

Plot_roc_curve is deprecated in newer version. Just refer below link for a ref on how to plot roc curve with sklearn: scikit-learn.org/1.0/visualizations.html#visualizations

@bonitaparangjaya7891 Жыл бұрын

@@AshutoshTripathi_AI Thank you very much

@HARIKRISHNAN-be7jh Жыл бұрын

Great explanation sir. Artifacts are loading for long time but not showing in the UI. how to solve that?

@AshutoshTripathi_AI Жыл бұрын

You might be referring wrong UI. What backend you have used file base or sqlite?

@HARIKRISHNAN-be7jh Жыл бұрын

@@AshutoshTripathi_AI File base sir

@Trends_Forever Жыл бұрын

Once we save ml models , for multiple tasks, how to track all models for multiple tasks at same time and , how to use saved pickle files to track metrics on mlflow, i have 10 files for 10 tasks I have saved pickle files for predicting,how to track and changes the code in mlflow

@AshutoshTripathi_AI Жыл бұрын

Each task I am assuming is related with one model. So simply each one as a separate experiment inside the mlflow. Then track it. However if you want to track all together and if so t want to use MLflow UI then you have to write the custom functions which will extract the tracking data from mlflow UI for all the tasks in one go and that you can display in some dashboard or console for tracking purposes.

@sornajavadi4883 Жыл бұрын

Hi Ashutosh great video on mlflow. When is part two of the video available?

@AshutoshTripathi_AI Жыл бұрын

Working on it. Mostly it will be available in the first week of November.

@sornajavadi4883 Жыл бұрын

Thank you we’re looking forward to it

@AshutoshTripathi_AI Жыл бұрын

Hi, just uploaded the part 2 which explains the model serving using MLFlow model registry: kzfaq.info/get/bejne/Z6uJibR4yb66f6M.html

@noamills1130 Жыл бұрын

Thanks for sharing! Why do you use pip install instead of conda install?

@AshutoshTripathi_AI Жыл бұрын

You can conda...no specific reason.

@tanmaykapil5362 Жыл бұрын

Hi sir i am getting an error AttributeError: module 'sklearn.metrics' has no attribute 'plot_roc_curve' and same in confusion matrix

@AshutoshTripathi_AI Жыл бұрын

From sklearn 1.2 onwards plot_roc_curve is deprecated. In newer version you can use like below from sklearn.metrics import RocCurveDisplay svc_disp = RocCurveDisplay.from_estimator(svc, X_test, y_test) rfc_disp = RocCurveDisplay.from_estimator(rfc, X_test, y_test, ax=svc_disp.ax_)

@henrystevens3993 Жыл бұрын

Sir in our project we are stuck in a situation of time forecasting using Holt Winters model, the tracking page is giving us 2500 models each time it gets trained, is it possible to get only one model output in tracking page at a cumulative level?

@AshutoshTripathi_AI Жыл бұрын

I think it is possible. You need to customize the code then log the final model under tracking in mlflow.

@Element-fi6lg Жыл бұрын

Where is the 2nd part of the video where you explain how to deploy the trained model?

@AshutoshTripathi_AI Жыл бұрын

Coming soon. Thanks

@Element-fi6lg Жыл бұрын

4 months and still coming...come on man! :) just kidding

@Element-fi6lg Жыл бұрын

Are you available to be hired for the ML projects... pls let know.

@AshutoshTripathi_AI Жыл бұрын

@@Element-fi6lg we can discuss. Could you please drop me a mail with more detail at ashutosh.optimistic@gmail.com

@Element-fi6lg Жыл бұрын

Email sent.

@kashifsadiq4145 3 ай бұрын

Sir I am from Devops background Please start mlops course in hindi

@adityanjsg99 Жыл бұрын

first of a kind in mlflow, this content is not on udemy !

@AshutoshTripathi_AI Жыл бұрын

Thank you @adityanjsg99 glad you liked the content .

@AshutoshTripathi_AI Жыл бұрын

Large Language Model and Langchain based QnA bot building live demo and code explaination: kzfaq.info/get/bejne/ldt0p6h_3Lu3enU.html

@sai199226 Жыл бұрын

Can we also get the slides presented in this playlist for the revision purpose?

@mritunjay3723 Жыл бұрын

From where you copied the code ?

@AshutoshTripathi_AI Жыл бұрын

What do you mean, Mritunjay

@ariesvazquez1349 2 жыл бұрын

The English is so hard to understand

@AshutoshTripathi_AI 2 жыл бұрын

OK If you have any suggestions do let me know I will try to work on that.

@elvigilante99 Жыл бұрын

@@AshutoshTripathi_AI Great work man! Just a suggestion on this comment, create subtitles for native English speakers. Keep posting such videos! 👍

@pritijorvekar6789 2 жыл бұрын

Hi Ashutosh, ML flow concept nicely explained.. I tried to execute notebook but after create_experiment() i got following error ConnectionRefusedError MlflowException: API request to localhost:5000/api/2.0/mlflow/experiments/list failed with exception HTTPConnectionPool(host='localhost', port=5000): Max retries exceeded with url: /api/2.0/mlflow/experiments/list?view_type=ALL (Caused by NewConnectionError(': Failed to establish a new connection: [WinError 10061] No connection could be made because the target machine actively refused it')) can you please help me to resolve this error

@AshutoshTripathi_AI 2 жыл бұрын

Just remove the line mlflow.set_tracking_url("localhost:5000") from cell in[50] This line will be required if you want to use sqlite or any other database as a backend store. Please watch the video then you will be clear. Simply running the notebook might get you an error.

@pritijorvekar6789 2 жыл бұрын

@@AshutoshTripathi_AI Thank you. It is working