Chevron Left
返回到 Production Machine Learning Systems

學生對 Google 云端平台 提供的 Production Machine Learning Systems 的評價和反饋

825 個評分
91 條評論


In the second course of this specialization, we will dive into the components and best practices of a high-performing ML system in production environments. Prerequisites: Basic SQL, familiarity with Python and TensorFlow...



Unlike pure technical courses, this one specially packs you with knowledge that you may find yourself face to. The course is really well designed and the content is crystal clear, just Awesome !


It is very good course, gives good overview over large ML systems on cloud, a lot of examples from real implementations gives good understunding about problematics in projects realisations


76 - Production Machine Learning Systems 的 90 個評論(共 90 個)

創建者 Santosh L






創建者 choisungwook



創建者 길경완



創建者 Rebecca S


Some of the content was really interesting, particularly about the hybrid ML systems, dynamically training models, distributed training and data parallelism, but overall, the information was mostly high level with few exercises or labs to delve into actually designing and implementing this stuff. I wrote lots of notes and found myself asking 'how' a lot with no answer. It also felt like I was constantly being pitched to buy and use GCP services. And finally, to actually build a product off these tools that could be considered 'production', I don't think having a bunch of notebooks and random CLI commands to launch stuff is the most robust and traceable architecture. Great presenters though, really liked your style folks!

創建者 Harold M


Overall rating is 3 out of 5, as I expected more of the initial line in the first course. The optional Kubeflow lab has issues, as the ksonnet apply command line halts. Also, the last lab was expected to allow the student to code more, as this is the only way to make a person to gain more insights on the architecture.

創建者 Lloyd P


The module on hybrid systems was weak. The time it would take to cover the material would be prohibitive so why do the intro that then apologize for not having the time to explain the material. Leave it out...

創建者 JJ


While there is definitely some good and useful content in this course, not all of the material is useful. ~40% of the course felt like a sales pitch, at least to me.

創建者 Junhwan Y


This course include deep contexts about Machine Learning. But, It's somewhat boring.

創建者 김유상


Some errors in Kubeflow quicklabs.

創建者 KimNamho


thank you

創建者 Melissa K R


I was really hoping I'd gain some real practical skills and knowledge about the different aspects of building and deploying a machine learning model on GCP. Even though a lot of real estate was covered in this course, most of it was theoretical, and I cannot say that I "really" learned how to implement them if I were working on a big machine learning project, which was exactly why I took this course. The only labs that had some practical aspects to them were also disappointing; one only looked at a number of Java modules and the other was a demo of Kubeflow that I couldn't follow at all, and was different from the lab itself! First off, the fact that Java was used in the first instance took me by surprise, and I wonder why the same thing couldn't be accomplished with Python. I have zero knowledge of Java and that was uncalled for, but tried very hard to make sense of the code. But I won't definitely be able to write it myself. And in the case of the last demo, I simply couldn't understand what the instructor was doing and where!

I expected a much much much higher standard from this course, but overall it was quite disappointing and I cannot say I took anything away from this course other than some theoretical concepts about various subtleties when it comes to ML on GCP! It would've really really helped if there were more actual lab work included in this course, just like the previous one and each concept was accompanied by one such hands-on lab, and concepts were explained step by step.

The other thing that was very odd to me (and is the same for every other course in this specialization) is that a ton of material is squeezed in Two Weeks. It would've helped if they were separated over multiple weeks. This change in the organization of material would really help learners to visualize the flow of topics. Right now, it all seems a load of crammed topics that have been merely glossed over!

創建者 Alireza K


The Qwiklabs should be more than copy pasting commands. Also I think this course is suitable for people with many years of experience in software development not people like me just came out from university!

創建者 Hamze G


This course is assuming too much IT engineering skill which could be challenging if you are not an IT professional.

創建者 Maxim


This specialization consists of 5 courses:

Course1: End-to-End Machine Learning with TensorFlow on GCP

Course2: Production Machine Learning Systems

Course3: Image Understanding with TensorFlow on GCP

Course4: Sequence Models for Time Series and Natural Language Processing

Course5: Recommendation Systems with TensorFlow on GCP

In specialization's FAQ say nothing about "audit" option. Are You know what is it ? "Audit" means that You can use course video material even after You subscriptions ended.

By fact, only "Course 1" has such ability. Before pay for specialization, carefully check FAQ for EACH separated course in specialization:

courses 2-5 has special items in FAQ:

"Why can’t I audit this course?

This course is one of a few offered on Coursera that are currently available only to learners who have paid or received financial aid, when available.


"Who have paid" means that after You subscriptions ended, you lost access to video materials in this courses.


1 star only for "Audit", content and lecturers are rated higher - at least 4 stars