Hi
i have been trying to run this practice lab of week 3, but fail owing to a file not found error which occurs from the Display call at the very start.
This could be linked to incompatible modules
Can you please share which version of python, pip , jupyter notebook etc should be used to build our local environment.
i dont see the environment specs anywhere in the course
since typically after subscription ends we rely only on our local environment, since this does not run, effectively i cannot leverage the course beyond the susbscription
this will prompt many users to switch to other elearnings and stop taking further subscriptions of coursera (i was considering deep learning specialization) but will swithc to OReilly if environment replication locally is a major challnege as it is today
I haven’t tried running the notebooks locally, so if you find any helpful tips on how to do this, please share them with us. See the pinned comment below regarding the C3W3 lab not working in a local environment. I hope this helps!
Thanks. Its actually a nightmare. The course exercises have not been well designed from a learners perspective. Its all built so very well in coursera environment that hte exercise is a cakewalk if replayed there. The moment i have to run it locally, its a compete day wasted with no clues of whats happening. Is it even worth it? I mean, the course doesnt really prepare the student to get independent. You need to make sure that exercises are run locally else there is no learning, you see. We fail to get hte ocnfidence of doing real life project. I doubt if anybody could do it, having done this course. You need to really give it a thought.
Moreoer my subscription ended and I have no access to the lab to actually help me resolve my local issues. Atleast coursera must give me an extension until i get this sorted out.
hi @Ganesh.Rao
Discourse community provides support for learners having issue when they’re doing any assignment. For a learner to practice the assignment codes locally is a task practically I or you as a learner want to get confident to work with, so in the File ==> Open section, they provide all the files one learners needs for successful run of the codes.
Being said that even after downloading all the files, one needs to understand all the libraries and modules comes with version match dependencies which for a learner it matches or not, cannot be handled by deeplearning.AI
Even if your python version is different than the assignment codes were prepared, you would still encounter issue.
For this part of personal practice honestly deeplearning.ai is not responsible.
Now about coursera not extending the course subscription for you to practice is again a time-limited course subscription and cannot be allowed for access without time limit.
I understand your discomfort, but I sincerely suggest you to save your work or lab to work in the local environment.
For better understanding when you are encountering issue while learning labs locally, you can share your issue in discourse community but believe me this doesn’t come under responsibility of deeplearning.AI on how to run code assignment locally as the assignment comes with Code of conduct.
Being said that there learners who still post their query and we as volunteering mentors try our best to address their issue.
Regards
DP
Thanks Deepti.
I believe one crucial thing that we miss in your courses lab is the confidence of running things locally. Now when I followed a simple 30 mins youtube video, i got to know that miniconda environment handles dependencies and it did. That is so crucial for a AI ML learner to know, to be able to confidently work on real life projects. So i believe that was a huge miss which I hopefully expected coursera to address, and not a random youtube video which accidentally we encounter
Nevertheless, I am sure you have chosen the best path at coursera and my views could be entirely personal. But honestly i felt a lot let down and now a lot contented that a random youtube video gave those useful hints and not a subscription which i paid for.
Also can you imagine the waste of time in these threads all of us are doing, discussing over environmental issues the solution of which was as simple as recommending miniconda. Moreover if you give a lifetime susbscription atleast its understood the labs are there forever but having withdrawn it in a few months, you also wash your hands off, when students struggle with local environments. This is certainly not done.
2 Likes
I feel your pain, but the loss of access to lab materials has been consistent from Deeplearning.ai at least since I took my first Specialization over 7 years ago. IIRC the course FAQs recommend downloading files before the subscription terminates for this reason. My lessons learned over the years from first doing it wrong and then turning it into a best practice:
-
always add pip list to a course notebook and get a snapshot of the complete operational context before the subscription ends.
-
always download all files, including data, unit test, helper files, and images before the subscription ends.
-
use virtual environments locally to fine tune and micromanage the package inventories. The requirements to run AI for Medicine locally are different than for Deep Learning which may be different from MLS etc. Keeping them all in the same environment is recipe for trouble. Virtual envs allow multiple, different, internally consistent configurations to not adversely impact one another or your base system. Sometimes there are work arounds or extra steps necessary if you’re on Windows or MacOS, but for the most part, if you use the inventory from 1) above to provision your virtual env, it really is not super difficult.
I wish I had access to not only the versions of the many courses I have completed over the years, but to the updates when they are refreshed and/or bugs are fixed later on. Unfortunately, that isn’t Deeplearning’s business model. I assume it is in part to manage server load; hundreds of thousands of people have taken these courses and giving us all perpetual access is a very different cost structure to undertake.
2 Likes
As I said earlier these are company policy and limitations which we as learners or mentors don’t have any say, I am not defending DeepLearning.AI but providing a knowledge platform is what I appreciate about discourse community.
About the coursera subscription limitation I suppose it comes with financial dependencies. As far as I ca. understand they don’t want to give life time subscription would be again their own policy. Honestly I look forward more on finding something myself
But you raised a really good point about pointing a guidance protocol about details given about these dependencies and to give you some news these points were actually addressed in future updated course like tensorflow developer professional specialisation where they provide a requirement.txt file about all the library and module versions.
But I will surely forward your feedback to staff as we look forward to improving experience for learners.
Regards
DP
Thanks ai_curios. This indeed helps. I missed doing a pip list and in their suggestions they had solely asked to download all files, not do a pip list. I believe that should have come from coursera rather than this community where I have spent almost 3-4 days struggling with this environment issues. With me, i see a whole lot of people struggling with this set up locally topic. It is so crucial that these tips should have been in the course content rather than community particularly if their subscription model restricts lab access.
I was keen on taking the deep learning specialization here but now with the way the courses work, i plan to switch to another eLearning platform where local environment setup is taught, rather than leaving students at the mercy of communities where we have to help each other. I can still understand community support but not something as basic as this which should have come from Coursera. Must say, a major let down. 
thanks Deepti. ai_curios spoke of pip list being run to get a list of preqs. Coursera has never suggested that. All they say is download files which is so obvious. And then we all waste our time here with fixing hte environment locally. The only reason i wasted 3 days is because i wished to converge this logically. I wished to see the lunar lander and i saw that eventually thanks to a random youtube video recommended by our own fellow worried student.
Well, if we have to fend for ourselves or with our fellow students, why not as well swtich to OReilly or other platforms wherein the subscription model is better off. No doubt sandboxes exist but that is for an Year in A Cloudguru and that is good enough time for one to check how to make that run locally. and that is for saving cost on AWS. This environment comes for free locally and you see the major learning is when you do the setup locally. What peopele actually do and obivous are expected to do with the way it was designed is to copy the hints between start code and end code and get certified. The credibility was lost at that moment when we did that. Didtn it?
btw do share your email ids please, deepti and ai_curios. Now that we need to fend for ourselves, why have this baggage of community. we may as well directly talk.
For transparency let me say that I am not now and never have been an employee or compensated in any manner by Deeplearning. Or Coursera. I don’t understand or love every choice either company makes. But I don’t think Coursera has much insight into the course content - they are merely the platform, and manage the enrollment and subscriptions stuff. The individual course developers are responsible for their own content. So maybe not fair to expect Coursera to recommend how students interact with all the diverse content available on the platform.
Additionally, I don’t think of this community forum as baggage. I’ve been voluntarily participating in it for 7 years. I’ve learned a lot along the way, and hopefully a few folks have benefited from experience and ideas I have shared. I think the historical and collective wisdom you find here is quite valuable, and allows Deeplearning to make content available at a ridiculously low cost to us individual subscribers. Especially for the courses taught directly by Professor Ng, a month of access to him (and all of us) for $50 US? Are you kidding me? Customers paid many hundreds for one hour of my time when I was at IBM, so I can only imagine what kind of consulting rate he could command. And to reinforce, I and all the other mentors and forum contributors charge $0.
I started at Coursera and have stayed here because for me the experience was a better value/cost ratio than edx or udemy etc. that I tried. I like the fact that the mentors volunteer to help in a public forum so I can benefit even if I wasn’t in on the original thread - this scales knowledge in a way that point-to-point email doesn’t. At the end of the day, though, you should vote with your eyeballs and your wallet, and find the learning platform that works best for you. Cheers
3 Likes
true that.
to each his own