About the Intro to Machine Learning category


(Jeremy Howard (Admin)) #1

Welcome to Introduction to Machine Learning for Coders! Please use this category for any questions, issues, comments (and of course answers!) related to this course.

Wiki links

Lesson resources

Course notes from hiromi

Unofficial course notes


Official MOOC launch today!
(Prince Grover) #2

Just curious, is fastai library also available with ~ pip install … ?
I checked but couldn’t install it. Guess it’s not accessible using pip.


(Jeremy Howard (Admin)) #3

Nope, it’s not.


(Danai Avgerinou) #4

I was trying to move files from AWS to my computer and the other way around. I tried this,

scp –i path/mypem.pem –r path/myfile.py ubuntu@ec2-54-190-113-250.us-west-2.compute.amazonaws.com:~

but I get the following error
Permission denied (publickey).
lost connection

even though my path for the file is fine. After some search in the web I tried this,

scp –i path/mypem.pem –r path/myfile.py ubuntu@ec2-54-190-113-250.us-west-2.compute.amazonaws.com aws dns name:~

which gave me another error

ssh: Could not resolve hostname name: nodename nor servname provided, or not known
lost connection

Did anyone encounter anything similar? Any suggestion?


(Prince Grover) #5

Have you tried ‘sudo scp -i …’?


(Danai Avgerinou) #6

Yes and I get the exact same error about permissions


(parrt) #7

You have to reverse the order of the source and target arguments to that SCP command. It also matters whether you do it from your laptop or the AWS computer. Do it from your laptop and reverse the order of the arguments. Your laptop does not have the SSH port open probably


(Shikhar Gupta) #8

scp -i shikhar.pem ec2-user@ec2-54-213-11-248.us-west-2.compute.amazonaws.com:~/data_aws/abc.csv ~/data_local/
this worked for me @Danai


(Danai Avgerinou) #9

Thank you! Got it


(Igor Kasianenko) #12

I’d like to report typo in http://www.fast.ai/2018/09/26/ml-launch/ post:
Lesson 1: “Bull Book for Bulldozers” Kaggle competition,
should be Blue Book


#13

How to use the model we trained on a training set to predict dependent variables in a test set?

If I do m.predict(X_test) on the testing set downloaded from Kaggle, I get an error saying values are stored as string.

I thought that once your had a model you could use it to predict any new data rows whose dependent variable is unknown. How do we use models to predict unknown values then?


(SA) #14

Sir, if the ML course is based on the older version of fastai library then will I be at a disadvantage? Is their going to be an ML course with fastai library v1?
I started the ML course, and will watch the archived versions of DL1 v3 (since it will be 6:30 in the morning and I have to attend college).
My only worry is concerning the older versions of the library.


#15

I’d like to start a course from fast.ai … but which should I do first?

ML or AI ?

What’s the relationship between the two - is it better to have an understanding of one before doing the other? Anything else I should consider in weighing the two options?

Thanks!


#16

Specifically, I’m referring to deep learning vs machine learning


#17

Hi, setup the Jupyter on Crestle but it’s not showing the ML1 course and is only showing DL1. Can anyone help how to fix this…

47%20pm 33%20pm


(Viet Vu) #18

Is there any requirements.txt for the packages that needed for the courses? At first I thought that install fastai should be enough, but there are many, like feather or some ploting utils.

I have to run the notebooks and install each time error show for a missing packages.


(Tamlyn Rhodes) #19

The situation right now is very confusing. fastai v1 is out which is totally incompatible with older versions. (It also is not following semver and has frequent breaking changes in minor updates.)

It should theoretically be possible to clone an older version of the library from the repo and use that with these old notebooks. But I’ve been hacking away at it all evening and not got very far. Maybe someone with a better understanding of Python dependency management could write some instructions for it. Without this, I’d say it’s not worth starting the ML course at this point.

Hopefully when v3 of the DL1 course is over in January, Jeremy will have time to update these notebooks for fastai v1 (which will also hopefully be more stable by then).


#20

Hello, I was wondering if there was going to be a Intro to Machine Learning 2 because all we covered in this course was random forest, some computer vision and some nlp.


#21

Figured this out a while ago and sorry for not responding earlier!

Had registered on Crestle.ai by mistake and it has exactly the same interface and there ml1 course was not listed. Then I registered back on crestle.com and could find the course there.

So it’s a bit confusing having crestle having .ai and .com exactly similar but out of sync.