General course chat

matdmiller · January 3, 2019, 12:48am

Were you able to resolve your issue? I am having the same problem on my local machine.

amitkayal · January 3, 2019, 4:11pm

Hello All,

I am getting following error while apply TTA. It is a multi class image classification.

log_preds,y = learn50_dog_breed.TTA()
probs = np.mean(np.exp(log_preds),0)

================================
TypeError Traceback (most recent call last)
in ()
----> 1 probs = np.mean(np.exp(log_preds),0)

/mnt/disks/user/anaconda3/lib/python3.6/site-packages/numpy/core/fromnumeric.py in mean(a, axis, dtype, out, keepdims)
2952 pass
2953 else:
-> 2954 return mean(axis=axis, dtype=dtype, out=out, **kwargs)
2955
2956 return _methods._mean(a, axis=axis, dtype=dtype,

TypeError: mean() missing 3 required positional argument: “dim”, “keepdim”, “dtype”

Xiwang · January 3, 2019, 5:13pm

Yes, Use the updated instruction at

github.com

fastai/fastai/blob/master/README.md#installation

[![Build Status](https://dev.azure.com/fastdotai/fastai/_apis/build/status/fastai.fastai)](https://dev.azure.com/fastdotai/fastai/_build/latest?definitionId=1)
[![pypi fastai version](https://img.shields.io/pypi/v/fastai.svg)](https://pypi.python.org/pypi/fastai)
[![Conda fastai version](https://img.shields.io/conda/v/fastai/fastai.svg)](https://anaconda.org/fastai/fastai)

[![Anaconda-Server Badge](https://anaconda.org/fastai/fastai/badges/platforms.svg)](https://anaconda.org/fastai/fastai)
[![fastai python compatibility](https://img.shields.io/pypi/pyversions/fastai.svg)](https://pypi.python.org/pypi/fastai)
[![fastai license](https://img.shields.io/pypi/l/fastai.svg)](https://pypi.python.org/pypi/fastai)

# fastai

The fastai library simplifies training fast and accurate neural nets using modern best practices. See the [fastai website](https://docs.fast.ai) to get started. The library is based on research into deep learning best practices undertaken at [fast.ai](http://www.fast.ai), and includes \"out of the box\" support for [`vision`](https://docs.fast.ai/vision.html#vision), [`text`](https://docs.fast.ai/text.html#text), [`tabular`](https://docs.fast.ai/tabular.html#tabular), and [`collab`](https://docs.fast.ai/collab.html#collab) (collaborative filtering) models. For brief examples, see the [examples](https://github.com/fastai/fastai/tree/master/examples) folder; detailed examples are provided in the full [documentation](https://docs.fast.ai/). For instance, here's how to train an MNIST model using [resnet18](https://arxiv.org/abs/1512.03385) (from the [vision example](https://github.com/fastai/fastai/blob/master/examples/vision.ipynb)):

```python
untar_data(MNIST_PATH)
data = image_data_from_folder(MNIST_PATH)
learn = create_cnn(data, tvm.resnet18, metrics=accuracy)
learn.fit(1)
```

## Note for [course.fast.ai](http://course.fast.ai) students

This file has been truncated. show original

cbaumgartner · January 3, 2019, 8:28pm

Is anyone else having audio problems with v3 Lesson 7? I can’t really understand anything Jeremy is saying.

avinashj · January 5, 2019, 4:06am

audio working fine for me.

sandeepsign · January 7, 2019, 7:43am

Why in v2 Part 1, jeremy did not discuss TTA at all? has TTA been made implicit in v3?

kurianbenoy · January 9, 2019, 12:42pm

create_cnn() method is not working in Kaggle kernel. Couldn’t solved this issue for a long time now. I wanted to use kaggle dataset, but using kaggle library in google colabs is not possible.

Can anyone help me

amitkayal · January 9, 2019, 5:09pm

Hello,

I have built a model and it uses image normalization. Now i am developing a web app for this. Do i need to normalize the input image which are being fed into web app for classification?

Thanks

Bhuvana_ka · January 9, 2019, 5:41pm

Hi,

This link has details on getting data from Kaggle to Colab

You can also look at the code here in my github, where i have downloaded dataset form Kaggle.

Preka · January 11, 2019, 12:22pm

Yes, I think you should do so in order to help the model you have trained to get the best results. What normalization does is just reduce all inputs into a unique scale. Most of the time, this unique scale corresponds to normal distribution of mean 0 and standard deviation of 1.

hellobharadwaj · January 11, 2019, 9:19pm

Where are the resources for part 2 of the program(dates and times etc). How do we register? If it is not yet available, when will it be available?

edwardjross · January 13, 2019, 9:01am

There is a in-person registration form linked from fast.ai homepage.

This states Part 2 will run March 18 - April 30 (7 weeks).

It’s currently not decided whether it will be possible to register for a livestream like part 1.

dhoa · January 14, 2019, 9:06am

The GCP discussion thread seems disappear right ? I cannot access to it anymore

amitkayal · January 14, 2019, 10:18am

Hello,

I am using ImageItemList.from_csv to read my data and my aim is to display the file name for which prediction went wrong from validation set. So i tried following but it returned me only the index image.

interpCat = ClassificationInterpretation.from_learner(learn50_cat_breed)
interpCat.most_confused(min_val=2)
interpCat.plot_top_losses(9, figsize=(15,11))
losses, indexes = interpCat.top_losses()

top_loss_paths = data_cat_breed.valid_ds.x[indexes]

Here top_loss_paths returning me indexes from validation data set which have gone wrong and i can see the images by following simple command.

data_cat_breed.valid_ds.x[130]

I need to get file name but unable to get the name from index. Can anyone please help me how to get the filename?

technocratius · January 22, 2019, 9:37pm

Yeah I’m having similar issues on android (audio works fine on PC…)

cbaumgartner · January 22, 2019, 9:40pm

I haven’t figured out the cause yet. It works fine on my iPhone and on the computers of my colleagues sitting right next to me, but not on my laptop for some reason. I tried all the standard browser troubleshooting (clearing cache, different browsers, etc.) nothing worked. No idea whats going on.

technocratius · January 22, 2019, 10:13pm

If possible, it gets even weirder here: so I only have the issue when I play over my phone’s speaker. Other yt videos play just fine over the speaker.if I then connect headphones the sound from the lesson is perfectly fine…

No clue, I guess we’ll never know haha…

jithinrocs · January 24, 2019, 2:31pm

Going through the video lectures, several lecture notes caught my attention. It was easy to search for and find topics. One which caught my eye was that shared by @hiromi here: https://github.com/hiromis/notes/

Here’s a little change (only for lesson 1) that I made and put them under Github pages. See example: https://jithinjk.github.io/notes/Lesson1.md.html

I used Markdeep here.

Please let me know what you think about it? Is it a good idea to add similar pages to http://docs.fast.ai for each lesson?

xjdeng · January 25, 2019, 4:45pm

I’ve got a 16 hour flight coming up… downloading all of the videos to watch on the flight as we speak… Many thanks to Jeremy and his team!

Alican1 · January 25, 2019, 5:33pm

Hello,

Just started setting up the new course. When I use the promo code for paperspace it says " Error! Could not claim promo code (code is expired)". Just wanted to let you know.

Cheers,
Alican