2020-09-15-deep-learning-flux tutorial crashes

[pojeda]

Hi,

I tried the tutorial on this repo that is mentioned in the title. I am using Julia v. 1.8.5, and

[052768ef] CUDA v4.0.1
[587475ba] Flux v0.13.12
[eb30cadb] MLDatasets v0.7.9

ERROR: LoadError: Scalar indexing is disallowed.
Invocation of getindex resulted in scalar indexing of a GPU array.
This is typically caused by calling an iterating implementation of a method.
Such implementations do not execute on the GPU, but very slowly on the CPU

According to the line numbers (in my file script), the issue seems to come from:

for i in 1:10
preds = m(test[i][1])
lab = test[i][2]
for j = 1:1000
pred_class = findmax(preds[:, j])[2]
actual_class = findmax(lab[:, j])[2]
if pred_class == actual_class
class_correct[pred_class] += 1
end
class_total[actual_class] += 1
end
end

Besides this, I saw the warning:
┌ Warning: CIFAR10.testdata() is deprecated, use CIFAR10(split=:test)[:] instead.

/Pedro

[mcabbott]

I presume you mean http://fluxml.ai/tutorialposts/2020-09-15-deep-learning-flux/ . Unfortunately tutorials there have been allowed to rot, and ought to be deleted.

Some have more recently updated cousins here: https://github.com/FluxML/model-zoo#examples-in-the-model-zoo . Those should all have project/manifest files, so that julia --project loads a known-good set of versions. The version of Flux is noted for each; the ones which say v0.13 will work on latest Flux as above (i.e. without using their own manifest). Ideally they would all be so updated, of course...

[pojeda]

I read about the license of Flux but I wanted to double check with you if it is fine if I take the example you wrote and copy/paste into some material I am developing for a course which will be publicly available?