I tried to reproduce the steps on: http://mxnet.incubator.apache.org/faq/finetune.html on a Amazon Deep Learning AMI V5.0 instance, but could not get it to work past im2rec.py.
Here is the error message:
(mxnet_p27) :~$ python /home/ubuntu/anaconda3/envs/mxnet_p27/lib/python2.7/site-packages/mxnet/tools/im2rec.py --list True --recursive True caltech-256-60-train caltech_256_train_60
usage: im2rec.py [-h] [–list] [–exts EXTS [EXTS …]] [–chunks CHUNKS]
[–train-ratio TRAIN_RATIO] [–test-ratio TEST_RATIO]
[–recursive] [–no-shuffle] [–pass-through]
[–resize RESIZE] [–center-crop] [–quality QUALITY]
[–num-thread NUM_THREAD] [–color {-1,0,1}]
[–encoding {.jpg,.png}] [–pack-label]
prefix root
im2rec.py: error: unrecognized arguments: caltech-256-60-train caltech_256_train_60
I know MXNet is still in incubator stage, but I am very disappointed by the fact that even the simplest thing on its tutorial pages failed to work. Most real open source projects manage to do a much better job. One would assume that companies like Amazon with all its resources and the bonfire on AWS/MXNet could afford to hire someone competent enough to keep the basic decency and consistency on its project. What a disgrace.
Thanks for reporting this bug and sorry you had a bad experience. Apache MXNet is developed by hundreds of collaborators, a lot of them contributing to the project in their own time and not affiliated with Amazon, please remember to remain civil and courteous.
The solution to your problem, as indicated in the help documentation of im2rec.py (see your post above), is to use this notation instead:
python /home/ubuntu/anaconda3/envs/mxnet_p27/lib/python2.7/site-packages/mxnet/tools/im2rec.py --list --recursive caltech-256-60-train caltech_256_train_60
The change is related to this commit that improved the command line arguments style of im2rec.py
The tutorials are versioned, and a fix had been put in already, but is linked to the ‘master’ version. We are looking into pointing tutorials only to master, to avoid poor experiences like the one you had.
Follow this link to have the latest version of the fine-tuning tutorial (or use the Version drop-down on the website):
https://mxnet.incubator.apache.org/versions/master/faq/finetune.html
This led me to investigate and I found a different example in the github repo that had been missed when upgrading to the new syntax: https://github.com/apache/incubator-mxnet/pull/10192 . Thanks
2 Likes