Thanks for all. I didn’t expect that was the way to do but I will spend the whole day on it.
I agree, it should be ok
No problem, thanks for all
Thanks for all. I didn’t expect that was the way to do but I will spend the whole day on it.
I agree, it should be ok
No problem, thanks for all
frustration is the word that describes my participation in this challenge. I spent around 1h to build the actual model but days to make a successful submission. Participation should definitely be more straightforward. A must is to enable “debug” with instant error logs. Besides
How did you make the environment.yml file ? You just export it from your current environment ?
hmm… actually are we supposed to train the model or can we upload our own model ?
just exported yml with conda. train locally and inference in docker
The weight of my model is 600 Mo…
@amapic in case your files are larger then 30-50 MB, you will need to use git-lfs for uploading those files. Please read about it here: How to upload large files (size) to your submission
@shivam Hi, thanks for that. Git lfs is working for me but what is git lfs migrate for ? Where must I store the real model ? Is there something to adapt in the submission for that ?
Hi git lfs migrate
is for transferring any older commit to start using lfs. This is useful in case you have lots of older commit (intended/unintended) and want those files to migrate to LFS based in future.
What must we keep from the initial file env.yml (besides aicrowd-api) ?
What’s wrong with my yml file ??
@amapic Your master branch contains the aicrowd-api but you submission branch does not. The environment.yml file in submission-v0.22 does not contain the api.
Could you please clarify on the last date for the contest? On the home page it shows “2 days remaining” but Timeline mentions Jan,17,2020.
Hi @gokuleloop,
Thanks for pointing it out. We have updated the last date to Jan 17, 2020 on website as well.
@ignasimg Hi,thanks for providing your help. What value did you set for AICROWD_TEST_IMAGES_PATH and AICROWD_PREDICTIONS_OUTPUT_PATH ?
do not change the default
What test did you use to detect corrupt images ?