Closed manisnesan closed 2 years ago
@muellerzr - I created this simple PR to understand first time contributor workflow.
I'd rather we state exactly how they should install it. E.g. pip install -e .[dev]
As those should be valid in the setup.py. if you need to change the requirements in settings.ini, don't hesitate to do that.
https://discord.com/channels/689892369998676007/933769297052975194/942610147614793728 - Note to self about Error in installing Dev.
Capturing the gist from discord
fastai/fastai
. The image is built using docker-containers repository. This image set to use python 3.8. Hence CI only tests with python version 3.8. min_python = 3.6
. Ideally we may want to run test-notebooks on python versions (3.6, 3.7, 3.8, 3.9) that fastai support. EDIT: Updated the reason why CI did not catch this dependency issue.
@manisnesan proposed solution I'd love to see:
@muellerzr
I created a new conda environment and did pip install -e ".[dev]"
for both python versions 3.6 & 3.8. Both were successfully able to install azureml-sdk
along with transitive dependencies (azureml-train-automl-client -> this is the one you mentioned having trouble with). Could you provide the python version you are having issues with??
Python 3.6
(azureml_test_3_6) ➜ fastai-docment-sprint git:(manisnesan_test_dep) ✗ python --version
Python 3.6.13 :: Anaconda, Inc.
$ pip freeze | grep azure
azure-common==1.1.28
azure-core==1.21.1
azure-graphrbac==0.61.1
azure-identity==1.7.0
azure-mgmt-authorization==0.61.0
azure-mgmt-containerregistry==8.2.0
azure-mgmt-core==1.3.0
azure-mgmt-keyvault==9.3.0
azure-mgmt-resource==20.1.0
azure-mgmt-storage==19.0.0
azureml-automl-core==1.38.0
azureml-core==1.38.0.post2
azureml-dataprep==2.26.0
azureml-dataprep-native==38.0.0
azureml-dataprep-rslex==2.2.0
azureml-dataset-runtime==1.38.0
azureml-pipeline==1.38.0
azureml-pipeline-core==1.38.0
azureml-pipeline-steps==1.38.0
azureml-sdk==1.38.0
azureml-telemetry==1.38.0
azureml-train-automl-client==1.38.0.post2
azureml-train-core==1.38.0
azureml-train-restclients-hyperdrive==1.38.0
msrestazure==0.6.4
Python 3.8
(azureml_test) ➜ fastai-docment-sprint git:(manisnesan_test_dep) ✗ python --version
Python 3.8.12
(azureml_test) ➜ fastai-docment-sprint git:(manisnesan_test_dep) ✗ pip freeze | grep azure
azure-common==1.1.28
azure-core==1.21.1
azure-graphrbac==0.61.1
azure-identity==1.7.0
azure-mgmt-authorization==0.61.0
azure-mgmt-containerregistry==8.2.0
azure-mgmt-core==1.3.0
azure-mgmt-keyvault==9.3.0
azure-mgmt-resource==20.1.0
azure-mgmt-storage==19.0.0
azureml-automl-core==1.38.0
azureml-core==1.38.0.post2
azureml-dataprep==2.26.0
azureml-dataprep-native==38.0.0
azureml-dataprep-rslex==2.2.0
azureml-dataset-runtime==1.38.0
azureml-pipeline==1.38.0
azureml-pipeline-core==1.38.0
azureml-pipeline-steps==1.38.0
azureml-sdk==1.38.0
azureml-telemetry==1.38.0
azureml-train-automl-client==1.38.0.post2
azureml-train-core==1.38.0
azureml-train-restclients-hyperdrive==1.38.0
msrestazure==0.6.4
The root cause is azureml-train-automl-client
requires Python >=3.6, <3.9. Hence it may not work for python environments greater than 3.9.
Based on this, I will go with
- Eliminate (for now) having azureml in the dev_requirements in settings.
- State that when building the docs, ignore errors for azure integration.
@muellerzr - Completed the steps.
@manisnesan don't forget to also update index.ipynb, as that's what builds the README
@muellerzr - My bad. Now I know why there is a conflict.:) I will update index.original.ipynb with the changes.
Closing this PR. See https://github.com/muellerzr/fastai-docment-sprint/pull/39
PR to update README