`hi-ml-azure` not specified in environment, but needed in `azure_util.py`
peterhessey opened this issue · comments
Is there an existing issue for this?
- I have searched the existing issues
Bug summary
When importing any components from InnerEye/Azure/azure_util.py
a ModuleNotFoundError is raised if one has not installed the hi-ml-azure
package manually using pip.
Code for reproduction
Run in cmd line:
python -c "from InnerEye.ML.photometric_normalization import PhotometricNormalization
Actual outcome
ModuleNotFoundError
Error messages
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "/home/azureuser/git/InnerEye-DeepLearning/InnerEye/ML/photometric_normalization.py", line 13, in <module>
from InnerEye.ML.config import PhotometricNormalizationMethod, SegmentationModelBase
File "/home/azureuser/git/InnerEye-DeepLearning/InnerEye/ML/config.py", line 25, in <module>
from InnerEye.ML.deep_learning_config import ModelCategory
File "/home/azureuser/git/InnerEye-DeepLearning/InnerEye/ML/deep_learning_config.py", line 16, in <module>
from InnerEye.Azure.azure_util import DEFAULT_CROSS_VALIDATION_SPLIT_INDEX, RUN_CONTEXT, is_offline_run_context
File "/home/azureuser/git/InnerEye-DeepLearning/InnerEye/Azure/azure_util.py", line 16, in <module>
from health_azure.utils import create_run_recovery_id, get_aml_run_from_run_id
ModuleNotFoundError: No module named 'health_azure'
Expected outcome
Successful import
System info
No response
I think that the problem here is that the conda environment creation doesn't include setup of any of the packages that come with the InnerEye repository. With a suitable setup.py
file, adding - -e .
as the final pip
dependency could be a solution.
On Mac, with the current setup.py
, and in the InnerEye environment, the command:
pip install -e .
results in the error:
ERROR: Could not find a version that satisfies the requirement cucim==21.10.1 (from innereye==0.0.1) (from versions: none)
ERROR: No matching distribution found for cucim==21.10.1 (from innereye==0.0.1)