Azure Machine Learning
An Azure machine learning service for building and deploying models.
2,729 questions
AutoML MLFlow Batch Endpoint - Image build failed
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(204.8, 91%, 29%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EM%3C/text%3E%3C/svg%3E)
Michael
0
Reputation points
When deploying an MLFlow model generated by Azure AutoML as a batch endpoint, using no-code option, the following "Image build failed" error occurs:
ERROR: Ignored the following versions that require a different python version: 0.7 Requires-Python >=3.6, <3.7; 0.8 Requires-Python >=3.6, <3.7; 1.19.0 Requires-Python >=3.5,<3.8; 1.20.0 Requires-Python >=3.5,<3.8; 1.20.0.post1 Requires-Python >=3.5,<3.8; 1.21.0 Requires-Python >=3.5,<3.8; 1.21.0.post1 Requires-Python >=3.5,<3.8; 1.22.0 Requires-Python >=3.5,<3.8; 1.22.0.post1 Requires-Python >=3.5,<3.8; 1.22.0.post2 Requires-Python >=3.5,<3.8; 1.22.0.post2 Requires-Python >=3.6,<3.8; 1.23.0 Requires-Python >=3.5,<3.8; 1.23.0.post1 Requires-Python >=3.5,<3.8; 1.24.0 Requires-Python >=3.5,<3.8; 1.25.0 Requires-Python >=3.5,<3.8; 1.25.0.post1 Requires-Python >=3.5,<3.8; 1.26.0 Requires-Python >=3.5,<3.8; 1.26.0.post1 Requires-Python >=3.5,<3.8; 1.27.0 Requires-Python >=3.5,<3.8; 1.27.0.post1 Requires-Python >=3.5,<3.8; 1.27.0.post2 Requires-Python >=3.5,<3.8; 1.28.0 Requires-Python >=3.5,<3.8; 1.28.0.post1 Requires-Python >=3.5,<3.8; 1.28.0.post2 Requires-Python >=3.5,<3.8; 1.29.0 Requires-Python >=3.6,<3.8; 1.29.0.post1 Requires-Python >=3.6,<3.8; 1.30.0 Requires-Python >=3.6,<3.8; 1.31.0 Requires-Python >=3.6,<3.8; 1.32.0 Requires-Python >=3.6,<3.8; 1.33.0 Requires-Python >=3.6,<3.8; 1.33.1 Requires-Python >=3.6,<3.8; 1.33.1.post1 Requires-Python >=3.6,<3.8; 1.34.0 Requires-Python >=3.6,<3.8; 1.34.0.post1 Requires-Python >=3.6,<3.8; 1.34.1 Requires-Python >=3.6,<3.8; 1.34.1.post1 Requires-Python >=3.6,<3.8; 1.35.0 Requires-Python >=3.6,<3.8; 1.35.1 Requires-Python >=3.6,<3.8; 1.36.0 Requires-Python >=3.6,<3.8; 1.36.1 Requires-Python >=3.6,<3.8; 1.37.0 Requires-Python >=3.6,<3.8; 1.38.0 Requires-Python >=3.6,<3.8; 3.2.0.dev1 Requires-Python >=3.9
ERROR: Could not find a version that satisfies the requirement en-core-web-sm==2.1.0 (from versions: none)
ERROR: No matching distribution found for en-core-web-sm==2.1.0
{count} votes
-
romungi-MSFT 43,696 Reputation points Microsoft Employee2023-04-20T07:51:57.98+00:00 @Michael I think in this case Azure ML was unable to generate environment with the required dependency of Spacy. This environment generation happens by reading the conda dependencies specified in the MLflow model. However, such action is performed at the best of its knowledge and there may be cases when the detection doesn't reflect your intentions or requirements. This is documented here from MLFlow deployment guidelines.
In such cases the recommendation is to use logging models with a custom conda dependencies definition which means that the cli or SDK method is required to deploy the MLFlow models.
Another workaround could be to use a scoring script with custom environment from CLI or SDK. This can help to use the model already generated through studio and you can deploy it by downloading the artifacts and updating conda.yml and then creating a batch deployment.
Sign in to comment