ModuleNotFoundError: No module named 'packaging'

[Licini]

Your issue may already be reported!
Please search on the issue tracker before creating one.

Context

• torchserve version: 0.2.1
• torch version: 1.6.0
• Operating System and version: Ubuntu 20.04 with Conda python3.8

Your Environment

• Installed using source? [yes/no]: No
• Are you planning to deploy it using docker container? [yes/no]: Yes
• Is it a CPU or GPU environment?: GPU
• Using a default/custom handler? [If possible upload/share custom handler/model]: Object_Detector
• What kind of model is it e.g. vision, text, audio?: Vision, MaskRCNN/FastRCNN

Current Behavior

Models using Object_Detector not able to load. Error log suggesting ModuleNotFoundError: No module named 'packaging'

Possible Solution

pip install packaging solved the issue. Please consider adding it to dependencies

Failure Logs [if any]

ts_log copy.log

[harshbafna]

@Licini: The TorchServe version reported by you is 0.2.1 and the same can be seen in the provided logs as well. However, the official latest version of TorchServe is 0.2.0 only. Is this some custom implementation of TorchServe that you are using?

The packaging module used in object_detector default handler is already a part of TorchServe dependency defined in setup.py

[Licini]

@Licini: The TorchServe version reported by you is 0.2.1 and the same can be seen in the provided logs as well. However, the official latest version of TorchServe is 0.2.0 only. Is this some custom implementation of TorchServe that you are using?

The packaging module used in object_detector default handler is already a part of TorchServe dependency defined in setup.py

@harshbafna Thanks for the quick reply, I used conda install torchserve. It is mysteriously one version ahead ...
https://anaconda.org/pytorch/torchserve

[harshbafna]

@Licini: Thanks for the update on the steps used for installation. I will try and reproduce this at my end and revert.

[harshbafna]

@Licini: Thanks for reporting this. I can confirm this is a bug with the conda binary and the packaging module does not get installed as part of TorchServe dependency. We will fix this in the next release.