Reproduce the results of pred_and_eval_ss.ipynb

[mmuneeburahman]

I am trying to reproduce the result of pred_and_eval_ss.ipynb. But I am facing this error, refer to the notebook
Notebook Link.

---------------------------------------------------------------------------
ConfigError                               Traceback (most recent call last)
[<ipython-input-3-6dafbbada4cc>](https://localhost:8080/#) in <cell line: 3>()
      1 from rastervision.pytorch_learner import SemanticSegmentationLearner
      2 
----> 3 learner = SemanticSegmentationLearner.from_model_bundle(bundle_uri, training=False)

2 frames
[/usr/local/lib/python3.9/dist-packages/rastervision/pytorch_learner/learner.py](https://localhost:8080/#) in from_model_bundle(cls, model_bundle_uri, tmp_dir, cfg, training, **kwargs)
    967 
    968             config_dict = file_to_json(config_path)
--> 969             config_dict = upgrade_config(config_dict)
    970 
    971             learner_pipeline_cfg: 'LearnerPipelineConfig' = build_config(

[/usr/local/lib/python3.9/dist-packages/rastervision/pipeline/config.py](https://localhost:8080/#) in upgrade_config(config_dict)
    240     """
    241     plugin_versions = config_dict.get('plugin_versions')
--> 242     plugin_versions = upgrade_plugin_versions(plugin_versions)
    243     if plugin_versions is None:
    244         raise ConfigError(

[/usr/local/lib/python3.9/dist-packages/rastervision/pipeline/config.py](https://localhost:8080/#) in upgrade_plugin_versions(plugin_versions)
    218             new_plugin_versions[plugin] = version
    219         else:
--> 220             raise ConfigError(
    221                 'The plugin_versions field contains an unrecognized '
    222                 f'plugin name: {alias}.')

ConfigError: The plugin_versions field contains an unrecognized plugin name: rastervision.gdal_vsi.

[AdeelH]

This is a known issue (4th item in the list in #960). This happens because the model bundle was created in an environment where rastervision_gdal_vsi was installed, but is being loaded in an environment where it has not beel installed (pip install rastervision does not install it by default).

You can work around this by installing rastervision_gdal_vsi. You will need to install gdal before that and pip install gdal usually doesn't work. Add the following to your Colab notebook:

!pip install -q condacolab
import condacolab
condacolab.install()

!conda install -c conda-forge gdal==3.5.2 -y
!pip install rastervision_gdal_vsi

You will have to restart the notebook after running the above.