Clarify OER_DATASET_PATH and contents

[robmarkcole]

The docs are a bit vague on this in FinetuneOlmoEarthSegmentation stating

export OER_DATASET_PATH=/path/to/your/oerun_dataset/folder # Replace with desired dataset folder path

I created a new folder and used that, then prepare_labeled_windows fails with

ValueError: Annotation task features file not found: /teamspace/studios/this_studio/data/oerun_dataset/annotation_task_features.geojson

However I see that file has been created at some point at

olmoearth_projects/docs/tutorials/FinetuneOlmoEarthSegmentation/config/annotation_task_features.geojson

So assume this needs to be copied over to oerun_dataset

[robmarkcole]

Ok the previous command created the geojson, but we should probably pass the path to OER_DATASET_PATH to that command, rather than having the files in the docs

⚡ main ~/olmoearth_projects uv run ./scripts/oer_annotation_creation.py $SRC_DATA_DIR/label_data/Calfire_2020-2025_bbox.gdb --outdir $PROJECT_PATH --id-col polygon_id --taskgeom-col task_geom
Reading file: /teamspace/studios/this_studio/data/label_data/Calfire_2020-2025_bbox.gdb

Successfully loaded 4026 features
Available columns: ['YEAR_', 'STATE', 'AGENCY', 'UNIT_ID', 'FIRE_NAME', 'INC_NUM', 'IRWINID', 'ALARM_DATE', 'CONT_DATE', 'C_METHOD', 'CAUSE', 'COMPLEX_NA', 'COMPLEX_ID', 'OBJECTIVE', 'GIS_ACRES', 'COMMENTS', 'FIRE_NUM', 'Shape_Leng', 'Shape_Area', 'label', 'polygon_id', 'start_time', 'end_time', 'bbox_area_', 'bbox_pix10', 'max_side_p', 'min_side_p', 'task_geom', 'geometry']
Assuming task geometry CRS is the same as the input geometry CRS: EPSG:4326

Label encoding
------------------
Label column: 'label': {'burnt': 0, 'unburnt': 1}

Processing rows: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 4026/4026 [00:13<00:00, 288.83it/s]

✓ Success!
  Wrote: docs/tutorials/FinetuneOlmoEarthSegmentation/config/annotation_task_features.geojson
  Wrote: docs/tutorials/FinetuneOlmoEarthSegmentation/config/annotation_features.geojson

I assume OER_DATASET_PATH should point to olmoearth_projects/docs/tutorials/FinetuneOlmoEarthSegmentation although I am now getting the error:

FileExistsError: [Errno 17] File exists: 'docs/tutorials/FinetuneOlmoEarthSegmentation/config/dataset.json' -> '/teamspace/studios/this_studio/data/oerun_dataset/dataset.json'

But if I delete that file I then get

FileNotFoundError: Required project file not found: docs/tutorials/FinetuneOlmoEarthSegmentation/config/dataset.json

It appears the cause of that error is the use of a relative path for

export PROJECT_PATH=./docs/tutorials/FinetuneOlmoEarthSegmentation/config

If I use an absolute path for PROJECT_PATH the prepare_labeled_windows script runs as expected