Tao PR

looks like version disparity to me. did you install the python packages by the requirement.txt? ``` pip install -r requirements.txt ```

very likely your pillow package is running on a different version than the code expects.

All subdirectories of `data/*` can be removed manually in case you want to. However, only `data/raw` is where the raw training data is kept before expanding into a larger training...

Exactly, all training vectors and their labelled values are kept as `trainset.csv`.

Hello @loveJasmine Two fun facts might surprise you. 1. The issue is caused by a too-heavy concurrent data loading from MongoDB. So the cursor is busy when it iterates to...

Warnings are fine and ignorable. The error on the bottom part means you are querying for the word which doesn't exist in the vector space. I would suggest you to...

as suggested, have you tried updating the query function in `pylib/knowledge/datasource.py#L20` so it becomes?: ``` query = self.src.find(conditions) if skip==0 else self.src.find(filter=conditions,skip=skip,no_cursor_timeout=True) ``` With the cursor timeout option added in...

interesting. can you please let me know your mongodb version?

Based on your output, the word space data `graph-data.js` should be written with 118 nodes and 100 links. Can you try opening `html/graph-universe.html`? The second part of your output suggests...

`build_index.py` will generate the index data onto your OrientDB. The output database will be created as `vorindex`. Can you check if it exists and whether it has some nodes/edges inside?