Closed username123062 closed 3 years ago
Hi, the documentation at the moment is very scarce, sorry. I'm working on it and I'll put all instructions in the README.
In general, training configurations are parsed from yaml files, just like the one at https://github.com/baldassarreFe/ws-vrd/blob/master/config/train.yaml
Then you launch the job as:
python -m xib.train path/to/config.yaml
The config parser is actually quite flexible, it even allows something like this
python -m xib.train config1.yaml config2.yaml session.max_epochs=35
Thanks a lot!
Thanks for sharing your code! You share us with the data preparation and pre-processing. Could you share more instructions for running the /xib/train.py and testing the predicate classification model? Like the following:
python -m xib.preprocessing.vrd \ --confidence-threshold=.3 \ --d2-dir="models/detectron_vrd_train" \ --data-dir="./data" \ --output-dir="data/vrd/processed"
Thanks a lot.