Is it worth adding the Gaussian score identity to `flow.py`, so the score for the flow-matching models can be evaluated?
Is it worth adding the Gaussian score identity to
flow.py, so the score for the flow-matching models can be evaluated?