Rectified Flow

python run.py flow_main

---

Note:

• The loss is set to mse - var. This is to get a sense of how good the model is trained by looking at the loss (loss = 0 means fully trained). Loss sometimes goes negative as a result of stochastic nature of the mean and variance estimates.

• You might want to increase config.plot_every to get faster training.