search

shrimai / Focused-Attention-Improves-Document-Grounded-Generation

MIT License
21 stars 4 forks source link

about DOHA #7

Open lalisaa opened 2 years ago

lalisaa commented 2 years ago

when i run doha.py on cmu_dog, there is an error: division by zero

self.args.train_batch_size=2; self.args.gradient_accumulation_steps=8

    train_batch_size = int(self.args.train_batch_size / self.args.gradient_accumulation_steps)
    num_train_steps = int(len(train_features) / train_batch_size / self.args.gradient_accumulation_steps * self.args.num_train_epochs)