File size: 339 Bytes
57bdca5 |
1 2 3 4 5 6 |
For example, if a problem starts happening at batch number 150. So you can dump traces for batches 149 and 150 and compare where numbers started to diverge. You can also specify the batch number after which to stop the training, with: python debug_overflow = DebugUnderflowOverflow(model, trace_batch_nums=[1, 3], abort_after_batch_num=3). |