Closed Nomination-NRB closed 4 months ago
I‘m greatful for the training code you provided, but I also have some question:
1.I wonder what the normal loss curve looks like(like start at? end at?). 2.How many epochs should I set. 3.Is the result shown in the readme document the result of test.py or the log during training.
1.The batchsize is 1, the loss is unstable. 2. I train it for 10w steps, but only 5000 steps it can realize a result not bad, You can save 10000 steps and test the result. 3. The readme document is the result of test.py
I‘m greatful for the training code you provided, but I also have some question: 1.I wonder what the normal loss curve looks like(like start at? end at?). 2.How many epochs should I set. 3.Is the result shown in the readme document the result of test.py or the log during training.
1.The batchsize is 1, the loss is unstable. 2. I train it for 10w steps, but only 5000 steps it can realize a result not bad, You can save 10000 steps and test the result. 3. The readme document is the result of test.py
got it, thanks a lot
I‘m greatful for the training code you provided, but I also have some question:
1.I wonder what the normal loss curve looks like(like start at? end at?). 2.How many epochs should I set. 3.Is the result shown in the readme document the result of test.py or the log during training.