Thanks for the source code provided by the author, but there is such a problem. Whether using single card or multi card training, you can train normally when using batchsize=1. However, when the batchsize is not 1, an error of "stack expectations each tense to be equal size, but got [2, 2400, 3] at entry 0 and [2, 2332, 3] at entry 1" will occur.
How to solve it? I'm sorry to trouble you.
Thanks for the source code provided by the author, but there is such a problem. Whether using single card or multi card training, you can train normally when using batchsize=1. However, when the batchsize is not 1, an error of "stack expectations each tense to be equal size, but got [2, 2400, 3] at entry 0 and [2, 2332, 3] at entry 1" will occur. How to solve it? I'm sorry to trouble you.