awni/speech

transducer results

BestSonny opened this issue · 1 comments

@awni Do you have results for the transducer model?

@awni When I set a batch size of 64, I will get "*** Error in `python': double free or corruption (!prev): 0x0000000077e2ec40 ***". And if with a batch size of 128, it will get a "segmentation fault".

I ensure that the GPU memory is large enough and the transducer can work with a small batch size (such as 10). Could you give me any suggestion on solving this problem?