@SeanNaren I want to know whether CTCLoss in pytorch support multi-gpu. When I use CTCLoss trainig crnn model, I find CTCLoss can't support multi-gpu, is it right?
I'm not sure if the actual loss can be put inside the data parallel table, but the current implementation in the master repo should work! What error are you running into?
@SeanNaren I want to know whether CTCLoss in pytorch support multi-gpu. When I use CTCLoss trainig crnn model, I find CTCLoss can't support multi-gpu, is it right?