Xianyan Jia

+1

same issue

torch fp32 with xla works fine in my experiment. As the inputs is an arithmetic progression sequence, the norm results for the three samples should be exactly the same. The...

seems ResNet50 is well supported and tested. How about alexnet? I'm trying very hard to train alexnet in order to get close to the published 58%+ top1, but can only...

@tfboyd @reedwm thanks for your explanation and clarification. In order to train benchmark with similar accuracy as that in paper, I have fixed the model, data preprocess, some hyperparameters(e.g. learning...

same issue

Make PriorBox a keras layer, will it compute input gradient in backward?

+1

可以看看 NCCL_SOCKET_IFNAME 是否配置正确

you can use tf.train.MonitoredTrainingSession instead of tf.Session , and `global_variables_initializer` is not necessary when using MonitoredTrainingSession, you can refer https://github.com/alibaba/FastNN/blob/73b70c633117ccff4f1a270f461bacb96e0fc4ee/resnet/resnet_dp.py#L67