cjolivier01 commented on issue #8751: Distributed Training has inverse results when imported (8 GPUS is slower than 1!) URL: https://github.com/apache/incubator-mxnet/issues/8751#issuecomment-346670273 I have twp GTX 1080's on my home machine here, and I get roughly the same speed on two GPUs: file_1.py /usr/bin/python2.7 /home/coolivie/src/DeepLearning/python/file_1.py --gpus 2 Defining network Sequential( (0): Conv2D(None -> 20, kernel_size=(3, 3), stride=(1, 1)) (1): MaxPool2D(size=(2, 2), stride=(2, 2), padding=(0, 0), ceil_mode=False) (2): Conv2D(None -> 50, kernel_size=(5, 5), stride=(1, 1)) (3): MaxPool2D(size=(2, 2), stride=(2, 2), padding=(0, 0), ceil_mode=False) (4): Flatten (5): Dense(None -> 128, Activation(relu)) (6): Dense(None -> 10, linear) ) running with 2 GPUs Running on [gpu(0), gpu(1)] loading mnist mnist loaded Batch size is 128 initalizing parameters initalizing trainer Epoch 0, training time = 6.3 sec Validation accuracy = 0.9525 Epoch 1, training time = 5.9 sec Validation accuracy = 0.9792 Epoch 2, training time = 6.0 sec Validation accuracy = 0.9815 Epoch 3, training time = 5.9 sec Validation accuracy = 0.9838 Epoch 4, training time = 5.9 sec Validation accuracy = 0.9825 file_2.py [GO]: Parallel Run Running on 2 gpus [gpu(0), gpu(1)] [INIT]: net parameters [INIT]: trainer Epoch 0, training time = 6.8 sec Validation Accuracy = 0.0000 Epoch 1, training time = 6.0 sec Validation Accuracy = 0.0000 Epoch 2, training time = 6.0 sec Validation Accuracy = 0.0000 Epoch 3, training time = 6.0 sec Validation Accuracy = 0.0000 Epoch 4, training time = 6.4 sec Validation Accuracy = 0.0000 Epoch 5, training time = 6.5 sec Validation Accuracy = 0.0000
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services