Open zhaofenqiang opened 6 years ago
Only port the ActivationLayer, including ReLU and softmax funciton. First forward was normal, second forward crashed.
Program ouput:
zfq@zfq-pi:~/mobile$ ./inference --merged_model ./mobilenet_flowers102.paddle --input_size 150528 I0411 06:22:18.676442 32307 Util.cpp:166] commandline: Time of init paddle 2.93423 ms. Time of create from merged model file 710.877 ms. image conv_0 batch_norm_0 running relu forward... running on acl... conv_1 batch_norm_1 running relu forward... running on acl... conv_2 batch_norm_2 running relu forward... running on acl... conv_3 batch_norm_3 running relu forward... running on acl... conv_4 batch_norm_4 running relu forward... running on acl... conv_5 batch_norm_5 running relu forward... running on acl... conv_6 batch_norm_6 running relu forward... running on acl... conv_7 batch_norm_7 running relu forward... running on acl... conv_8 batch_norm_8 running relu forward... running on acl... conv_9 batch_norm_9 running relu forward... running on acl... conv_10 batch_norm_10 running relu forward... running on acl... conv_11 batch_norm_11 running relu forward... running on acl... conv_12 batch_norm_12 running relu forward... running on acl... conv_13 batch_norm_13 running relu forward... running on acl... conv_14 batch_norm_14 running relu forward... running on acl... conv_15 batch_norm_15 running relu forward... running on acl... conv_16 batch_norm_16 running relu forward... running on acl... conv_17 batch_norm_17 running relu forward... running on acl... conv_18 batch_norm_18 running relu forward... running on acl... conv_19 batch_norm_19 running relu forward... running on acl... conv_20 batch_norm_20 running relu forward... running on acl... conv_21 batch_norm_21 running relu forward... running on acl... conv_22 batch_norm_22 running relu forward... running on acl... conv_23 batch_norm_23 running relu forward... running on acl... conv_24 batch_norm_24 running relu forward... running on acl... conv_25 batch_norm_25 running relu forward... running on acl... conv_26 batch_norm_26 running relu forward... running on acl... pool_0 fc_layer_0 running softmax act.... running on acl... Time of first run time 2969.11 ms. image conv_0 batch_norm_0 running relu forward... running on acl... Thread [1995748352] Forwarding batch_norm_0__, Aborted at 1523442142 (unix time) try "date -d @1523442142" if you are using GNU date PC: @ 0x0 (unknown) SIGSEGV (@0x1) received by PID 32307 (TID 0x76f4b400) from PID 1; stack trace: @ 0x767d7270 (unknown) @ 0x75cf4ece arm_compute::Iterator::Iterator() @ 0x75cf9170 arm_compute::NEActivationLayerKernel::activation<>() @ 0x75a2b350 arm_compute::CPPScheduler::schedule() @ 0x76bb0a5a paddle::ACLOperator::run() @ 0x76bb0c6a paddle::ACLOperator::acl_run() @ 0x76bb1948 paddle::reluActivation::forward() @ 0x76aade38 paddle::Layer::forwardActivation() @ 0x76b21c5e paddle::BatchNormalizationLayer::forward() @ 0x76b4c91e paddle::NeuralNetwork::forward() @ 0x76a17e06 paddle_gradient_machine_forward @ 0x1168c infer() @ 0x11d28 main @ 0x767c88aa libc_start_main Segmentation fault
Only port the ActivationLayer, including ReLU and softmax funciton. First forward was normal, second forward crashed.
Program ouput: