Issue summary

I had succeeded to training bvlc-alexnet and bvlc-googlenet models in single MI25 GPU. When I changed the number of training GPU from 1 to all, caffe show the below message.. CPU memory:256GB swap:16GB db:imagenet lmdb batchsize:64 bvlc_alexnet:

I0719 10:51:50.941951 2540 solver.cpp:279] Solving AlexNet I0719 10:51:50.941956 2540 solver.cpp:280] Learning Rate Policy: step I0719 10:51:50.955250 2540 solver.cpp:337] Iteration 0, Testing net (#0) I0719 10:54:02.507711 2540 solver.cpp:404] Test net output #0: accuracy = 0.00109375 I0719 10:54:02.508229 2540 solver.cpp:404] Test net output #1: loss = 6.91062 (* 1 = 6.91062 loss) Memory access fault by GPU node-2 on address 0x422ea6b000. Reason: Page not present or supervisor privilege. *** Aborted at 1500432842 (unix time) try "date -d @1500432842" if you are using GNU date *** PC: @ 0x7f64489dc428 gsignal *** SIGABRT (@0x9ec) received by PID 2540 (TID 0x7f642c526700) from PID 2540; stack trace: *** @ 0x7f644ddd0390 (unknown) @ 0x7f64489dc428 gsignal @ 0x7f64489de02a abort @ 0x7f644d9401c9 (unknown) @ 0x7f644d9464e5 (unknown) @ 0x7f644d91e9d7 (unknown) @ 0x7f644ddc66ba start_thread @ 0x7f6448aae3dd clone @ 0x0 (unknown)

db:imagenet lmdb batchsize:32 bvlc_googlenet:

I0719 00:12:28.380522 7405 solver.cpp:279] Solving GoogleNet I0719 00:12:28.380544 7405 solver.cpp:280] Learning Rate Policy: step Memory access fault by GPU node-2 on address 0x42309ba000. Reason: Page not present or supervisor privilege. *** Aborted at 1500394348 (unix time) try "date -d @1500394348" if you are using GNU date *** PC: @ 0x7f4078d7a428 gsignal *** SIGABRT (@0x1ced) received by PID 7405 (TID 0x7f405c8c4700) from PID 7405; stack trace: *** @ 0x7f407e16e390 (unknown) @ 0x7f4078d7a428 gsignal @ 0x7f4078d7c02a abort @ 0x7f407dcde1c9 (unknown) @ 0x7f407dce44e5 (unknown) @ 0x7f407dcbc9d7 (unknown) @ 0x7f407e1646ba start_thread @ 0x7f4078e4c3dd clone @ 0x0 (unknown)

Steps to reproduce

Using the latest ROCm from debian packages.

My caffe configuration:

USE_CUDNN := 0 USE_MIOPEN := 1 USE_LMDB := 1 BLAS := open BLAS_INCLUDE := /opt/openBlas/include BLAS_LIB := /opt/openBlas/lib

Your system configuration

Operating system: Ubuntu 16.04.2 LTS with 4.9.0-kfd-compute-rocm-rel-1.6-77 Compiler: GCC v5.4.0, HCC clang 5.0 CUDA version (if applicable): not applicable CUDNN version (if applicable): not applicable BLAS: OpenBlas Python or MATLAB version (for pycaffe and matcaffe respectively): not applicable

Issue summary

I have a problem with lenet.prototxt. ./caffe time -model '../../examples/mnist/lenet.prototxt' - OK ./caffe time -model '../../examples/mnist/lenet.prototxt' -gpu 0 - fails with this message:

Network initialization done.
I0714 10:10:04.648849 13230 caffe.cpp:355] Performing Forward
terminate called after throwing an instance of 'char const*'
*** Aborted at 1500016205 (unix time) try "date -d @1500016205" if you are using GNU date ***
PC: @     0x7f76dc314428 gsignal
*** SIGABRT (@0x3e8000033ae) received by PID 13230 (TID 0x7f76e4675b00) from PID 13230; stack trace: ***
    @     0x7f76e3b0a390 (unknown)
    @     0x7f76dc314428 gsignal
    @     0x7f76dc31602a abort
    @     0x7f76e295970d __gnu_cxx::__verbose_terminate_handler()
    @     0x7f76e294f4b6 __cxxabiv1::__terminate()
    @     0x7f76e294f501 std::terminate()
    @     0x7f76e294f609 __cxa_throw
    @     0x7f76e294dddc hipblasSgemm
    @           0x6b9449 caffe::caffe_gpu_gemm<>()
    @           0x654ec4 caffe::InnerProductLayer<>::Forward_gpu()
    @           0x430b2f caffe::Layer<>::Forward()
    @           0x6ff4d7 caffe::Net<>::ForwardFromTo()
    @           0x6ff3ef caffe::Net<>::Forward()
    @           0x42fa5e time()
    @           0x431412 main
    @     0x7f76dc2ff830 __libc_start_main
    @           0x850cf9 _start
    @                0x0 (unknown)

Steps to reproduce

Freshly compiled hipCaffe with default Makefile.config. Compilation flags = -O3 I am using the latest ROCm from debian packages. Running the MNIST lenet example works fine only with CPU. test_all.testbin show problems with other networks with GPU.

Your system configuration

Operating system: Ubuntu 16.04.2 LTS with 4.9.0-kfd-compute-rocm-rel-1.6-77 Compiler: GCC v5.4.0, HCC clang 5.0 CUDA version (if applicable): not applicable CUDNN version (if applicable): not applicable BLAS: ATLAS Python or MATLAB version (for pycaffe and matcaffe respectively): not applicable

Hardware: Radeon Pro WX 7100

Issue summary

A SIGABRT is received when running the unit tests or the examples on a Vega RX with Ubuntu.

Steps to reproduce

I followed all the steps on https://github.com/ROCmSoftwarePlatform/hipCaffe/blob/hip/README.ROCm.md and I got:

$ ./build/test/test_all.testbin --gtest_filter='DataLayerTest/2.*'
Cuda number of devices: 1
Current device id: 0
Current device name: Vega [Radeon RX Vega]
Note: Google Test filter = DataLayerTest/2.*
[==========] Running 12 tests from 1 test case.
[----------] Global test environment set-up.
[----------] 12 tests from DataLayerTest/2, where TypeParam = caffe::GPUDevice<float>
[ RUN      ] DataLayerTest/2.TestReadLevelDB
Memory access fault by GPU node-1 on address 0x28000. Reason: Page not present or supervisor privilege.
*** Aborted at 1507162515 (unix time) try "date -d @1507162515" if you are using GNU date ***
PC: @     0x7f80572b077f gsignal
*** SIGABRT (@0x3e800001f89) received by PID 8073 (TID 0x7f803a184700) from PID 8073; stack trace: ***
    @     0x7f805c6b1670 (unknown)
    @     0x7f80572b077f gsignal
    @     0x7f80572b237a abort
    @     0x7f805c220559 (unknown)
    @     0x7f805c226885 (unknown)
    @     0x7f805c1fe9d7 (unknown)
    @     0x7f805c6a76da start_thread
    @     0x7f8057383d7f clone
    @                0x0 (unknown)
Aborted (core dumped)

Stack with gdb:

#0  0x00007ffff1c8f77f in __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:58
#1  0x00007ffff1c9137a in __GI_abort () at abort.c:89
#2  0x00007ffff6bff559 in  () at /opt/rocm/hsa/lib/libhsa-runtime64.so.1
#3  0x00007ffff6c05885 in  () at /opt/rocm/hsa/lib/libhsa-runtime64.so.1
#4  0x00007ffff6bdd9d7 in  () at /opt/rocm/hsa/lib/libhsa-runtime64.so.1
#5  0x00007ffff70866da in start_thread (arg=0x7fffd4b63700) at pthread_create.c:456
#6  0x00007ffff1d62d7f in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:105

I have the same stack with ./examples/mnist/train_lenet.sh

Your system configuration

Operating system: Ubuntu 17.04 x86-64 Compiler: g++ 6.3.0 BLAS: libblas 3.7.0-1, hcblas 0.1.1, hipblas 0.4.4.6 Python or MATLAB version (for pycaffe and matcaffe respectively): Hardware (should be supported by Rocm):

• Vega RX 56
• Intel(R) Core(TM) i7-4770K CPU @ 3.50GHz

Hello! The reason I am opening issue here is because I didn't find better place for discussion of caffe running on AMD gpus. If there is better place for discussion, please say so.

So I was able to install ROCm 1.5 on my Ubuntu 16.04 running on i7 6700 and Radeon RX 480 8GB.

I did manage to build hipCaffe and run mnist example using AMD GPU. I also tried same sample on cpu (using multithreaded OpenBLAS) and the result were not that impressive.

I am testing using command time ./examples/mnist/train_lenet.sh When using CPU only (8 threads), the result is:

real	7m38.942s
user	22m0.692s
sys	34m6.000s

When using GPU (rx 480 + ROCm 1.5):

real	5m46.945s
user	5m27.120s
sys	0m7.256s

The speed is better also CPU is more free. Compared to out "gpu grid" server when using 1x nvidia Titan X:

real	2m0.855s
user	1m34.332s
sys	0m43.948s

I do understand that Titan X is much faster than RX 480 and nvidia has put MUCH more resources in to deep learning optimisation. Are those results expected, or should I get better? Anyone else tried hipCaffe on any AMD GPU?

Your system configuration

Operating system: Ubuntu 16.04 + ROCm 1.5 (kernel 4.9-kfd) + RX 480 8GB BLAS: OpenBLAS (and amd hip BLAS variant) Python or MATLAB version (for pycaffe and matcaffe respectively): no python/matlab

Issue summary

Initially reported in #19 that I am getting issues with test failures as well as core dumps, but just reporting on core dumps here for now.

In short, NetTest/0.TestReshape, as well as my attempts at running the MNIST and CaffeNet all end with a core dump. Data for CIFAR-10 has an integrity problem...

(I've been out of the game long enough that I'm not sure how to get the stack trace with gdb... I'm happy to look into this further.)

Steps to reproduce

If you are having difficulty building Caffe or training a model, please ask the caffe-users mailing list. If you are reporting a build error that seems to be due to a bug in Caffe, please attach your build configuration (either Makefile.config or CMakeCache.txt) and the output of the make (or cmake) command.

Your system configuration

Operating System: Ubuntu 16.04.3 LTS, Linux kernel 4.13.0, ROCm & DKMS from the ROCm PPA. GPU: AMD RX 580, drivers from rocm PPA CPU: Threadripper 1900X on X399 chipset Compiler: (I think this is the one you want?) hcc version=1.2.18063-7e18c64-ac8732c-710f135, workweek (YYWWD) = 18063 BLAS: ROCBLAS (I assume - that's the default in Makefile.config) Python or MATLAB version (for pycaffe and matcaffe respectively): standard Ubuntu Python 2.7, no matlab

Issue summary

I am not sure if this a problem. I training ImageNet on VGG-16 network with multiple GPUs (MI25 x 4). After interrupted current training job, and re-run this training again, caffe shows below error message. But if I change the training command to use any single GPU in my system, no error found.
The workaround env "export HCC_UNPINNED_COPY_MODE=2" has been applied before training.

Error Message: *** Aborted at 1509489472 (unix time) try "date -d @1509489472" if you are using GNU date *** PC: @ 0x7f36ae941faa (unknown) *** SIGSEGV (@0x15b475060) received by PID 8045 (TID 0x7ef26b7fe700) from PID 1531400288; stack trace: *** @ 0x7f36b4a1c390 (unknown) @ 0x7f36ae941faa (unknown) @ 0x7f36b34b1315 Mrg32k3aCreateStream() @ 0x7f36b34b1235 hcrngMrg32k3aCreateOverStreams() @ 0x7f36b352943f hiprngSetPseudoRandomGeneratorSeed @ 0x522ffc caffe::Caffe::SetDevice() @ 0x722dca caffe::P2PSync<>::InternalThreadEntry() @ 0x7f36b0eb15d5 (unknown) @ 0x7f36b4a126ba start_thread @ 0x7f36ae8fb3dd clone @ 0x0 (unknown) Segmentation fault (core dumped)

Steps to reproduce

1. Run the below command to train my network amax@p47:/data/ImageNet/imagenet$ /home/amax/hipCaffe/build/tools/caffe train -solver=/data/ImageNet/models/vgg_net/vgg_solver.prototxt -gpu 0,1,2,3
2. Use Ctrl-C to interrupt this job
3. Re-Run the command in the step 1. It will show the error message above.
4. If I change the '-gpu' to any single gpu, the program will run but only one GPU.
5. If I change the '-gpu' to any number of gpu, the program still show above error message. But during my testing, there still 1/10 chance to run without error message.
6. This error message also happened when I run the train job when the system just boot up.
7. After waiting several minutes, some times the job could be run again but some times not.

I am not sure the root cause, from the error message it seems the device resource has not be released after the previous job terminated.

Your system configuration

Operating system: Ubuntu 16.04.3 Compiler: gcc version 5.4.0 20160609 (Ubuntu 5.4.0-6ubuntu1~16.04.5) CUDA version (if applicable): CUDNN version (if applicable): BLAS: USE_ROCBLAS := 1 Python or MATLAB version (for pycaffe and matcaffe respectively): 2.7.12 Other: miopen-hip 1.1.4 miopengemm 1.1.5 rocm-libs 1.6.180 Server: Inventec P47 GPU: AMD MI25 x4 CPU: AMD EPYC 7601 x2 Memory: 512GB

• Fix one memory leak

• Reduce HIP stream numbers by using default HIP stream

• Reduce MIOpen handle numbers by one handle for fwd / bwd weight / bwd data

• Polish debug message output

• Add debug message for CuDNNConvolutionLayer count

• Remove unused CUDNN integration logic

I followed the instruction but I'm getting this error when building hipCaffe.

.build_release/src/caffe/proto/caffe.pb.h:12:2: error: This file was generated by a newer version of protoc which is #error This file was generated by a newer version of protoc which is ^

Clean set up of Ubuntu 16.04.3 and ROCm-1.6-180. Any advice?

Issue summary

test_all.testbin fais at the following test. All preceding tests are OK:

[----------] 2 tests from InternalThreadTest
[ RUN      ] InternalThreadTest.TestStartAndExit
[       OK ] InternalThreadTest.TestStartAndExit (14 ms)
[ RUN      ] InternalThreadTest.TestRandomSeed
*** Aborted at 1499717161 (unix time) try "date -d @1499717161" if you are using GNU date ***
PC: @     0x7f4a6bb45116 (unknown)
*** SIGSEGV (@0x0) received by PID 7887 (TID 0x7f4a719e2b80) from PID 0; stack trace: ***
    @     0x7f4a70ea0390 (unknown)
    @     0x7f4a6bb45116 (unknown)
    @     0x7f4a4f35d75d HSACopy::syncCopyExt()
    @     0x7f4a4f35c8bc Kalmar::HSAQueue::copy_ext()
    @     0x7f4a715daa5b ihipStream_t::locked_copySync()
    @     0x7f4a716020bf hipMemcpy
    @     0x7f4a6f9da4f8 hiprngSetPseudoRandomGeneratorSeed
    @           0xc41b60 caffe::Caffe::set_random_seed()
    @           0x48827b caffe::InternalThreadTest_TestRandomSeed_Test::TestBody()
    @           0xb47884 testing::internal::HandleExceptionsInMethodIfSupported<>()
    @           0xb47745 testing::Test::Run()
    @           0xb488f0 testing::TestInfo::Run()
    @           0xb49137 testing::TestCase::Run()
    @           0xb4f517 testing::internal::UnitTestImpl::RunAllTests()
    @           0xb4ef64 testing::internal::HandleExceptionsInMethodIfSupported<>()
    @           0xb4ef19 testing::UnitTest::Run()
    @           0xf83cda main
    @     0x7f4a6ba17830 __libc_start_main
    @           0xf76c19 _start
    @                0x0 (unknown)

Steps to reproduce

Freshly compiled hipCaffe with default Makefile.config. Compilation flags = -O3 I am using the latest ROCm from debian packages. Running the MNIST lenet example works fine (with or without MIOpen).

Your system configuration

Operating system: Linux Mint 18 (Debian derivative) Compiler: GCC v5.4.0, HCC clang 5.0 CUDA version (if applicable): not applicable CUDNN version (if applicable): not applicable BLAS: ATLAS Python or MATLAB version (for pycaffe and matcaffe respectively): not applicable

This fix enables caffe to use MIOpen for MAX_POOLING. This was disabled for some reason which cuDNN does not / did not support. Looking at NVIDIA's github itself makes me believe that they have fixed this limitation.

Highly recommend to run training with this fix to ensure this does not manifest in accuracy issues. I am suspecting all this while we were not using MIOpen for pooling in our training runs.

cc\ @ashishfarmer This might also be the reason for the discrepancy you saw between ocl and hip caffe for pooling.

If I follow the ROCm readme to the letter, I end up successfully building everything, but tests don't run:

$ ./build/test/test_all.testbin
### HCC RUNTIME ERROR: Fail to find compatible kernel at file:/home/jenkins/jenkins-root/workspace/compute-rocm-rel-1.5/external/hcc-tot/lib/mcwamp.cpp line:344

This is aside from my other problem, which is that installing ROCm completely destabilised my Ubuntu install, and on a fresh reinstall of 16.04 it stops any GUI from appearing, so I'm working from TTY1.. When will ROCm work with a real kernel?

@xmaxk

We are going to enforce two factor authentication in (https://github.com/ROCmSoftwarePlatform/) organization on 29th April, 2022 . Since we identified you as outside collaborator for ROCmSoftwarePlatform organization, you need to enable two factor authentication in your github account else you shall be removed from the organization after the enforcement. Please skip if already done.

To set up two factor authentication, please go through the steps in below link:

https://docs.github.com/en/free-pro-team@latest/github/authenticating-to-github/configuring-two-factor-authentication

Please email "[email protected]" for queries

Why this "hipCaffe" is not a simple fork from bvlc caffe repo? I know it's a stupid question, but, maybe it will be easier to update the "hip" with the other caffe version (like bvlc or opencl version).

Issue summary

I'm reposting this after closing my original issue because I am now quite confident the install was completely canonical.

Running ./build/test/test_all.testbin ultimately drops core. It drops core same place with either RX470 or RX Vega 64.

A few tests fail but at a certain point it always drops core.

Steps to reproduce

• Clean install Ubuntu 18.04.1 LTS Server.
• Use Ubuntu's stock kernels provided by only the apt-get update / upgrade mechanism (ie the kernel in use is that provided by Ubuntu after apt-get dist-upgrade: I have not installed an upstream kernel.
• Install ROCm 2.0 & hipCaffe per the hipCaffe instructions.
• Run several of the examples without error.
• Run ./build/test/test_all.testbin... drops core.

Problem:

...
[ RUN      ] NetTest/2.TestForcePropagateDown
[       OK ] NetTest/2.TestForcePropagateDown (2 ms)
[ RUN      ] NetTest/2.TestAllInOneNetTrain
[       OK ] NetTest/2.TestAllInOneNetTrain (3 ms)
[ RUN      ] NetTest/2.TestAllInOneNetVal
[       OK ] NetTest/2.TestAllInOneNetVal (4 ms)
[ RUN      ] NetTest/2.TestAllInOneNetDeploy
[       OK ] NetTest/2.TestAllInOneNetDeploy (1 ms)
[----------] 26 tests from NetTest/2 (772 ms total)

[----------] 26 tests from NetTest/3, where TypeParam = caffe::GPUDevice<double>
[ RUN      ] NetTest/3.TestHasBlob
[       OK ] NetTest/3.TestHasBlob (4 ms)
[ RUN      ] NetTest/3.TestGetBlob
[       OK ] NetTest/3.TestGetBlob (4 ms)
[ RUN      ] NetTest/3.TestHasLayer
[       OK ] NetTest/3.TestHasLayer (4 ms)
[ RUN      ] NetTest/3.TestGetLayerByName
[       OK ] NetTest/3.TestGetLayerByName (4 ms)
[ RUN      ] NetTest/3.TestBottomNeedBackward
[       OK ] NetTest/3.TestBottomNeedBackward (4 ms)
[ RUN      ] NetTest/3.TestBottomNeedBackwardForce
[       OK ] NetTest/3.TestBottomNeedBackwardForce (4 ms)
[ RUN      ] NetTest/3.TestBottomNeedBackwardEuclideanForce
[       OK ] NetTest/3.TestBottomNeedBackwardEuclideanForce (1 ms)
[ RUN      ] NetTest/3.TestBottomNeedBackwardTricky
[       OK ] NetTest/3.TestBottomNeedBackwardTricky (5 ms)
[ RUN      ] NetTest/3.TestLossWeight
[       OK ] NetTest/3.TestLossWeight (21 ms)
[ RUN      ] NetTest/3.TestLossWeightMidNet
[       OK ] NetTest/3.TestLossWeightMidNet (16 ms)
[ RUN      ] NetTest/3.TestComboLossWeight
[       OK ] NetTest/3.TestComboLossWeight (18 ms)
[ RUN      ] NetTest/3.TestBackwardWithAccuracyLayer
MIOpen Error: /home/dlowell/MIOpenPrivate/src/ocl/softmaxocl.cpp:59: Only alpha=1 and beta=0 is supported
F0116 04:57:34.752313 24321 cudnn_softmax_layer_hip.cpp:27] Check failed: status == miopenStatusSuccess (7 vs. 0)  miopenStatusUnknownError
*** Check failure stack trace: ***
    @     0x7f2ab9f720cd  google::LogMessage::Fail()
    @     0x7f2ab9f73f33  google::LogMessage::SendToLog()
    @     0x7f2ab9f71c28  google::LogMessage::Flush()
    @     0x7f2ab9f74999  google::LogMessageFatal::~LogMessageFatal()
    @          0x15364ce  caffe::CuDNNSoftmaxLayer<>::Forward_gpu()
    @           0x4cb540  caffe::Layer<>::Forward()
    @          0x1ea2073  caffe::SoftmaxWithLossLayer<>::Forward_gpu()
    @           0x4cb540  caffe::Layer<>::Forward()
    @          0x1b459d7  caffe::Net<>::ForwardFromTo()
    @          0x1b458f0  caffe::Net<>::Forward()
    @           0x967796  caffe::NetTest_TestBackwardWithAccuracyLayer_Test<>::TestBody()
    @          0x108be34  testing::internal::HandleExceptionsInMethodIfSupported<>()
    @          0x108bcf6  testing::Test::Run()
    @          0x108ceb1  testing::TestInfo::Run()
    @          0x108d5c7  testing::TestCase::Run()
    @          0x1093967  testing::internal::UnitTestImpl::RunAllTests()
    @          0x10933a4  testing::internal::HandleExceptionsInMethodIfSupported<>()
    @          0x1093359  testing::UnitTest::Run()
    @          0x201545a  main
    @     0x7f2ab4b00b97  __libc_start_main
    @          0x20148fa  _start
Aborted (core dumped)

Your system configuration

I can provide what ever info you need, just tell me what you want.

Operating system: Ubuntu 18.04.1 LTS Compiler: gcc (Ubuntu 7.3.0-27ubuntu1~18.04) 7.3.0 CUDA version (if applicable): N/A CUDNN version (if applicable): N/A BLAS: rocblas 2.0.0.0 Python or MATLAB version (for pycaffe and matcaffe respectively): Python 2.7.15rc1

Hardware: RX Vega 64, or RX 470 4GB. Ryzen 5 2600X 16 GB RAM X470 mobo SR-IOV is turned off IOMMU enabled or disabled - same result.

Please use the caffe-users list for usage, installation, or modeling questions, or other requests for help. Do not post such requests to Issues. Doing so interferes with the development of Caffe.

Please read the guidelines for contributing before submitting this issue.

Issue summary

Compilation error with latest and greatest checkout

Steps to reproduce

Usual process of CMake followed by make If you are having difficulty building Caffe or training a model, please ask the caffe-users mailing list. If you are reporting a build error that seems to be due to a bug in Caffe, please attach your build configuration (either Makefile.config or CMakeCache.txt) and the output of the make (or cmake) command.

Your system configuration

Operating system: ubuntu 18.04 Compiler: hcc (latest version - CLang 8.0) CUDA version (if applicable): CUDNN version (if applicable): BLAS: Python or MATLAB version (for pycaffe and matcaffe respectively):

[ 2%] Building CXX object src/caffe/CMakeFiles/caffe.dir/layers/absval_layer_hip.cpp.o /home/naths/srcs_new/hipCaffe/src/caffe/layers/absval_layer_hip.cpp:13:3: error: use of undeclared identifier 'caffe_gpu_abs' caffe_gpu_abs(count, bottom[0]->gpu_data(), top_data); ^ /home/naths/srcs_new/hipCaffe/src/caffe/layers/absval_layer_hip.cpp:29:1: note: in instantiation of member function 'caffe::AbsValLayer::Forward_gpu' requested here INSTANTIATE_LAYER_GPU_FUNCS(AbsValLayer); ^ /home/naths/srcs_new/hipCaffe/include/caffe/common.hpp:85:3: note: expanded from macro 'INSTANTIATE_LAYER_GPU_FUNCS' INSTANTIATE_LAYER_GPU_FORWARD(classname);
^ /home/naths/srcs_new/hipCaffe/include/caffe/common.hpp:67:35: note: expanded from macro 'INSTANTIATE_LAYER_GPU_FORWARD' template void classname::Forward_gpu(
^ /home/naths/srcs_new/hipCaffe/src/caffe/layers/absval_layer_hip.cpp:13:3: error: use of undeclared identifier 'caffe_gpu_abs' caffe_gpu_abs(count, bottom[0]->gpu_data(), top_data); ^ /home/naths/srcs_new/hipCaffe/src/caffe/layers/absval_layer_hip.cpp:29:1: note: in instantiation of member function 'caffe::AbsValLayer::Forward_gpu' requested here INSTANTIATE_LAYER_GPU_FUNCS(AbsValLayer); ^ /home/naths/srcs_new/hipCaffe/include/caffe/common.hpp:85:3: note: expanded from macro 'INSTANTIATE_LAYER_GPU_FUNCS' INSTANTIATE_LAYER_GPU_FORWARD(classname);
^ /home/naths/srcs_new/hipCaffe/include/caffe/common.hpp:70:36: note: expanded from macro 'INSTANTIATE_LAYER_GPU_FORWARD' template void classname::Forward_gpu(
^ /home/naths/srcs_new/hipCaffe/src/caffe/layers/absval_layer_hip.cpp:24:5: error: use of undeclared identifier 'caffe_gpu_sign' caffe_gpu_sign(count, bottom_data, bottom_diff); ^ /home/naths/srcs_new/hipCaffe/src/caffe/layers/absval_layer_hip.cpp:29:1: note: in instantiation of member function 'caffe::AbsValLayer::Backward_gpu' requested here INSTANTIATE_LAYER_GPU_FUNCS(AbsValLayer); ^ /home/naths/srcs_new/hipCaffe/include/caffe/common.hpp:86:3: note: expanded from macro 'INSTANTIATE_LAYER_GPU_FUNCS' INSTANTIATE_LAYER_GPU_BACKWARD(classname) ^ /home/naths/srcs_new/hipCaffe/include/caffe/common.hpp:75:35: note: expanded from macro 'INSTANTIATE_LAYER_GPU_BACKWARD' template void classname::Backward_gpu(
^ /home/naths/srcs_new/hipCaffe/src/caffe/layers/absval_layer_hip.cpp:25:5: error: use of undeclared identifier 'caffe_gpu_mul' caffe_gpu_mul(count, bottom_diff, top_diff, bottom_diff); ^ /home/naths/srcs_new/hipCaffe/src/caffe/layers/absval_layer_hip.cpp:24:5: error: use of undeclared identifier 'caffe_gpu_sign' caffe_gpu_sign(count, bottom_data, bottom_diff); ^ /home/naths/srcs_new/hipCaffe/src/caffe/layers/absval_layer_hip.cpp:29:1: note: in instantiation of member function 'caffe::AbsValLayer::Backward_gpu' requested here INSTANTIATE_LAYER_GPU_FUNCS(AbsValLayer); ^ /home/naths/srcs_new/hipCaffe/include/caffe/common.hpp:86:3: note: expanded from macro 'INSTANTIATE_LAYER_GPU_FUNCS' INSTANTIATE_LAYER_GPU_BACKWARD(classname) ^ /home/naths/srcs_new/hipCaffe/include/caffe/common.hpp:79:36: note: expanded from macro 'INSTANTIATE_LAYER_GPU_BACKWARD' template void classname::Backward_gpu(
^ /home/naths/srcs_new/hipCaffe/src/caffe/layers/absval_layer_hip.cpp:25:5: error: use of undeclared identifier 'caffe_gpu_mul' caffe_gpu_mul(count, bottom_diff, top_diff, bottom_diff); ^ 6 errors generated. src/caffe/CMakeFiles/caffe.dir/build.make:254: recipe for target 'src/caffe/CMakeFiles/caffe.dir/layers/absval_layer_hip.cpp.o' failed make[2]: *** [src/caffe/CMakeFiles/caffe.dir/layers/absval_layer_hip.cpp.o] Error 1 CMakeFiles/Makefile2:235: recipe for target 'src/caffe/CMakeFiles/caffe.dir/all' failed make[1]: *** [src/caffe/CMakeFiles/caffe.dir/all] Error 2 Makefile:129: recipe for target 'all' failed make: *** [all] Error 2

Issue summary

Two straightforward questions:

1. to build python caffe interface do I run "make", or "make py" or "make python"?

2. Will this code work on an RX470, or with an RX Vega 64? I'd rather experiment using the RX 470 if I can get away with it.

Thanks!

Please use the caffe-users list for usage, installation, or modeling questions, or other requests for help. Do not post such requests to Issues. Doing so interferes with the development of Caffe.

Please read the guidelines for contributing before submitting this issue.

Issue summary

Fresh install of Ubuntu 16.x Desktop, ROCM and started to build hipCafee however ran into a error when running ./build/test/test_all.testbin

Stack Trace:

[ RUN      ] SoftmaxWithLossLayerTest/2.TestGradientUnnormalized
[       OK ] SoftmaxWithLossLayerTest/2.TestGradientUnnormalized (89 ms)
[----------] 4 tests from SoftmaxWithLossLayerTest/2 (2434 ms total)

[----------] 4 tests from SoftmaxWithLossLayerTest/3, where TypeParam = caffe::GPUDevice<double>
[ RUN      ] SoftmaxWithLossLayerTest/3.TestGradient
MIOpen Error: /home/dlowell/MIOpenPrivate/src/ocl/softmaxocl.cpp:59: Only alpha=1 and beta=0 is supported
F1227 11:47:35.865128 23305 cudnn_softmax_layer_hip.cpp:27] Check failed: status == miopenStatusSuccess (7 vs. 0)  miopenStatusUnknownError
*** Check failure stack trace: ***
    @     0x7f6f6c7d15cd  google::LogMessage::Fail()
    @     0x7f6f6c7d3433  google::LogMessage::SendToLog()
    @     0x7f6f6c7d115b  google::LogMessage::Flush()
    @     0x7f6f6c7d3e1e  google::LogMessageFatal::~LogMessageFatal()
    @          0x14a1f2e  caffe::CuDNNSoftmaxLayer<>::Forward_gpu()
    @           0x4ed860  caffe::Layer<>::Forward()
    @          0x1d9a153  caffe::SoftmaxWithLossLayer<>::Forward_gpu()
    @           0x4ed860  caffe::Layer<>::Forward()
    @           0x51c010  caffe::GradientChecker<>::CheckGradientSingle()
    @           0x51b76d  caffe::GradientChecker<>::CheckGradientExhaustive()
    @           0xa21807  caffe::SoftmaxWithLossLayerTest_TestGradient_Test<>::TestBody()
    @          0x1033034  testing::internal::HandleExceptionsInMethodIfSupported<>()
    @          0x1032ef6  testing::Test::Run()
    @          0x1034051  testing::TestInfo::Run()
    @          0x10348b7  testing::TestCase::Run()
    @          0x103ada7  testing::internal::UnitTestImpl::RunAllTests()
    @          0x103a7e4  testing::internal::HandleExceptionsInMethodIfSupported<>()
    @          0x103a799  testing::UnitTest::Run()
    @          0x1ef98fa  main
    @     0x7f6f6783f830  __libc_start_main
    @          0x1ef8d99  _start
    @              (nil)  (unknown)
Aborted (core dumped)

Scrolling up to see if any errors I did find this , I believe this was on make test.

Expected: data[i]
Which is: 1.6294847
src/caffe/test/test_inner_product_layer.cpp:384: Failure
Value of: data_t[i]
  Actual: 2.9882355
Expected: data[i]
Which is: 2.474798
src/caffe/test/test_inner_product_layer.cpp:384: Failure
Value of: data_t[i]
  Actual: 2.1015618
Expected: data[i]
Which is: 2.0466099

Thoughts, suggestions I could try..

Your system configuration

Operating system: Ubuntu 16 Desktop Compiler: gcc version 5.4.0 20160609 (Ubuntu 5.4.0-6ubuntu1~16.04.11) CUDA version (if applicable): NA CUDNN version (if applicable): BLAS: Python or MATLAB version (for pycaffe and matcaffe respectively):