Closed cdasl closed 2 years ago
artyom-beilis
commented
2 years ago Do you have two GPUs?
This is some trivial internal test (not real test - there are better tests under test dir) but it is strange it fails.
What is output of clinfo?
What is output of:
OPENCL_DEBUG_MODE=1 python test.pyAnd
OPENCL_DEBUG_MODE=2 python test.pyAlso what is your operating system?
version of dlprimitives and pytorch_dlprim you can run git rev-parse HEAD to check
cdasl
commented
2 years ago clinfo output is
Number of platforms 1
Platform Name NVIDIA CUDA
Platform Vendor NVIDIA Corporation
Platform Version OpenCL 1.2 CUDA 10.2.95
Platform Profile FULL_PROFILE
Platform Extensions cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64 cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll cl_nv_copy_opts cl_nv_create_buffer cl_khr_int64_base_atomics cl_khr_int64_extended_atomics
Platform Extensions function suffix NV
Platform Name NVIDIA CUDA
Number of devices 2
Device Name Tesla V100S-PCIE-32GB
Device Vendor NVIDIA Corporation
Device Vendor ID 0x10de
Device Version OpenCL 1.2 CUDA
Driver Version 440.33.01
Device OpenCL C Version OpenCL C 1.2
Device Type GPU
Device Topology (NV) PCI-E, 3b:00.0
Device Profile FULL_PROFILE
Device Available Yes
Compiler Available Yes
Linker Available Yes
Max compute units 80
Max clock frequency 1597MHz
Compute Capability (NV) 7.0
Device Partition (core)
Max number of sub-devices 1
Supported partition types None
Max work item dimensions 3
Max work item sizes 1024x1024x64
Max work group size 1024
Preferred work group size multiple 32
Warp size (NV) 32
Preferred / native vector sizes
char 1 / 1
short 1 / 1
int 1 / 1
long 1 / 1
half 0 / 0 (n/a)
float 1 / 1
double 1 / 1 (cl_khr_fp64)
Half-precision Floating-point support (n/a)
Single-precision Floating-point support (core)
Denormals Yes
Infinity and NANs Yes
Round to nearest Yes
Round to zero Yes
Round to infinity Yes
IEEE754-2008 fused multiply-add Yes
Support is emulated in software No
Correctly-rounded divide and sqrt operations Yes
Double-precision Floating-point support (cl_khr_fp64)
Denormals Yes
Infinity and NANs Yes
Round to nearest Yes
Round to zero Yes
Round to infinity Yes
IEEE754-2008 fused multiply-add Yes
Support is emulated in software No
Address bits 64, Little-Endian
Global memory size 34089730048 (31.75GiB)
Error Correction support Yes
Max memory allocation 8522432512 (7.937GiB)
Unified memory for Host and Device No
Integrated memory (NV) No
Minimum alignment for any data type 128 bytes
Alignment of base address 4096 bits (512 bytes)
Global Memory cache type Read/Write
Global Memory cache size 2621440 (2.5MiB)
Global Memory cache line size 128 bytes
Image support Yes
Max number of samplers per kernel 32
Max size for 1D images from buffer 268435456 pixels
Max 1D or 2D image array size 2048 images
Max 2D image size 32768x32768 pixels
Max 3D image size 16384x16384x16384 pixels
Max number of read image args 256
Max number of write image args 32
Local memory type Local
Local memory size 49152 (48KiB)
Registers per block (NV) 65536
Max number of constant args 9
Max constant buffer size 65536 (64KiB)
Max size of kernel argument 4352 (4.25KiB)
Queue properties
Out-of-order execution Yes
Profiling Yes
Prefer user sync for interop No
Profiling timer resolution 1000ns
Execution capabilities
Run OpenCL kernels Yes
Run native kernels No
Kernel execution timeout (NV) No
Concurrent copy and kernel execution (NV) Yes
Number of async copy engines 7
printf() buffer size 1048576 (1024KiB)
Built-in kernels
Device Extensions cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64 cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll cl_nv_copy_opts cl_nv_create_buffer cl_khr_int64_base_atomics cl_khr_int64_extended_atomics
Device Name Tesla V100S-PCIE-32GB
Device Vendor NVIDIA Corporation
Device Vendor ID 0x10de
Device Version OpenCL 1.2 CUDA
Driver Version 440.33.01
Device OpenCL C Version OpenCL C 1.2
Device Type GPU
Device Topology (NV) PCI-E, d8:00.0
Device Profile FULL_PROFILE
Device Available Yes
Compiler Available Yes
Linker Available Yes
Max compute units 80
Max clock frequency 1597MHz
Compute Capability (NV) 7.0
Device Partition (core)
Max number of sub-devices 1
Supported partition types None
Max work item dimensions 3
Max work item sizes 1024x1024x64
Max work group size 1024
Preferred work group size multiple 32
Warp size (NV) 32
Preferred / native vector sizes
char 1 / 1
short 1 / 1
int 1 / 1
long 1 / 1
half 0 / 0 (n/a)
float 1 / 1
double 1 / 1 (cl_khr_fp64)
Half-precision Floating-point support (n/a)
Single-precision Floating-point support (core)
Denormals Yes
Infinity and NANs Yes
Round to nearest Yes
Round to zero Yes
Round to infinity Yes
IEEE754-2008 fused multiply-add Yes
Support is emulated in software No
Correctly-rounded divide and sqrt operations Yes
Double-precision Floating-point support (cl_khr_fp64)
Denormals Yes
Infinity and NANs Yes
Round to nearest Yes
Round to zero Yes
Round to infinity Yes
IEEE754-2008 fused multiply-add Yes
Support is emulated in software No
Address bits 64, Little-Endian
Global memory size 34089730048 (31.75GiB)
Error Correction support Yes
Max memory allocation 8522432512 (7.937GiB)
Unified memory for Host and Device No
Integrated memory (NV) No
Minimum alignment for any data type 128 bytes
Alignment of base address 4096 bits (512 bytes)
Global Memory cache type Read/Write
Global Memory cache size 2621440 (2.5MiB)
Global Memory cache line size 128 bytes
Image support Yes
Max number of samplers per kernel 32
Max size for 1D images from buffer 268435456 pixels
Max 1D or 2D image array size 2048 images
Max 2D image size 32768x32768 pixels
Max 3D image size 16384x16384x16384 pixels
Max number of read image args 256
Max number of write image args 32
Local memory type Local
Local memory size 49152 (48KiB)
Registers per block (NV) 65536
Max number of constant args 9
Max constant buffer size 65536 (64KiB)
Max size of kernel argument 4352 (4.25KiB)
Queue properties
Out-of-order execution Yes
Profiling Yes
Prefer user sync for interop No
Profiling timer resolution 1000ns
Execution capabilities
Run OpenCL kernels Yes
Run native kernels No
Kernel execution timeout (NV) No
Concurrent copy and kernel execution (NV) Yes
Number of async copy engines 7
printf() buffer size 1048576 (1024KiB)
Built-in kernels
Device Extensions cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_fp64 cl_khr_byte_addressable_store cl_khr_icd cl_khr_gl_sharing cl_nv_compiler_options cl_nv_device_attribute_query cl_nv_pragma_unroll cl_nv_copy_opts cl_nv_create_buffer cl_khr_int64_base_atomics cl_khr_int64_extended_atomics
NULL platform behavior
clGetPlatformInfo(NULL, CL_PLATFORM_NAME, ...) No platform
clGetDeviceIDs(NULL, CL_DEVICE_TYPE_ALL, ...) No platform
clCreateContext(NULL, ...) [default] No platform
clCreateContext(NULL, ...) [other] Success [NV]
clCreateContextFromType(NULL, CL_DEVICE_TYPE_DEFAULT) No platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_CPU) No devices found in platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_GPU) No platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_ACCELERATOR) No devices found in platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_CUSTOM) Invalid device type for platform
clCreateContextFromType(NULL, CL_DEVICE_TYPE_ALL) No platformOPENCL_DEBUG_MODE=1 python test.py output is
Accessing device #1:Tesla V100S-PCIE-32GB on NVIDIA CUDA
Exception from at::Tensor& ptdlprim::fill_(at::Tensor&, const c10::Scalar&)
Traceback (most recent call last):
File "test.py", line 16, in <module>
t1=torch.ones((20,10),requires_grad=True,device=dev)
RuntimeError: clSetKernelArgOPENCL_DEBUG_MODE=2 python test.py output is
in: at::Tensor ptdlprim::allocate_empty(c10::IntArrayRef, c10::optional<c10::ScalarType>, c10::optional<c10::Layout>, c10::optional<c10::Device>, c10::optional<bool>, c10::optional<c10::MemoryFormat>)
Accessing device #1:Tesla V100S-PCIE-32GB on NVIDIA CUDA
in: at::Tensor& ptdlprim::fill_(at::Tensor&, const c10::Scalar&)
Traceback (most recent call last):
File "test.py", line 16, in <module>
t1=torch.ones((20,10),requires_grad=True,device=dev)
RuntimeError: clSetKernelArgOS is Ubuntu 18.04.1
pytorch_dlprim: 032a6933fed7e9e0cdcf0dbadf9c7f246804fa3b dlprimitives: 2867462c531d80d6404e616c7290b80c086f4620
artyom-beilis
commented
2 years ago Can you please run these in pytorch_dlprim:
python tests/test_op.py --device=opencl:0
python tests/validate_network.py --model mnist_cnn --device=opencl:0And these in dlprimitives/build directory:
./test_net 0:0 ../tests/test_net.json ../tests/test_weights.json
./test_context 0:0And
./test_context 0:1
./test_net 0:1 ../tests/test_net.json ../tests/test_weights.jsonNote different devices.
cdasl
commented
2 years ago python tests/test_op.py --device=opencl:0
/home/xxx/anaconda3/envs/pytorch_opencl/lib/python3.7/site-packages/torchvision/io/image.py:11: UserWarning: Failed to load image Python extension: libc10_cuda.so: cannot open shared object file: No such file or directory
warn(f"Failed to load image Python extension: {e}")
Mean 1d
Accessing device #0:Tesla V100S-PCIE-32GB on NVIDIA CUDA
Traceback (most recent call last):
File "tests/test_op.py", line 165, in <module>
test_all(r.device)
File "tests/test_op.py", line 93, in test_all
test_fwd_bwd([([2,3,4],-1)],lambda x:torch.mean(x,dim=0,keepdim=True),device)
File "tests/test_op.py", line 49, in test_fwd_bwd
x_dev = x_cpu.to(device)
RuntimeError: clEnqueueWriteBufferpython tests/validate_network.py --model mnist_cnn --device=opencl:0
/home/xxx/anaconda3/envs/pytorch_opencl/lib/python3.7/site-packages/torchvision/io/image.py:11: UserWarning: Failed to load image Python extension: libc10_cuda.so: cannot open shared object file: No such file or directory
warn(f"Failed to load image Python extension: {e}")
Testing mnist_cnn
Accessing device #0:Tesla V100S-PCIE-32GB on NVIDIA CUDA
Traceback (most recent call last):
File "tests/validate_network.py", line 272, in <module>
main(r)
File "tests/validate_network.py", line 220, in main
train_on_images(m,batch,args.device,args.eval,iter_size = args.iter_size,opt_steps = args.opt,fwd=args.fwd)
File "tests/validate_network.py", line 95, in train_on_images
data_dev = data.to(device)
RuntimeError: clEnqueueWriteBuffer./test_net 0:0 ../tests/test_net.json ../tests/test_weights.json
Testing for Tesla V100S-PCIE-32GB on NVIDIA CUDA
Checking Diffs
Testing bn2a.2
Testing bn2a.3
Testing cnv1.0
Testing cnv1.1
Testing cnv2.0
Testing cnv2.1
Testing cnv2a.0
Testing cnv2b.0
Testing cnv2b.1
Testing fc.0
Testing fc.1
Checking Param Updates
Testing bn2a.0
Testing bn2a.1
Testing bn2a.2
Testing bn2a.3
Testing cnv1.0
Testing cnv1.1
Testing cnv2.0
Testing cnv2.1
Testing cnv2a.0
Testing cnv2b.0
Testing cnv2b.1
Testing fc.0
Testing fc.1./test_context 0:0
Basic context
Tesla V100S-PCIE-32GB on NVIDIA CUDA
Ok./test_context 0:1
Basic context
Tesla V100S-PCIE-32GB on NVIDIA CUDA
Ok./test_net 0:1 ../tests/test_net.json ../tests/test_weights.json
Testing for Tesla V100S-PCIE-32GB on NVIDIA CUDA
Checking Diffs
Testing bn2a.2
Testing bn2a.3
Testing cnv1.0
Testing cnv1.1
Testing cnv2.0
Testing cnv2.1
Testing cnv2a.0
Testing cnv2b.0
Testing cnv2b.1
Testing fc.0
Testing fc.1
Checking Param Updates
Testing bn2a.0
Testing bn2a.1
Testing bn2a.2
Testing bn2a.3
Testing cnv1.0
Testing cnv1.1
Testing cnv2.0
Testing cnv2.1
Testing cnv2a.0
Testing cnv2b.0
Testing cnv2b.1
Testing fc.0
Testing fc.1I find the problem.
Just set USE_CL_HPP=OFF and all can run normally.
Thanks for your project, it is really awesome!!
artyom-beilis
commented
2 years ago I find the problem. Just set
USE_CL_HPP=OFFand all can run normally. Thanks for your project, it is really awesome!!
Thanks, interesting I'll check. I expected that it should work with both CL/cl.hpp and CL/cl2.hpp I'll check it.
After installing pytorch_dlprim, when
python test.py, get error.