AStupidBear
commented
7 years ago I have checked my GPU and there's no problem.
Closed AStupidBear closed 7 years ago
AStupidBear
commented
7 years ago I have checked my GPU and there's no problem.
denizyuret
commented
7 years ago Knet 0.8.3 (which works with AutoGrad 0.0.7) was just released. Can you try with these? (If you haven't switched to Knet master, Pkg.update() and Pkg.build("Knet") should be sufficient to upgrade. If you have, you may need to run Pkg.free("AutoGrad") and Pkg.free("Knet") first).
On Sun, May 21, 2017 at 10:00 AM Yao Lu notifications@github.com wrote:
I have checked my GPU and there's no problem.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/denizyuret/Knet.jl/issues/125#issuecomment-302919227, or mute the thread https://github.com/notifications/unsubscribe-auth/ABvNpiW2GDUKZcmh0ZklqsMsqAX5M1Imks5r7-DzgaJpZM4NheG9 .
AStupidBear
commented
7 years ago Pkg.test("Knet") still failed.
AStupidBear
commented
7 years ago That's strange. ArraryFire.jl functions well.
denizyuret
commented
7 years ago Can you send me the error message, versioninfo(), Pkg.status()?
Did you try Pkg.build("Knet") after update?
Did you try a clean install: just set JULIA_PKGDIR to a temp directory and try Pkg.add("Knet") there. On Sun, May 21, 2017 at 2:27 PM Yao Lu notifications@github.com wrote:
That's strange.
— You are receiving this because you commented.
Reply to this email directly, view it on GitHub https://github.com/denizyuret/Knet.jl/issues/125#issuecomment-302930798, or mute the thread https://github.com/notifications/unsubscribe-auth/ABvNphJv94f5BwnLvGMSDTsy-ilXfb8cks5r8B-XgaJpZM4NheG9 .
AStupidBear
commented
7 years ago Some sample errors for a clean install
indexing: Test Failed
Expression: a[i...] == k[i...]
Evaluated: [0.958081 0.638335 0.764553 0.548269; 0.921487 0.505448 0.739217 0.166488] == Knet.KnetArray{Float64,2}(Knet.KnetPtr(Ptr{Void} @0x
0000002309874c00,64,1,nothing),(2,4))
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:45 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: a == k
Evaluated: [0.0 0.0 0.0 0.0; 0.160203 0.486415 0.988154 0.918841; 0.0 0.0 0.0 0.0] == Knet.KnetArray{Float64,2}(Knet.KnetPtr(Ptr{Void} @0x000
000230985f800,96,1,nothing),(3,4))
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:49 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: gradcheck(getindex,k,i...)
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:54 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: a[i...] == k[i...]
Evaluated: [0.958081 0.764553; 0.160203 0.988154; 0.921487 0.739217] == Knet.KnetArray{Float64,2}(Knet.KnetPtr(Ptr{Void} @0x000000230985f600,
48,1,nothing),(3,2))
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:45 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: a == k
Evaluated: [0.0 0.638335 0.0 0.548269; 0.0 0.486415 0.0 0.918841; 0.0 0.505448 0.0 0.166488] == Knet.KnetArray{Float64,2}(Knet.KnetPtr(Ptr{Vo
id} @0x000000230985f800,96,1,nothing),(3,4))
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:49 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: gradcheck(getindex,k,i...)
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:54 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: a[i...] == k[i...]
Evaluated: [0.958081,0.921487] == Knet.KnetArray{Float64,1}(Knet.KnetPtr(Ptr{Void} @0x000000230987da00,16,1,nothing),(2,))
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:45 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: a == k
Evaluated: [0.0 0.638335 0.764553 0.548269; 0.160203 0.486415 0.988154 0.918841; 0.0 0.505448 0.739217 0.166488] == Knet.KnetArray{Float64,2}
(Knet.KnetPtr(Ptr{Void} @0x000000230985f800,96,1,nothing),(3,4))
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:49 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: gradcheck(getindex,k,i...)
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:54 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: a[i...] == k[i...]
Evaluated: [0.160203,0.160203] == Knet.KnetArray{Float64,1}(Knet.KnetPtr(Ptr{Void} @0x0000002309886400,16,1,nothing),(2,))
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:45 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: a == k
Evaluated: [0.958081 0.638335 0.764553 0.548269; 0.0 0.486415 0.988154 0.918841; 0.921487 0.505448 0.739217 0.166488] == Knet.KnetArray{Float
64,2}(Knet.KnetPtr(Ptr{Void} @0x000000230985f800,96,1,nothing),(3,4))
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:49 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times)
in process_options(::Base.JLOptions) at ./client.jl:262
in _start() at ./client.jl:318
indexing: Test Failed
Expression: gradcheck(getindex,k,i...)
in record(::Base.Test.DefaultTestSet, ::Base.Test.Fail) at ./test.jl:428
in do_test(::Base.Test.Returned, ::Expr) at ./test.jl:281
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:54 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:22 [inlined]
in macro expansion; at ./test.jl:672 [inlined]
in macro expansion; at /tmp/v0.5/Knet/test/karray.jl:14 [inlined]
in anonymous at ./<missing>:?
in include_from_node1(::String) at ./loading.jl:488 (repeats 2 times) julia> versioninfo()
Julia Version 0.5.0
Commit 3c9d753 (2016-09-19 18:14 UTC)
Platform Info:
System: Linux (x86_64-pc-linux-gnu)
CPU: Intel(R) Xeon(R) CPU E5-2676 v3 @ 2.40GHz
WORD_SIZE: 64
BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Haswell)
LAPACK: libopenblas64_
LIBM: libopenlibm
LLVM: libLLVM-3.7.1 (ORCJIT, haswell) julia> Pkg.status()
1 required packages:
- Knet 0.8.3
2 additional packages:
- AutoGrad 0.0.7
- Compat 0.25.2These look normal. How about cuda library versions? Can get those using include(Pkg.dir("Knet/test/gpu.jl")). And the model of gpu (run nvidia-smi in shell). Also try "make clean; make" under .julia/v0.5/Knet/src.
On Sun, May 21, 2017, 17:08 Yao Lu notifications@github.com wrote:
julia> Pkg.status() 1 required packages:
- Knet 0.8.3 2 additional packages:
- AutoGrad 0.0.7
- Compat 0.25.2
— You are receiving this because you commented.
Reply to this email directly, view it on GitHub https://github.com/denizyuret/Knet.jl/issues/125#issuecomment-302938982, or mute the thread https://github.com/notifications/unsubscribe-auth/ABvNpsiXjCW_DRUajwkUlEiLSY5qoBMmks5r8EVKgaJpZM4NheG9 .
AStupidBear
commented
7 years ago julia> include(Pkg.dir("Knet/test/gpu.jl"))
(Knet.cudaDriverVersion,Knet.cudaRuntimeVersion,Knet.cublasVersion,Knet.cudnnVersion) = (8000,8000,8000,5105)
Test Summary: | Pass Total
gpu | 8 8+-----------------------------------------------------------------------------+
| NVIDIA-SMI 375.39 Driver Version: 375.39 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce GTX 980 Ti Off | 0000:03:00.0 On | N/A |
| 26% 38C P8 21W / 260W | 44MiB / 6075MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 1 GeForce GTX 980 Ti Off | 0000:81:00.0 Off | N/A |
| 20% 44C P8 16W / 260W | 3MiB / 6078MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 1871 G /usr/lib/xorg/Xorg 40MiB |
+-----------------------------------------------------------------------------+After running make clean; make the problem is still there.
Have you tested Knet on other computers with GPUs? I cannot access other testing platforms.
denizyuret
commented
7 years ago There is a machine image Knet-0.8.3 on amazon aws ohio. It has also been tested on K20, K40, K80 gpus, other gpus on amazon, and cpu only linux and macs.
Your errors seem to be with float64 arrays. Is it possible your gpu does not support float64?
On Mon, May 22, 2017, 16:43 Yao Lu notifications@github.com wrote:
After running make clean; make the problem is still there.
Have you tested Knet on other computers with GPUs? I cannot access other testing platforms.
— You are receiving this because you commented.
Reply to this email directly, view it on GitHub https://github.com/denizyuret/Knet.jl/issues/125#issuecomment-303103889, or mute the thread https://github.com/notifications/unsubscribe-auth/ABvNptecp_nFqNgIyi6RFnoQQ39BZiBuks5r8ZEMgaJpZM4NheG9 .
AStupidBear
commented
7 years ago julia> using Knet; a = KnetArray(ones(Float32, 10))
10-element Knet.KnetArray{Float32,1}:
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
julia> a + a
10-element Knet.KnetArray{Float32,1}:
2.29589f-41
2.3319f-41
2.36819f-41
2.40477f-41
2.44162f-41
2.47876f-41
2.51617f-41
2.55387f-41
2.59184f-41
2.6301f-41julia> using ArrayFire; setBackend(AF_BACKEND_CUDA); a = AFArray(ones(Float32, 10)); a + a
10-element ArrayFire.AFArray{Float32,1}:
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0
2.0Does this mean my hardware and nvidia driver are OK?
AStupidBear
commented
7 years ago The test for CUBLAS.jl also passed.
julia> Pkg.test("CUBLAS")
INFO: Testing CUBLAS
WARNING: Method definition test_trsm(Any) in module Main at /home/luyao/.julia/v0.5/CUBLAS/test/runtests.jl:1362 overwritten at /home/luyao/.julia/v0.5/CUBLAS/test/runtests.jl:1417.
INFO: CUBLAS tests passedHi Yao: Thanks for your support debugging this issue. Can you confirm that Knet-0.8.2 and AutoGrad-0.0.6 were working ok before the upgrade?
Enis, I think there is a problem with our new kernels running on the GTX-980 GPU. Are you using any instructions that are not supported on this platform? If so can we make them conditional on the chipset? Yao has found out that a+b is not working for KnetArrays!
best, deniz
On Tue, May 23, 2017 at 3:57 AM Yao Lu notifications@github.com wrote:
The test https://github.com/JuliaGPU/CUBLAS.jl/blob/master/test/runtests.jl for CUBLAS.jl also passed.
julia> Pkg.test("CUBLAS") INFO: Testing CUBLAS WARNING: Method definition test_trsm(Any) in module Main at /home/luyao/.julia/v0.5/CUBLAS/test/runtests.jl:1362 overwritten at /home/luyao/.julia/v0.5/CUBLAS/test/runtests.jl:1417. INFO: CUBLAS tests passed
— You are receiving this because you commented.
Reply to this email directly, view it on GitHub https://github.com/denizyuret/Knet.jl/issues/125#issuecomment-303260049, or mute the thread https://github.com/notifications/unsubscribe-auth/ABvNpo5gb0D9JLc2Mgca3pnusMICKVAnks5r8i7sgaJpZM4NheG9 .
EnisBerk
commented
7 years ago Hi All, a+a is same size array addition so it is not computed by new kernels, it is computed by src/cuda11.jl, Yao, can you try with arrays of different sizes or share results of test/broadcast.jl
other than that we have "-arch=compute_30 -code=sm_30 " flags in line 12 of Makefile for minimum compute architecture used in cuda13.jl, but in code we handled those cases with conditionals. Also Yao's architecture is suitable (NVIDIA-SMI 375.39 ).
Should we remove "-arch=compute_30 -code=sm_30 " and test again ?
best, Enis Berk
denizyuret
commented
7 years ago I think the problem may not be with the driver version but the compute capability of the chip itself: GTX-980. Do we have any way to access machines with that compute capability for testing?
On Tue, May 23, 2017 at 10:05 AM Enis Berk Çoban notifications@github.com wrote:
Hi All, a+a is same size array addition so it is not computed by new kernels, it is computed by src/cuda11.jl, Yao, can you try with arrays of different sizes or share results of test/broadcast.jl
other than that we have "-arch=compute_30 -code=sm_30 " flags in line 12 of Makefile for minimum compute architecture used in cuda13.jl, but in code we handled those cases with conditionals. Also Yao's architecture is suitable (NVIDIA-SMI 375.39 ).
Should we remove "-arch=compute_30 -code=sm_30 " and test again ?
best, Enis Berk
— You are receiving this because you commented.
Reply to this email directly, view it on GitHub https://github.com/denizyuret/Knet.jl/issues/125#issuecomment-303309239, or mute the thread https://github.com/notifications/unsubscribe-auth/ABvNpicQRF0TqIdKKRma6tZOWqT1Jw05ks5r8oVLgaJpZM4NheG9 .
AStupidBear
commented
7 years ago After using
Pkg.pin("Knet", v"0.8.2")
Pkg.pin("AutoGrad", v"0.0.6")and
cd ~/.julia/v0.5/Knet/src && make clean && makethe test passed.
julia> Pkg.test("Knet")
INFO: Computing test dependencies for Knet...
INFO: Installing BaseTestNext v0.2.2
INFO: Testing Knet
INFO: Knet using GPU 1
Test Summary: | Pass Total
kptr | 1 1
1.878766 seconds (1.12 M allocations: 49.284 MB, 13.09% gc time)
(Knet.cudaDriverVersion,Knet.cudaRuntimeVersion,Knet.cublasVersion,Knet.cudnnVersion) = (8000
Test Summary: | Pass Total
gpu | 8 8
0.895661 seconds (207.56 k allocations: 8.691 MB, 0.55% gc time)
Test Summary: | Pass Total
distributions | 3 3
1.559528 seconds (1.03 M allocations: 43.159 MB, 1.02% gc time)
Test Summary: | Pass Total
karray | 122 122
12.671813 seconds (7.98 M allocations: 332.267 MB, 0.86% gc time)
Test Summary: | Pass Total
linalg | 112 112
18.156013 seconds (9.72 M allocations: 405.184 MB, 0.76% gc time)
Test Summary: | Pass Total
broadcast | 3624 3624
30.212741 seconds (13.17 M allocations: 575.725 MB, 1.04% gc time)
Test Summary: | Pass Total
reduction | 1188 1188
25.244122 seconds (13.50 M allocations: 561.333 MB, 0.87% gc time)
Test Summary: | Pass Total
conv | 139 139
57.717182 seconds (15.43 M allocations: 623.457 MB, 0.56% gc time)
Test Summary: | Pass Total
unary | 1720 1720
41.423803 seconds (13.89 M allocations: 587.040 MB, 0.71% gc time) Hi Yao: could you try Enis' suggestion and recompile after changing the default options in Knet/src/Makefile: Should we remove "-arch=compute_30 -code=sm_30 " and test again ?
AStupidBear
commented
7 years ago @denizyuret @EnisBerk Thanks for your suggestions. The test passed after removing "-arch=compute_30 -code=sm_30" and make again.
denizyuret
commented
7 years ago bugun 9 11 arasi musait.
On Tue, May 23, 2017, 10:05 Enis Berk Çoban notifications@github.com wrote:
Hi All, a+a is same size array addition so it is not computed by new kernels, it is computed by src/cuda11.jl, Yao, can you try with arrays of different sizes or share results of test/broadcast.jl
other than that we have "-arch=compute_30 -code=sm_30 " flags in line 12 of Makefile for minimum compute architecture used in cuda13.jl, but in code we handled those cases with conditionals. Also Yao's architecture is suitable (NVIDIA-SMI 375.39 ).
Should we remove "-arch=compute_30 -code=sm_30 " and test again ?
best, Enis Berk
— You are receiving this because you commented.
Reply to this email directly, view it on GitHub https://github.com/denizyuret/Knet.jl/issues/125#issuecomment-303309239, or mute the thread https://github.com/notifications/unsubscribe-auth/ABvNpicQRF0TqIdKKRma6tZOWqT1Jw05ks5r8oVLgaJpZM4NheG9 .
AStupidBear
commented
7 years ago This issue also occurs for Titan X Pascal.
klowrey
commented
7 years ago If it helps, I was having a similar issue on a Maxwell Titan X after updating Knet.
Ubuntu 16.04 Cuda 8.0 Nvidia Driver 375.66
Packages CUDABLAS and CUDAdrv were failing due to not having gcc-4.7 and g++-4.7 installed on my system, so this probably affected Knet in some way.
apt install gcc-4.7 g++-4.7 # and recommended packages
then make clean; make in Knet/src directory fixed things for me (but Pkg.rm- add- build-("Knet") didnt...)
Thanks, Knet team!
denizyuret
commented
7 years ago The problem compiler options have been removed in latest release.
After using
Pkg.update()yesterday, Knet.jl no longer works anymore for KnetArray.Pkg.test("Knet")failed, andMy configurations is:
Ubuntu 14.04 Julia 0.5.0 AutoGrad.jl master Knet.jl master
If you don't have this issue, then it may be a coincidence that my gpu crashed at the same time of updating.