and will later pull request later.
Arnaud Bergeron 於 2018年10月19日 週五 上午1:31寫道:
>
> Le 18 oct. 2018 à 00:15, Wong Hang a écrit :
>
> Hi there,
>
> May I ask why theano gpuarray does not support float64 cholesky
> factorization and triangular solve?
> Are there any reason beh
Hi there,
May I ask why theano gpuarray does not support float64 cholesky
factorization and triangular solve?
Are there any reason behind that?
I think I can simply hack the code in theano/gpuarray/linalg.py to
invoke cusolver.cusolverDnDpotrf / cublas.cublasDtrsm / cublas.cublasDtrsv
rather
Thank you very much for your help
Arnaud Bergeron 於 2018年12月11日 週二 上午1:42寫道:
> The only option with Theano would be to write your own op that would run
> in parallel on the GPU.
>
> This is not an easy task.
>
> Le 10 déc. 2018 à 07:35, Wong Hang a écrit :
>
> Hi,
>
Hi,
I am working on a problem that is similar to this one:
import theano
import theano.tensor as T
import theano.tensor.slinalg as Tla
cholesky=Tla.cholesky
solve_lower_triangular=Tla.solve_lower_triangular
import numpy as np
def _LL(k,ret,data,mu,sigma):
jitter = 1e-6*T.eye(3)
L =
he version of cuDNN first?
moran li 於 2019年1月14日 週一 上午11:20寫道:
> Sorry to bother you. Actually, I'm not clear what shall I do after reading
> the contents. Is there any more details about your mentioned solution?
> Thanks again.
>
> Wong Hang 于2019年1月13日周日 下午10:30写道
This may be useful
https://stackoverflow.com/questions/47258882/theano-gpu-support-ptx-jit-compiler-library-not-found
Check the library path.
(I know your error is different from me)
moran li 於 2019年1月13日 週日 下午10:24寫道:
> *Here, I meet this problem while installing theano1.0.3 with
Check http://deeplearning.net/software/theano/library/config.html
and try to disable config.gpuarray.preallocate
moran li 於 2019年1月14日 週一 下午1:37寫道:
> Thanks for your reply. I have upgraded my cudnn version. And the pugpu
> error still exists.
> [image: image.png]
>
>
> Wong H
Hi Paul,
I am now struggling for my own project and therefore I don't think I have
time to work on any side project.
thenao is no longer maintained. One of my friend, who is working as data
scientist , told me that most researchers in academia have already switched
to pytorch.
I recently looked
ay.linalg.GpuCholesky() ) Both used the same precision
>>> (float32) so should give the same results.
>>> But you can see that at the end of the diagonal, the values go wild. It
>>> appears to be numericla errors.
>>> Thanks in advance!
>>> Paul
>>
1, Paul Baggenstoss
>>>> wrote:
>>>>>
>>>>> Hi Simon,I have uploaded the MATLAB format file with the matrix Cll,
>>>>> which is the original matrix, and R_cpu which was produced using CPU by
>>>>> slinalg.Cholesky( ), and R_cu
=27.3912 (max=25.9949)
row #198: err=0.915223 (max=0.915223)
row #199: err=3.60679e-13 (max=2.98261e-13)
494.5201252308407 49.755829752019224
494.5201252308407 49.755829752019224
I attached my test code in this message.
Wong Hang 於 2020年2月7日 週五 下午10:49寫道:
> Hi all,
>
> I found that the
I am quite sure I once get correct result even when the matrix is of size
>1000.
Let me do more research and test later and get back to you.
Wong Hang 於 2020年2月7日 週五 下午11:18寫道:
> I suddenly get the HEAD version of libgpuarray works
> I found that if I increase the size of the matrix,
;
> Nice work,
> Paul
>
>
> On Saturday, February 8, 2020 at 8:49:29 AM UTC+1, Wong Hang wrote:
>>
>> Hi Paul,
>>
>> I think I fixed the issue. Please check the PR
>> https://github.com/Theano/libgpuarray/pull/589
>> and you can try to use my branch of
of your implementation.
[global]
profile = True
Paul Baggenstoss 於 2020年2月8日 週六 上午12:30寫道:
> Hi Wong Hang,
> Yes, that's what I saw, the errors started near the end of the matrix.
> After that, the numbers appeared random.
> I'll try the older version and let you know what I
Hi,
The GPU cholesky decomposition relies on cuSOLVER or Magma. I believe
nvidia knows their hardware well and cuSOLVER should provide the best
efficient result.
Although cholesky decomposition is very numerical stable, when I write the
test case, I find that I will get trouble for relatively
Hi there,
Will your team maintain libgpuarray as well? I have a bug fix in pygpu but
haven't been merged:
https://github.com/Theano/libgpuarray/pull/589
Due to my employment arrangement, I need to seek approval before I can
submit any new PR to open source projects. Would you mind to merge and
16 matches
Mail list logo