libmir/public

Mathias L. Baumann

@Marenz

I still haven't figured out all those details

Mind if I create PRs that add what you just told me?

for the docu

Ilya Yaroshenko

@9il

Sure, hope you will improve English

Mathias L. Baumann

@Marenz

I didn't notice any weird English so far. But when I do I'll fix it

btw, I wrote my own mtimes() now that works like the one in lubeck but uses mir-glas and also expects an allocator to be passed

That way I can use e.g. the Region allocator for loops to reuse the memory

if you're curious: https://gist.github.com/Marenz/9f734331c01f65c19510fef93796830a

Ilya Yaroshenko

@9il

:+1:

Nicholas Wilson

@thewilsonator

@Marenz you should make alignment an enum, it doesn't change.

Nicholas Wilson

@thewilsonator

First set of docs for dcompute: libmir/dcompute#37

Mathias L. Baumann

@Marenz

I am trying to port a python example for a neural net with a hidden layer to D2/mir

And.. everything seems to be working, only that it.. doesn't

This my program: https://github.com/Marenz/neural_net_examples/blob/master/lstm/source/app.d

This is what I want to port it from: https://iamtrask.github.io/2015/11/15/anyone-can-code-lstm/?deDup267640=1&deDup936550=1

As far as I can tell, my code is identical

It's a lot to ask to compare those and get feedback on the code, so I understand if you feel like you don't want to ;)-

I just thought I try asking here because I am a bit out of ideas currently

Ilya Yaroshenko

@9il

Hi Mathias,
I would suggest to do step by step verification. It is common for SCI code to do so.
After quick review i found one small bug (but there may be also others): Random gen is not initialized.
You may want to use https://github.com/libmir/mir-random as wall as https://github.com/libmir/numir to generate random slices. Mir Random has default contrcuttor disabled for RNGs, so it is safer then Phobos.

Mathias L. Baumann

@Marenz

it actually is, here: https://github.com/Marenz/neural_net_examples/blob/master/lstm/source/app.d#L64 if you mean seed when you say initialized?

I tried to construct it in a way that it matches the random numbers generated in python

I got surprisingly close

to see where it starts diverging

but there was no clear point

Ilya Yaroshenko

@9il

Ah, ok. BTW keep in mind that gen is TLS variable

Mathias L. Baumann

@Marenz

I don't plan on more threads yet :)

Francis Nixon

@pirocks

Hello. I'm having trouble compiling dcompute on debian 9. My dub.json looks like this:

{
    "name": "compute-messing-around",
    "authors": [
        "Francis Nixon"
    ],
    "description": "A minimal D application.",
    "license": "proprietary",
    "dependencies": {
        "dcompute": "~>0.1.0"
    },
    "dflags": ["-mdcompute-targets=ocl-210,cuda-350","-oq"]
}

The error I'm currently getting is:

../../.dub/packages/dcompute-0.1.0/dcompute/source/dcompute/driver/error.d(143,13): Error: undefined identifier `fprintf`
../../.dub/packages/dcompute-0.1.0/dcompute/source/dcompute/driver/ocl/context.d(144,19): Error: undefined identifier `clCreateProgramWithIL`

Nicholas Wilson

@thewilsonator

Thats embarrassing. libmir/dcompute@cee8eb3

I'm not sure why your clCreateProgramWithIL can't be found. Are you using an up to date DerelictCL?

Francis Nixon

@pirocks

The clCreateProgramWithIL error went away after manually selecting the most recent version of DerelictCL. I then got an error in the same place as the fprintf error, except for toStringz. Adding an import fixed that, but now I'm getting the following:

../../.dub/packages/dcompute-0.1.0/dcompute/source/dcompute/driver/error.d(139,32): Error: cannot implicitly convert expression `__lambda1` of type `void delegate(Status _status) @system` to `immutable(void delegate(Status) nothrow @nogc)`
../../.dub/packages/dcompute-0.1.0/dcompute/source/dcompute/driver/error.d(139,32): Error: cannot implicitly convert expression `__lambda1` of type `void delegate(Status _status) @system` to `immutable(void delegate(Status) nothrow @nogc)`

Nicholas Wilson

@thewilsonator

Hmm, that does seem odd, but that's my fault for not testing it properly.

Francis Nixon

@pirocks

If relevant my ldc version is:

LDC - the LLVM D compiler (1.7.0git-958e58c):
  based on DMD v2.077.1 and LLVM 3.8.1
  built with DMD64 D Compiler v2.077.1
  Default target: x86_64-pc-linux-gnu
  Host CPU: broadwell

Nicholas Wilson

@thewilsonator

Unless you have need to use the D_betterC version try without it. That should "work" (note the @BUG@ just above). You will need to set onDriverError yourself because of it, see e.g.

https://github.com/libmir/dcompute/blob/master/source/dcompute/tests/main.d#L40

Thats not a compiler problem, thats me not testing properly. I really need to set up CI, but given the hardware required I haven't got around to it yet.

Thanks for pointing out the issues.

Francis Nixon

@pirocks

Without -betterC I get:

Invalid bitcast
  %3 = bitcast float addrspace(1)* %res_arg to float*
Invalid bitcast
  %5 = bitcast float addrspace(1)* %x_arg to float*
Invalid bitcast
  %9 = bitcast float addrspace(1)* %y_arg to float*
LLVM ERROR: Broken function found, compilation aborted!

Nicholas Wilson

@thewilsonator

Hmm, can you try with an LLVM that is 3.9 or greater (e.g. from the LDC release page)?

Also what registered targets does the LDC you are using have (just below the output of ldc2 --version you posted)?

Sebastian Wilzbach

@wilzbach

I have two good news:
1) libmir/mir-algorithm#122 - examples on the mir docs will be runnable soon (see http://files.wilzbach.me/dlang/mir-algorithm/mir_ndslice_algorithm.html)
2) https://tour.dlang.org/tour/en/dub/mir - the tour will be finally moving to integrate mir (thought writing a good one-page summary might turn out to be challenging)

Ilya Yaroshenko

@9il

Some news:
PR "Tarjan graph algorithm" libmir/mir-algorithm#121
Issue "ndslice based API for dopt" henrygouk/dopt#6
Issue "Dcompute based backend for dopt" henrygouk/dopt#6

Ilya Yaroshenko

@9il

EDIT: Issue "ndslice based API for dopt" henrygouk/dopt#7

Ilya Yaroshenko

@9il

Mir Random PR https://github.com/libmir/mir-random/pull/73/files

kerdemdemir

@kerdemdemir

Hi Everybody I am trying to use multivariateNormalVar in mir.random.ndvariable for creating two separated data clusters (it is done in pyhton like: https://beckernick.github.io/logistic-regression-from-scratch/)

It doesn't work if I don't select resulting vector same size as sigma matrix

double[100] x1;   // Only works if the size is only 2 
Random* gen = threadLocalPtr!Random;
auto mu = [0.0, 0.0].sliced;
auto sigma = [1.0, 0.75, 0.75, 1].sliced(2,2);
auto rv = multivariateNormalVar(mu, sigma);
rv(gen, x1[]);

How can I use multivariateNormalVar to create data sets bigger than size 2

Shigeki Karita

@ShigekiKarita

how about this? (3dim random normal x 10)

/+dub.sdl:
dependency "lubeck" version="~>0.0.4"
dependency "numir" version="~>0.1.0"
libs "blas"
+/

import mir.ndslice : map, sliced, slicedField, ndarray;
import mir.random : threadLocalPtr, Random;
import mir.random.variable : NormalVariable;
import mir.random.algorithm : field;
import lubeck : mtimes;

import numir : alongDim;
import std.stdio;

void main() {
    Random* gen = threadLocalPtr!Random;
    auto mu = [0.0, 0.0, 0.0].sliced;
    auto sigma = [1.0, 0.75, 0.0,
                  0.75, 1.0, 0.75,
                  0.0, 0.75, 1].sliced(3,3);
    auto xs = field(gen, NormalVariable!double(0, 1)).slicedField(10, 3);
    auto x1 = xs.mtimes(sigma).alongDim!1.map!(x => x + mu).ndarray;
    x1.writeln; // 10 x 3 dim
}

https://run.dlang.io/gist/bd6dd9a2f6606151c707a6bdf6d0be36?compiler=ldc&args=-release
The multi normal random value is just an affine transformation of the standard normal random values. https://en.wikipedia.org/wiki/Multivariate_normal_distribution#Affine_transformation

for creating two separated data clusters

Oh you wanna create two clusters! you can take my example in https://github.com/ShigekiKarita/d-tree/blob/master/example/plot_boundary/app.d

it looks like this

Where communities thrive

People

Repo info

Activity