Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Arx]

Raaagh I am trying to install Torch and it hates me.

Mainly wanted to get that off my chest, but if you have any advice that's welcome too.

Of course, I actually have no idea how much of my trouble is caused by dicky* varsity wifi.

*I just realised this may cause confusion. The slang meaning I am aiming for here is 'unreliable' not 'impolite'.

[Orange Wizard]

I finally figured out how to work Cython, and it's fast. My shittily-optimised code runs at an acceptable speed. I'm so happy.

[Bauglir]

Raaagh I am trying to install Torch and it hates me.

I found this to be a pain in the ass myself. What error messages is it shitting out at you? Also, what operating system, and are you trying to use a graphics card with it?

[Moghjubar]

Got Vulkan to actually work last night after a bunch of trouble [understatement].  Said trouble involved yet another AMD beta driver update (they didn't even support 1.0.3 out of the gate so demos didnt work) and gnashing of teeth (trying to get it to work manually in mingw and converting stuff over and I got all the way near the end and I got hard crashes on function calls, yaaay).  Finally reinstalled Visual Studio to get the C++ pack (and everything else: turns out if you checkmark everything so you don't have to reinstall ever again its like 40+ gigs).

Man, Vulkan is definitely a marathon.  I don't know if I can go back to it for awhile (still gotta figure out exacly where I messed up).  Fair warning for anyone that dives into it: get help if you can.

[alway]

Yeah, Vulkan's target audience is mostly graphics programmers who know enough to write graphics drivers. That's literally why it exists, so graphics programmers can take over some of the job of the driver programmers and tailor the low level details to their needs.

[EnigmaticHat]

I finally figured out how to work Cython, and it's fast. My shittily-optimised code runs at an acceptable speed. I'm so happy.

Nice!

I wouldn't worried about shittily-optimised code.  Some people think there's a right way and a wrong way but the truth is optimization is an infinitely deep well and its never truly finished.

[Arx]

Raaagh I am trying to install Torch and it hates me.

I found this to be a pain in the ass myself. What error messages is it shitting out at you? Also, what operating system, and are you trying to use a graphics card with it?

luajit: ./th:104: module 'trepl' not found:
   no field package.preload['trepl']
   no file './trepl.lua'
   no file '/usr/share/luajit-2.0.2/trepl.lua'
   no file '/usr/local/share/lua/5.1/trepl.lua'
   no file '/usr/local/share/lua/5.1/trepl/init.lua'
   no file '/usr/share/lua/5.1/trepl.lua'
   no file '/usr/share/lua/5.1/trepl/init.lua'
   no file './trepl.so'
   no file '/usr/local/lib/lua/5.1/trepl.so'
   no file '/usr/lib/x86_64-linux-gnu/lua/5.1/trepl.so'
   no file '/usr/lib/lua/5.1/trepl.so'
   no file '/usr/local/lib/lua/5.1/loadall.so'
stack traceback:
   [C]: in function 'require'
   ./th:104: in main chunk
   [C]: at 0x00404d60

On Linux Mint 17, and nope, I'm not trying to run it with a GPU.

Like, I assume that means that I don't have trepl installed, which isn't unreasonable (even though Torch is supposed to install trepl in their simplified install procedure). What does seem a bit odd is that the installation instructions for trepl are 'luarocks install trepl', which is an issue because luarocks doesn't seem to have trepl.

This is after I crawled through way too many hoops, including that LuaRocks does not seem to have all the stuff they think it does.

[Bauglir]

... Okay that's beyond my experience. In a sane and just world, luarocks would indicate that it couldn't connect to the repository to download stuff, which would distinguish it from situations where the repository simply doesn't have it. So you'd get something like "couldn't connect" instead of "couldn't find module". But I don't actually know if it does that, and it certainly should have those things.

You probably found this already, but maybe start a thread?

[Arx]

whaaaa

I just ran the Torch installer for the thousandth time and it inexplicably worked. Whaaaaat.

[jaked122]

whaaaa

I just ran the Torch installer for the thousandth time and it inexplicably worked. Whaaaaat.

What are you going to use that for? I've seen a lot about it recently in regards to machine learning, but what do you want to do with it?

[Arx]

whaaaa

I just ran the Torch installer for the thousandth time and it inexplicably worked. Whaaaaat.

What are you going to use that for? I've seen a lot about it recently in regards to machine learning, but what do you want to do with it?

Machine learning. Specifically, recurrant neural nets for character-level text modelling.

I'm using this specific widget to try to fake LSP here (using sample code), but I've been messing around with neural nets for quite a while. There are a few other applications I have in mind for this (since what it does is essentially map one vector of any size onto another vector of any size), but I have got to insufficient grips with it to try them.

E: Recursive and recurrent may have the same root, but they are not the same. Whoops.

[Bauglir]

I could swear recurrent neural nets were a subset of recursive neural nets, but perhaps my memory is flubbing this.

[Arx]

Better to be specific, either way.

[Magnumcannon]

Alright, so i'm trying to make a small python program which calculates quadratic equations. It's very buggy and inaccurate. Can someone give me a clue of why it is so inaccurate? I made this months ago and now i tried to improve it, but it gives me a "math domain error"

Code: [Select]

import math

a = int (raw_input ("Qual e o coeficiente 'a'?: "))
b = int (raw_input ("Qual e o coeficiente 'b'?: "))
c = int (raw_input ("Qual e o coeficiente 'c'?: "))
delta = (b**2)-4*a*c
xI = ((-b + (math.sqrt (delta))/2*a))
xII = ((-b - (math.sqrt (delta))/2*a))
xv = -b/(2*a)
yv = -delta/(4*a)
a_ccv = "cima" if a > 0 else "baixo"
b_crs = "crescendo" if b > 0 else "decrescendo"
n_raiz = "2 raizes diferentes" if delta > 0 else "2 raizes iguais" if delta == 0 else "nenhuma raiz real"
if (delta < 0):
    print ("a = %i b = %i, c = %i, delta nao e real (%i), xI nao e real e xII nao e real" %(a,b,c, delta))
else:
           print ("a = %i b = %i, c = %i, delta = %i, xI = %i e xII = %i" %(a,b,c,delta,xI,xII))       
print ("Xv = %i, Yv = %i" %(xv,yv))
print ("A concavidade da parabola se voltara para %s" %a_ccv)
print ("A parabola antingira o eixo y enquanto estiver %s" %b_crs)
print ("No eixo x ha %s" %n_raiz)
end0 = raw_input ('Digite qualquer coisa para sair')


[Orange Wizard]

math.sqrt() doesn't work with negative numbers (domain error), so probably delta is negative.