soumith

$ CUDA_VISIBLE_DEVICES=0 th -lcutorch -e "cutorch.test('sigmoid1')"
seed:   1471924845
Running 1 test
1/1 sigmoid1 ............................................................ [WAIT]torch.CudaTensor
input CPU:      0.31726333498955
input GPU:      0.31726333498955
Result CPU:     0.57865715026855
Result GPU:     0.5786572098732
torch.CudaDoubleTensor
input CPU:      0.31726332521066

soumith

import torch.multiprocessing as mp
from torch.multiprocessing import Semaphore
import sys


if sys.version_info[0] == 3:
    Barrier = mp.Barrier
else: # version 2
    # from http://stackoverflow.com/a/26703365/117844
    class Barrier:

soumith

import torch
import torch.nn as nn
import torch.nn.parallel

class DCGAN_D(nn.Container):
    def __init__(self, isize, nz, nc, ndf, ngpu, n_extra_layers=0):
        super(DCGAN_D, self).__init__()
        self.ngpu = ngpu
        assert isize % 16 == 0, "isize has to be a multiple of 16"

soumith

soumith

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable as Var

class TreeDecoder(nn.Module):
    NODE_DICT, NODE_LIST, NODE_STR = 0, 1, 2

    def __init__(self, input_size, max_key, max_ident, max_depth, max_length):
        super(TreeDecoder, self).__init__()

soumith

import torch
import torch.nn as nn
from torch.autograd import Variable
import numpy as np

BATCH_SIZE = 1
INPUT_DIM = 1
OUTPUT_DIM = 1
DTYPE = np.float32

soumith

[WARNING]: No mapping options supplied. 'Naive' options will be used which might fail compilation
[WARNING]: Autotuning results won't be cached. 'cache' option is not specified
[WARNING]: Using naive options for autotuning
 
template<typename T> inline __device__ T floord(T n, T d) {
  return n < 0 ? - (-n + d - 1)/d : n / d;
}

// Halide type handling
typedef int int32;

soumith

Scopes 0.3.1 backport   https://github.com/pytorch/pytorch/pull/5153
Cherry pick dataloader issue fix to 0.3.1   https://github.com/pytorch/pytorch/pull/5140
Fixed double memory accesses of several pointwise operations.   https://github.com/pytorch/pytorch/pull/5068
Broadcast output requires_grad only if corresponding input requires_grad   https://github.com/pytorch/pytorch/pull/5061
Fix topk work size computation   https://github.com/pytorch/pytorch/pull/5053
Fix maxpool3d / avgpool3d crashs   https://github.com/pytorch/pytorch/pull/5052
Fix blas addmm (gemm) condition check   https://github.com/pytorch/pytorch/pull/5048
Fix C FFI extension after moving TH to C++   https://github.com/pytorch/pytorch/pull/5005
make torch.set_num_threads also set MKL threads (take 2)   https://github.com/pytorch/pytorch/pull/5002
Fix reduction functions to respect the stride of the output   https://github.com/pytorch/pytorch/pull/4995

soumith

import torch
import torch.nn as nn
from torch.autograd import Variable
torch.manual_seed(1)
# do gradcheck
N = 1
C = 1
D = 1
H = 2
W = 1

soumith

FROM nvidia/cuda:8.0-devel-centos7

ENV LC_ALL en_US.UTF-8
ENV LANG en_US.UTF-8
ENV LANGUAGE en_US.UTF-8

RUN yum install -y wget curl perl cmake util-linux xz bzip2 git

RUN yum install -y patch
RUN yum install -y yum-utils centos-release-scl