import cudaconv2 # driver.init() # device_info = (0, 0) # for i in range(driver.Device.count()): # dev = driver.Device(i) # ctx = dev.make_context() # ctx.push() # free, total = driver.mem_get_info() # print 'Free Memory for Device', i, 'is', free / 1000000, 'MB' # # if device_info[1] < free: # device_info = (i, free) # # ctx.pop() # ctx.detach() # print 'Choose Device', device_info[0] # dev = driver.Device(device_info[0]) CONTEXT = cudaconv2.init()
import cudaconv2 cudaconv2.init() import pycuda.driver as cuda import sys from pycuda import gpuarray, driver import cudaconv2 import numpy as np from scipy.signal import convolve2d imgSize = 32 filterSize = 5 padding = 2 color = 1 imgNum = 1 filterNum = 64 stride = 1 modulesX = 1 + int(((2 * padding + imgSize - filterSize) / float(stride))) print 'Modules X', modulesX img = gpuarray.to_gpu(np.ones((imgSize * imgSize * color, imgNum)).astype(np.float32)) filter = gpuarray.to_gpu(np.ones((filterSize * filterSize * color, filterNum)).astype(np.float32)) target = gpuarray.to_gpu(np.ones((modulesX * modulesX * filterNum, imgNum)).astype(np.float32)) print 'standard output for convolution' print convolve2d(np.ones((imgSize, imgSize)).astype(np.float32), np.ones((filterSize, filterSize)).astype(np.float32),'valid') cudaconv2.convFilterActs(img, filter, target, imgSize, modulesX, modulesX, -padding, stride, color, 1, 0.0, 1.0)
#!/usr/bin/env python
'''A relatively simple distributed network implementation, using async SGD.'''
from fastnet import net, layer, data, parser, weights
from fastnet.util import EZTimer
from mpi4py import MPI
import ctypes
import cudaconv2
import numpy as np
import os
WORLD = MPI.COMM_WORLD
cudaconv2.init(WORLD.Get_rank())
print 'CUDA', os.environ.get('MV2_USE_CUDA')
MASTER = 0
WORKERS = range(1, WORLD.Get_size())
batch_size = 128
data_dir = '/ssd/nn-data/imagenet/'
data_provider = 'imagenet'
checkpoint_dir = './checkpoint'
param_file = 'config/imagenet.cfg'
train_range = range(101, 1301)
test_range = range(1, 101)
import cudaconv2 # driver.init() # device_info = (0, 0) # for i in range(driver.Device.count()): # dev = driver.Device(i) # ctx = dev.make_context() # ctx.push() # free, total = driver.mem_get_info() # print 'Free Memory for Device', i, 'is', free / 1000000, 'MB' # # if device_info[1] < free: # device_info = (i, free) # # ctx.pop() # ctx.detach() # print 'Choose Device', device_info[0] # dev = driver.Device(device_info[0]) CONTEXT = cudaconv2.init()
#!/usr/bin/env python
'''A relatively simple distributed network implementation, using async SGD.'''
from fastnet import net, layer, data, parser, weights
from fastnet.util import EZTimer
from mpi4py import MPI
import ctypes
import cudaconv2
import numpy as np
import os
WORLD = MPI.COMM_WORLD
cudaconv2.init(WORLD.Get_rank())
print 'CUDA', os.environ.get('MV2_USE_CUDA')
MASTER = 0
WORKERS = range(1, WORLD.Get_size())
batch_size = 128
data_dir = '/ssd/nn-data/imagenet/'
data_provider = 'imagenet'
checkpoint_dir = './checkpoint'
param_file = 'config/imagenet.cfg'
train_range = range(101, 1301)
test_range = range(1, 101)
data_provider = 'imagenet'
