Brian Broll
09b6ca0ad0
WIP #3. Added import tests WIP #3 Added more test-cases WIP #3 Added more tests WIP #3. Fixed concat.lua test WIP #3 minor changes WIP #3 Fixed concat-parallel.lua WIP #3 Added check-model helper WIP #3 Added more tests for model checker WIP #3 Added extra tests WIP #3 Changed check-model to GraphChecker WIP #3. multiple cases fail for ImportTorch... WIP #3 Fixed ImportTorch batch test case running WIP #3 Changed graph checker to use gme path for id WIP #3 Updated tests WIP #3. Tweaked to get all examples working locally w/ 'th' WIP #3 Fixed tests
52 linhas
1.4 KiB
Lua
52 linhas
1.4 KiB
Lua
-- thanks to https://github.com/soumith/imagenet-multiGPU.torch for this example
|
|
|
|
-- Achieves 62.6% top1 on validation set at 35 epochs with this regime:
|
|
-- { 1, 9, 1e-1, 5e-4, },
|
|
-- { 10, 19, 1e-2, 5e-4 },
|
|
-- { 20, 25, 0.001, 0 },
|
|
-- { 26, 30, 1e-4, 0 },
|
|
-- Trained model:
|
|
-- https://gist.github.com/szagoruyko/0f5b4c5e2d2b18472854
|
|
|
|
require 'nn'
|
|
local nin = nn.Sequential()
|
|
local function block(...)
|
|
local arg = {...}
|
|
local no = arg[2]
|
|
nin:add(nn.SpatialConvolution(...))
|
|
nin:add(nn.SpatialBatchNormalization(no,0.001))
|
|
nin:add(nn.ReLU(true))
|
|
nin:add(nn.SpatialConvolution(no, no, 1, 1, 1, 1, 0, 0))
|
|
nin:add(nn.SpatialBatchNormalization(no,0.001))
|
|
nin:add(nn.ReLU(true))
|
|
nin:add(nn.SpatialConvolution(no, no, 1, 1, 1, 1, 0, 0))
|
|
nin:add(nn.SpatialBatchNormalization(no,0.001))
|
|
nin:add(nn.ReLU(true))
|
|
end
|
|
|
|
local function mp(...)
|
|
nin:add(nn.SpatialMaxPooling(...))
|
|
end
|
|
|
|
block(3, 96, 11, 11, 4, 4, 5, 5)
|
|
mp(3, 3, 2, 2, 1, 1)
|
|
block(96, 256, 5, 5, 1, 1, 2, 2)
|
|
mp(3, 3, 2, 2, 1, 1)
|
|
block(256, 384, 3, 3, 1, 1, 1, 1)
|
|
mp(3, 3, 2, 2, 1, 1)
|
|
block(384, 1024, 3, 3, 1, 1, 1, 1)
|
|
|
|
nin:add(nn.SpatialAveragePooling(7, 7, 1, 1))
|
|
nin:add(nn.View(-1):setNumInputDims(3))
|
|
|
|
local model = nn.Sequential()
|
|
-- :add(makeDataParallel(nin, nGPU))
|
|
:add(nin)
|
|
:add(nn.Linear(1024,1000))
|
|
:add(nn.LogSoftMax())
|
|
|
|
model.imageSize = 256
|
|
model.imageCrop = 224
|
|
|
|
return model-- :cuda()
|