-
Notifications
You must be signed in to change notification settings - Fork 18
/
train_conv.lua
279 lines (247 loc) · 10.1 KB
/
train_conv.lua
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
require 'torch'
require 'nn'
require 'optim'
require 'lfs'
require 'gnuplot'
require 'util.print'
local MODEL_ID = torch.randn(1)[1]
local EEGMinibatchLoader = require 'util.EEGMinibatchLoader'
cmd = torch.CmdLine()
cmd:text()
cmd:text('Train a cnn to classify EEG recordings')
cmd:text()
cmd:text('Options')
-- data
cmd:option('-data_dir','data/preprocessed','data directory')
cmd:option('-prepro_dir','data/torch','torch data directory')
-- model prototype
cmd:option('-proto_file', 'cnn/proto/first_cnn.lua', 'file defining network structure')
-- optimization
cmd:option('-optim_algo','rmsprop','optimization algorithm')
cmd:option('-learning_rate',2e-3,'learning rate')
cmd:option('-learning_rate_decay',0.97,'learning rate decay')
cmd:option('-learning_rate_decay_after',10,'in number of epochs, when to start decaying the learning rate')
cmd:option('-decay_rate',0.95,'decay rate for rmsprop')
cmd:option('-dropout',0,'dropout for regularization (0 = no dropout)')
cmd:option('-seq_length',800,'batch length')
cmd:option('-batch_size',2,'number of sequences to train on in parallel')
cmd:option('-window_len',300,'cnn window size')
cmd:option('-max_epochs',50,'number of full passes through the training data')
cmd:option('-grad_clip',5,'clip gradients at this value')
-- checkpoints
cmd:option('-init_from', '', 'initialize network parameters from checkpoint at this path')
-- bookkeeping
cmd:option('-seed',123,'torch manual random number generator seed')
cmd:option('-print_every',1,'how many steps/minibatches between printing out the loss')
cmd:option('-eval_val_every',1000,'every how many iterations should we evaluate on validation data?')
cmd:option('-checkpoint_dir', 'cv', 'output directory where checkpoints get written')
cmd:option('-savefile','cnn','filename to autosave the checkpont to. Will be inside checkpoint_dir/')
-- GPU/CPU
cmd:option('-gpuid',0,'which gpu to use. -1 = use CPU')
cmd:text()
-- parse input params
opt = cmd:parse(arg)
torch.manualSeed(opt.seed)
-- initialize cunn/cutorch for training on the GPU and fall back to CPU gracefully
if opt.gpuid >= 0 then
local ok, cunn = pcall(require, 'cunn')
local ok2, cutorch = pcall(require, 'cutorch')
if not ok then printRed('package cunn not found!') end
if not ok2 then printRed('package cutorch not found!') end
if ok and ok2 then
printGreen('using CUDA on GPU ' .. opt.gpuid .. '...')
cutorch.setDevice(opt.gpuid + 1) -- note +1 to make it 0 indexed! sigh lua
cutorch.manualSeed(opt.seed)
else
printYellow('If cutorch and cunn are installed, your CUDA toolkit may be improperly configured.')
printYellow('Check your CUDA toolkit installation, rebuild cutorch and cunn, and try again.')
printYellow('Falling back on CPU mode')
opt.gpuid = -1 -- overwrite user setting
end
end
-- create the data loader class
local loader = EEGMinibatchLoader.create(opt)
-- make sure output directory exists
if not path.exists(opt.checkpoint_dir) then lfs.mkdir(opt.checkpoint_dir) end
-- define the model: prototypes for one timestep, then clone them in time
local do_random_init = true
local start_iter = 1
if string.len(opt.init_from) > 0 then
printRed('Checkpoints aren\'t supported yet!')
os.exit()
else
print('creating CNN')
dofile(opt.proto_file)
criterion = nn.BCECriterion()
end
-- ship the model to the GPU if desired
if opt.gpuid >= 0 then
cnn:cuda()
criterion:cuda()
end
-- evaluate the loss over an entire split
function eval_split(split_index)
print('evaluating loss over split index ' .. split_index)
loader:reset_batch_pointer(split_index) -- move batch iteration pointer for this split to front
local loss = 0
-- TODO: dirty hack. will work as long as there are less then 1e6 batches in a file
function get_batch_id()
return loader.file_idx[split_index] * 1e6 + loader.batch_idx[split_index]
end
-- iterate over batches in the split
local ct = 0
local last_batch_id = -1
while get_batch_id() > last_batch_id do
last_batch_id = get_batch_id()
-- fetch a batch
local x, y = loader:next_batch(split_index)
if opt.gpuid >= 0 then -- ship the input arrays to GPU
-- have to convert to float because integers can't be cuda()'d
x = x:float():cuda()
y = y:float():cuda()
end
-- forward pass
local batch_size = x:size(1)
local num_steps = x:size(2) - opt.window_len + 1
local partial_loss = 0
for first_sample = 1, num_steps do
local last_sample = first_sample + opt.window_len - 1
local x_mini = x:sub(1, batch_size, first_sample, last_sample)
local y_mini = y[{{}, last_sample, {}}]
partial_loss = partial_loss + criterion:forward(cnn:forward(x_mini), y_mini)
end
loss = loss + (partial_loss / num_steps)
ct = ct + 1
if ct % 10 == 0 then
print('Evaluated: ' .. ct .. ' batches')
end
end
loss = loss / ct
return loss
end
local params, grad_params = cnn:getParameters()
params:uniform(-0.08, 0.08)
local feval = function(x)
if x ~= params then
params:copy(x)
end
grad_params:zero()
local x, y = loader:next_batch(1)
if opt.gpuid >= 0 then -- ship the input arrays to GPU
-- have to convert to float because integers can't be cuda()'d
x = x:float():cuda()
y = y:float():cuda()
end
local loss = 0
local batch_size = x:size(1)
local num_steps = x:size(2) - opt.window_len + 1
local has_printed = false
for first_sample = 1, num_steps do
local last_sample = first_sample + opt.window_len - 1
local x_mini = x:sub(1, batch_size, first_sample, last_sample)
local y_mini = y[{{}, last_sample, {}}]
-- print(cnn:forward(x_mini):size())
local partial_loss = criterion:forward(cnn:forward(x_mini), y_mini)
loss = loss + partial_loss
cnn:backward(x_mini, criterion:backward(cnn.output, y_mini))
if not has_printed and first_sample % 10 == 0 and y_mini[1]:sum() > 0 then
str = '\n'
for i = 1, 6 do
str = str .. string.format('%.2f ', cnn.output[1][1][i])
end
str = str .. '\n'
for i = 1, 6 do
str = str .. string.format('%.2f ', y_mini[1][i])
end
print(str)
has_printed = true
end
end
grad_params:div(num_steps)
grad_params:clamp(-5, 5)
loss = loss / num_steps
return loss, grad_params
end
function calculate_avg_loss(losses)
local smoothing = 40
local sum = 0
for i = #losses, math.max(1, #losses - smoothing + 1), -1 do
sum = sum + losses[i]
end
return sum / math.min(smoothing, #losses)
end
-- start optimization here
train_losses = train_losses or {}
train_losses_avg = train_losses_avg or {}
val_losses = val_losses or {}
local optim_fun, optim_state
if opt.optim_algo == 'rmsprop' then
optim_fun = optim.rmsprop
optim_state = {learningRate = opt.learning_rate, alpha = opt.decay_rate}
elseif opt.optim_algo == 'adadelta' then
optim_fun = optim.adadelta
optim_state = {rho = 0.95, eps = 1e-7}
end
local iterations = opt.max_epochs * loader.total_samples
local loss0 = nil
for i = start_iter, iterations do
local epoch = i / loader.total_samples
local timer = torch.Timer()
local _, loss = optim_fun(feval, params, optim_state)
local time = timer:time().real
local train_loss = loss[1] -- the loss is inside a list, pop it
train_losses[i] = train_loss
train_losses_avg[i] = calculate_avg_loss(train_losses)
if i % opt.print_every == 0 then
local grad_norm = grad_params:norm()
local param_norm = params:norm()
print(string.format("%d/%d (epoch %.3f), train_loss = %6.8f, grad/param norm = %6.4e, param norm = %.2e time/batch = %.2fs",
i, iterations, epoch, train_loss, grad_norm / param_norm, param_norm, time))
local ct = 0;
local xAxis = torch.Tensor(#train_losses_avg):apply(function() ct = ct + 1; return ct; end)
gnuplot.plot(xAxis, torch.Tensor(train_losses_avg))
end
-- exponential learning rate decay
if i % loader.total_samples == 0 and opt.learning_rate_decay < 1 then
if epoch >= opt.learning_rate_decay_after then
local decay_factor = opt.learning_rate_decay
optim_state.learningRate = optim_state.learningRate * decay_factor -- decay it
print('decayed learning rate by a factor ' .. decay_factor .. ' to ' .. optim_state.learningRate)
end
end
-- every now and then or on last iteration
if i % opt.eval_val_every == 0 or i == iterations then
-- evaluate loss on validation data
local val_loss = eval_split(2) -- 2 = validation
val_losses[i] = val_loss
local savefile = string.format('%s/lm_%s_epoch%.4f_%.2f.t7', opt.checkpoint_dir, opt.savefile, val_loss, epoch)
printGreen('saving checkpoint to ' .. savefile)
local checkpoint = {}
checkpoint.cnn = cnn
checkpoint.criterion = criterion
checkpoint.type = "cnn"
checkpoint.opt = opt
checkpoint.train_losses = train_losses
checkpoint.val_loss = val_loss
checkpoint.val_losses = val_losses
checkpoint.i = i
checkpoint.epoch = epoch
checkpoint.loader = {}
checkpoint.loader.file_idx = loader.file_idx
checkpoint.loader.batch_idx = loader.batch_idx
checkpoint.id = MODEL_ID
torch.save(savefile, checkpoint)
end
if i % 10 == 0 then collectgarbage() end
-- handle early stopping if things are going really bad
if loss[1] ~= loss[1] then
print('loss is NaN. This usually indicates a bug. Please check the issues page for existing issues, or create a new issue, if none exist. Ideally, please state: your operating system, 32-bit/64-bit, your blas version, cpu/cuda/cl?')
break -- halt
end
if loss0 == nil then loss0 = loss[1] end
if loss[1] > loss0 * 3 then
print('loss is exploding, aborting.')
break -- halt
end
end
print 'TRAINING DONE'