X-Git-Url: https://fleuret.org/cgi-bin/gitweb/gitweb.cgi?p=pysvrt.git;a=blobdiff_plain;f=cnn-svrt.py;h=63b11ee4c33831bd0ec8236f7892554bcab0b47a;hp=96fb498740625c551a3524df7e79ef68e35544ee;hb=4c7ff07760d015a2efad8b7eb0bd44dd9acc9106;hpb=231c2b2d912d7480af0ce9512b12a909a4fe2a3d

diff --git a/cnn-svrt.py b/cnn-svrt.py
index 96fb498..63b11ee 100755
--- a/cnn-svrt.py
+++ b/cnn-svrt.py
@@ -19,13 +19,17 @@
 #  General Public License for more details.
 #
 #  You should have received a copy of the GNU General Public License
-#  along with selector.  If not, see <http://www.gnu.org/licenses/>.
+#  along with svrt.  If not, see <http://www.gnu.org/licenses/>.
 
 import time
 import argparse
+import math
+import distutils.util
 
 from colorama import Fore, Back, Style
 
+# Pytorch
+
 import torch
 
 from torch import optim
@@ -35,55 +39,75 @@ from torch import nn
 from torch.nn import functional as fn
 from torchvision import datasets, transforms, utils
 
-import svrt
+# SVRT
+
+import svrtset
 
 ######################################################################
 
 parser = argparse.ArgumentParser(
-    description = 'Simple convnet test on the SVRT.',
+    description = "Convolutional networks for the SVRT. Written by Francois Fleuret, (C) Idiap research institute.",
     formatter_class = argparse.ArgumentDefaultsHelpFormatter
 )
 
 parser.add_argument('--nb_train_samples',
-                    type = int, default = 100000,
-                    help = 'How many samples for train')
+                    type = int, default = 100000)
 
 parser.add_argument('--nb_test_samples',
-                    type = int, default = 10000,
-                    help = 'How many samples for test')
+                    type = int, default = 10000)
 
 parser.add_argument('--nb_epochs',
-                    type = int, default = 100,
-                    help = 'How many training epochs')
+                    type = int, default = 50)
+
+parser.add_argument('--batch_size',
+                    type = int, default = 100)
 
 parser.add_argument('--log_file',
-                    type = str, default = 'cnn-svrt.log',
-                    help = 'Log file name')
+                    type = str, default = 'default.log')
+
+parser.add_argument('--compress_vignettes',
+                    type = distutils.util.strtobool, default = 'True',
+                    help = 'Use lossless compression to reduce the memory footprint')
+
+parser.add_argument('--deep_model',
+                    type = distutils.util.strtobool, default = 'True',
+                    help = 'Use Afroze\'s Alexnet-like deep model')
+
+parser.add_argument('--test_loaded_models',
+                    type = distutils.util.strtobool, default = 'False',
+                    help = 'Should we compute the test errors of loaded models')
+
+parser.add_argument('--problems',
+                    type = str, default = '1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23',
+                    help = 'What problems to process')
 
 args = parser.parse_args()
 
 ######################################################################
 
 log_file = open(args.log_file, 'w')
+pred_log_t = None
 
 print(Fore.RED + 'Logging into ' + args.log_file + Style.RESET_ALL)
 
-def log_string(s):
-    s = Fore.GREEN + time.ctime() + Style.RESET_ALL + ' ' + s
-    log_file.write(s + '\n')
-    log_file.flush()
-    print(s)
-
-######################################################################
+# Log and prints the string, with a time stamp. Does not log the
+# remark
+def log_string(s, remark = ''):
+    global pred_log_t
 
-def generate_set(p, n):
-    target = torch.LongTensor(n).bernoulli_(0.5)
     t = time.time()
-    input = svrt.generate_vignettes(p, target)
-    t = time.time() - t
-    log_string('DATA_SET_GENERATION {:.02f} sample/s'.format(n / t))
-    input = input.view(input.size(0), 1, input.size(1), input.size(2)).float()
-    return Variable(input), Variable(target)
+
+    if pred_log_t is None:
+        elapsed = 'start'
+    else:
+        elapsed = '+{:.02f}s'.format(t - pred_log_t)
+
+    pred_log_t = t
+
+    log_file.write('[' + time.ctime() + '] ' + elapsed + ' ' + s + '\n')
+    log_file.flush()
+
+    print(Fore.BLUE + '[' + time.ctime() + '] ' + Fore.GREEN + elapsed + Style.RESET_ALL + ' ' + s + Fore.CYAN + remark + Style.RESET_ALL)
 
 ######################################################################
 
@@ -101,14 +125,15 @@ def generate_set(p, n):
 # -- full(120x84)      -> 84         1
 # -- full(84x2)        -> 2          1
 
-class Net(nn.Module):
+class AfrozeShallowNet(nn.Module):
     def __init__(self):
-        super(Net, self).__init__()
+        super(AfrozeShallowNet, self).__init__()
         self.conv1 = nn.Conv2d(1, 6, kernel_size=21)
         self.conv2 = nn.Conv2d(6, 16, kernel_size=19)
         self.conv3 = nn.Conv2d(16, 120, kernel_size=18)
         self.fc1 = nn.Linear(120, 84)
         self.fc2 = nn.Linear(84, 2)
+        self.name = 'shallownet'
 
     def forward(self, x):
         x = fn.relu(fn.max_pool2d(self.conv1(x), kernel_size=2))
@@ -119,44 +144,94 @@ class Net(nn.Module):
         x = self.fc2(x)
         return x
 
-def train_model(train_input, train_target):
-    model, criterion = Net(), nn.CrossEntropyLoss()
+######################################################################
 
-    nb_parameters = 0
-    for p in model.parameters():
-        nb_parameters += p.numel()
-    log_string('NB_PARAMETERS {:d}'.format(nb_parameters))
+# Afroze's DeepNet
+
+class AfrozeDeepNet(nn.Module):
+    def __init__(self):
+        super(AfrozeDeepNet, self).__init__()
+        self.conv1 = nn.Conv2d(  1,  32, kernel_size=7, stride=4, padding=3)
+        self.conv2 = nn.Conv2d( 32,  96, kernel_size=5, padding=2)
+        self.conv3 = nn.Conv2d( 96, 128, kernel_size=3, padding=1)
+        self.conv4 = nn.Conv2d(128, 128, kernel_size=3, padding=1)
+        self.conv5 = nn.Conv2d(128,  96, kernel_size=3, padding=1)
+        self.fc1 = nn.Linear(1536, 256)
+        self.fc2 = nn.Linear(256, 256)
+        self.fc3 = nn.Linear(256, 2)
+        self.name = 'deepnet'
+
+    def forward(self, x):
+        x = self.conv1(x)
+        x = fn.max_pool2d(x, kernel_size=2)
+        x = fn.relu(x)
+
+        x = self.conv2(x)
+        x = fn.max_pool2d(x, kernel_size=2)
+        x = fn.relu(x)
+
+        x = self.conv3(x)
+        x = fn.relu(x)
+
+        x = self.conv4(x)
+        x = fn.relu(x)
+
+        x = self.conv5(x)
+        x = fn.max_pool2d(x, kernel_size=2)
+        x = fn.relu(x)
+
+        x = x.view(-1, 1536)
+
+        x = self.fc1(x)
+        x = fn.relu(x)
+
+        x = self.fc2(x)
+        x = fn.relu(x)
+
+        x = self.fc3(x)
+
+        return x
+
+######################################################################
+
+def train_model(model, train_set):
+    batch_size = args.batch_size
+    criterion = nn.CrossEntropyLoss()
 
     if torch.cuda.is_available():
-        model.cuda()
         criterion.cuda()
 
-    optimizer, bs = optim.SGD(model.parameters(), lr = 1e-2), 100
+    optimizer = optim.SGD(model.parameters(), lr = 1e-2)
 
-    for k in range(0, args.nb_epochs):
+    start_t = time.time()
+
+    for e in range(0, args.nb_epochs):
         acc_loss = 0.0
-        for b in range(0, train_input.size(0), bs):
-            output = model.forward(train_input.narrow(0, b, bs))
-            loss = criterion(output, train_target.narrow(0, b, bs))
+        for b in range(0, train_set.nb_batches):
+            input, target = train_set.get_batch(b)
+            output = model.forward(Variable(input))
+            loss = criterion(output, Variable(target))
             acc_loss = acc_loss + loss.data[0]
             model.zero_grad()
             loss.backward()
             optimizer.step()
-        log_string('TRAIN_LOSS {:d} {:f}'.format(k, acc_loss))
+        dt = (time.time() - start_t) / (e + 1)
+        log_string('train_loss {:d} {:f}'.format(e + 1, acc_loss),
+                   ' [ETA ' + time.ctime(time.time() + dt * (args.nb_epochs - e)) + ']')
 
     return model
 
 ######################################################################
 
-def nb_errors(model, data_input, data_target, bs = 100):
+def nb_errors(model, data_set):
     ne = 0
-
-    for b in range(0, data_input.size(0), bs):
-        output = model.forward(data_input.narrow(0, b, bs))
+    for b in range(0, data_set.nb_batches):
+        input, target = data_set.get_batch(b)
+        output = model.forward(Variable(input))
         wta_prediction = output.data.max(1)[1].view(-1)
 
-        for i in range(0, bs):
-            if wta_prediction[i] != data_target.narrow(0, b, bs).data[i]:
+        for i in range(0, data_set.batch_size):
+            if wta_prediction[i] != target[i]:
                 ne = ne + 1
 
     return ne
@@ -164,38 +239,129 @@ def nb_errors(model, data_input, data_target, bs = 100):
 ######################################################################
 
 for arg in vars(args):
-    log_string('ARGUMENT ' + str(arg) + ' ' + str(getattr(args, arg)))
+    log_string('argument ' + str(arg) + ' ' + str(getattr(args, arg)))
 
-for problem_number in range(1, 24):
-    train_input, train_target = generate_set(problem_number, args.nb_train_samples)
-    test_input, test_target = generate_set(problem_number, args.nb_test_samples)
+######################################################################
 
-    if torch.cuda.is_available():
-        train_input, train_target = train_input.cuda(), train_target.cuda()
-        test_input, test_target = test_input.cuda(), test_target.cuda()
+def int_to_suffix(n):
+    if n >= 1000000 and n%1000000 == 0:
+        return str(n//1000000) + 'M'
+    elif n >= 1000 and n%1000 == 0:
+        return str(n//1000) + 'K'
+    else:
+        return str(n)
+
+class vignette_logger():
+    def __init__(self, delay_min = 60):
+        self.start_t = time.time()
+        self.last_t = self.start_t
+        self.delay_min = delay_min
+
+    def __call__(self, n, m):
+        t = time.time()
+        if t > self.last_t + self.delay_min:
+            dt = (t - self.start_t) / m
+            log_string('sample_generation {:d} / {:d}'.format(
+                m,
+                n), ' [ETA ' + time.ctime(time.time() + dt * (n - m)) + ']'
+            )
+            self.last_t = t
+
+######################################################################
+
+if args.nb_train_samples%args.batch_size > 0 or args.nb_test_samples%args.batch_size > 0:
+    print('The number of samples must be a multiple of the batch size.')
+    raise
+
+if args.compress_vignettes:
+    log_string('using_compressed_vignettes')
+    VignetteSet = svrtset.CompressedVignetteSet
+else:
+    log_string('using_uncompressed_vignettes')
+    VignetteSet = svrtset.VignetteSet
+
+for problem_number in map(int, args.problems.split(',')):
+
+    log_string('############### problem ' + str(problem_number) + ' ###############')
+
+    if args.deep_model:
+        model = AfrozeDeepNet()
+    else:
+        model = AfrozeShallowNet()
+
+    if torch.cuda.is_available(): model.cuda()
+
+    model_filename = model.name + '_pb:' + \
+                     str(problem_number) + '_ns:' + \
+                     int_to_suffix(args.nb_train_samples) + '.param'
+
+    nb_parameters = 0
+    for p in model.parameters(): nb_parameters += p.numel()
+    log_string('nb_parameters {:d}'.format(nb_parameters))
+
+    ##################################################
+    # Tries to load the model
+
+    need_to_train = False
+    try:
+        model.load_state_dict(torch.load(model_filename))
+        log_string('loaded_model ' + model_filename)
+    except:
+        need_to_train = True
+
+    ##################################################
+    # Train if necessary
+
+    if need_to_train:
+
+        log_string('training_model ' + model_filename)
+
+        t = time.time()
+
+        train_set = VignetteSet(problem_number,
+                                args.nb_train_samples, args.batch_size,
+                                cuda = torch.cuda.is_available(),
+                                logger = vignette_logger())
+
+        log_string('data_generation {:0.2f} samples / s'.format(
+            train_set.nb_samples / (time.time() - t))
+        )
+
+        train_model(model, train_set)
+        torch.save(model.state_dict(), model_filename)
+        log_string('saved_model ' + model_filename)
+
+        nb_train_errors = nb_errors(model, train_set)
+
+        log_string('train_error {:d} {:.02f}% {:d} {:d}'.format(
+            problem_number,
+            100 * nb_train_errors / train_set.nb_samples,
+            nb_train_errors,
+            train_set.nb_samples)
+        )
+
+    ##################################################
+    # Test if necessary
 
-    mu, std = train_input.data.mean(), train_input.data.std()
-    train_input.data.sub_(mu).div_(std)
-    test_input.data.sub_(mu).div_(std)
+    if need_to_train or args.test_loaded_models:
 
-    model = train_model(train_input, train_target)
+        t = time.time()
 
-    nb_train_errors = nb_errors(model, train_input, train_target)
+        test_set = VignetteSet(problem_number,
+                               args.nb_test_samples, args.batch_size,
+                               cuda = torch.cuda.is_available())
 
-    log_string('TRAIN_ERROR {:d} {:.02f}% {:d} {:d}'.format(
-        problem_number,
-        100 * nb_train_errors / train_input.size(0),
-        nb_train_errors,
-        train_input.size(0))
-    )
+        log_string('data_generation {:0.2f} samples / s'.format(
+            test_set.nb_samples / (time.time() - t))
+        )
 
-    nb_test_errors = nb_errors(model, test_input, test_target)
+        nb_test_errors = nb_errors(model, test_set)
 
-    log_string('TEST_ERROR {:d} {:.02f}% {:d} {:d}'.format(
-        problem_number,
-        100 * nb_test_errors / test_input.size(0),
-        nb_test_errors,
-        test_input.size(0))
-    )
+        log_string('test_error {:d} {:.02f}% {:d} {:d}'.format(
+            problem_number,
+            100 * nb_test_errors / test_set.nb_samples,
+            nb_test_errors,
+            test_set.nb_samples)
+        )
 
 ######################################################################