def log(model, batch_idx):
if batch_idx % params["log_interval"] == 0:
entropy = model.entropy()
print("logging: {} learning iterations, entropy: {} / {}".format(
model.getLearningIterations(), float(entropy), model.maxEntropy()))
if params["create_plots"]:
plotDutyCycles(model.dutyCycle,
self.resultsDir + "/figure_" + str(epoch) + "_" +
str(model.getLearningIterations()))
def log(model, batch_idx):
if batch_idx % params["log_interval"] == 0:
entropy = model.entropy()
print("logging: {} learning iterations, entropy: {} / {}".format(
model.getLearningIterations(), float(entropy), model.maxEntropy()))
if params["create_plots"]:
plotDutyCycles(model.dutyCycle,
self.resultsDir + "/figure_" + str(epoch) + "_" +
str(model.getLearningIterations()))
def train(self, params, epoch, repetition):
"""
Train one epoch of this model by iterating through mini batches. An epoch
ends after one pass through the training set, or if the number of mini
batches exceeds the parameter "batches_in_epoch".
"""
# Check for pre-trained model
modelCheckpoint = os.path.join(
params["path"], params["name"],
"model_{}_{}.pt".format(repetition, epoch))
if os.path.exists(modelCheckpoint):
self.model = torch.load(modelCheckpoint, map_location=self.device)
return
self.model.train()
for batch_idx, (batch, target) in enumerate(self.train_loader):
data = batch["input"]
if params["model_type"] in ["resnet9", "cnn"]:
data = torch.unsqueeze(data, 1)
data, target = data.to(self.device), target.to(self.device)
self.optimizer.zero_grad()
output = self.model(data)
loss = F.nll_loss(output, target)
loss.backward()
self.optimizer.step()
# Log info every log_interval mini batches
if batch_idx % params["log_interval"] == 0:
entropy = self.model.entropy()
print("logging: ", self.model.getLearningIterations(),
" learning iterations, elapsedTime",
time.time() - self.startTime, " entropy:",
float(entropy), " / ", self.model.maxEntropy(), "loss:",
loss.item())
if params["create_plots"]:
plotDutyCycles(
self.model.dutyCycle,
self.resultsDir + "/figure_" + str(epoch) + "_" +
str(self.model.getLearningIterations()))
if batch_idx >= params["batches_in_epoch"]:
break
self.model.postEpoch()
# Save model checkpoint on every epoch
if params.get("save_every_epoch", False):
torch.save(self.model, modelCheckpoint)
def train(self, params, epoch):
"""
Train one epoch of this model by iterating through mini batches. An epoch
ends after one pass through the training set, or if the number of mini
batches exceeds the parameter "batches_in_epoch".
"""
self.model.train()
for batch_idx, batch in enumerate(self.train_loader):
data = batch["input"]
if params["model_type"] in ["resnet9", "cnn"]:
data = torch.unsqueeze(data, 1)
target = batch["target"]
data, target = data.to(self.device), target.to(self.device)
self.optimizer.zero_grad()
output = self.model(data)
loss = F.nll_loss(output, target)
loss.backward()
self.optimizer.step()
# Log info every log_interval mini batches
if batch_idx % params["log_interval"] == 0:
entropy = self.model.entropy()
print(
"logging: ",self.model.getLearningIterations(),
" learning iterations, elapsedTime", time.time() - self.startTime,
" entropy:", float(entropy)," / ", self.model.maxEntropy(),
"loss:", loss.item())
if params["create_plots"]:
plotDutyCycles(self.model.dutyCycle,
self.resultsDir + "/figure_"+str(epoch)+"_"+str(
self.model.getLearningIterations()))
if batch_idx >= params["batches_in_epoch"]:
break
self.model.postEpoch()