The part of the code that needs help is indicated with ###, the code below is appending a dict called results into history = [] and returning history.
def epoch_end(result,epoch):
print("epoch: {}, last_lr {}, Epoch_loss:{}, Epoch_accuracy {}, train_loss {}" .format(epoch, result["lrs"][-1], result["Loss"], result["Accuracy"], result["train loss"] ))
def fit(epochs,train_set,val_dl,model,lr): #model trainer
optimizer = torch.optim.Adam(model.parameters(), lr, weight_decay = 0.001) #defining the optimizer
scheduler = torch.optim.lr_scheduler.OneCycleLR(optimizer, lr*2, epochs = epochs,steps_per_epoch = len(train_set))# learning rate scheduler
def get_lr():
for param_group in optimizer.param_groups: #getting the learning rates of e
return param_group["lr"]
history = []
for epoch in range(epochs):
model.train()
train_loss = []
lrs = []
for batch in train_set:
loss = training_step(batch, model)
train_loss.append(loss)
loss.backward()
nn.utils.clip_grad_norm_(model.parameters(), 0.1)
optimizer.step()
optimizer.zero_grad()
lrs.append(get_lr())
scheduler.step()
#validation
results = validation_combine_loss(val_dl,model)
results["lrs"] = lrs
results["train loss"] = torch.stack(train_loss).mean().item()
epoch_end(results,epoch) ###
history.append(results) ###
return history
fit(20,train_set,val_set,model,0.01) #training model
The output will be
epoch: 0, last_lr 0.0010401048569784015, Epoch_loss:1.3724077939987183, Epoch_accuracy 0.5091797113418579, train_loss 1.513434886932373
epoch: 1, last_lr 0.002795876957832664, Epoch_loss:1.785091757774353, Epoch_accuracy 0.3656249940395355, train_loss 1.467275619506836
epoch: 2, last_lr 0.0051964283175401065, Epoch_loss:2.189635992050171, Epoch_accuracy 0.3041015565395355, train_loss 1.5236047506332397
epoch: 3, last_lr 0.0075979375927655805, Epoch_loss:1.843380331993103, Epoch_accuracy 0.36347657442092896, train_loss 1.593713641166687
epoch: 4, last_lr 0.009356326529838265, Epoch_loss:3.570767641067505, Epoch_accuracy 0.150390625, train_loss 1.6375536918640137
epoch: 5, last_lr 0.01, Epoch_loss:2.7340424060821533, Epoch_accuracy 0.142578125, train_loss 1.6638548374176025
epoch: 6, last_lr 0.009874640062350875, Epoch_loss:1.8366981744766235, Epoch_accuracy 0.31230467557907104, train_loss 1.6672502756118774
epoch: 7, last_lr 0.009504846320134737, Epoch_loss:2.4218339920043945, Epoch_accuracy 0.21992187201976776, train_loss 1.6705607175827026
epoch: 8, last_lr 0.0089091617757105, Epoch_loss:1.7438850402832031, Epoch_accuracy 0.3316406309604645, train_loss 1.6584593057632446
However when I try to append epoch_end(results, epoch) to history instead of appending the dict results to history the outputs are completely different.
for epoch in range(epochs):
model.train()
train_loss = []
lrs = []
for batch in train_set:
loss = training_step(batch, model)
train_loss.append(loss)
loss.backward()
nn.utils.clip_grad_norm_(model.parameters(), 0.1)
optimizer.step()
optimizer.zero_grad()
lrs.append(get_lr())
scheduler.step()
#validation
results = validation_combine_loss(val_dl,model)
results["lrs"] = lrs
results["train loss"] = torch.stack(train_loss).mean().item()
epoch_end(results,epoch) ###
string_formatting = epoch_end(results,epoch) ###
history.append(string_formatting) ###
return history
The output will be, comparing the two outputs, this output seems to have duplicates, is there a reason for this?
epoch: 0, last_lr 0.002080209713956803, Epoch_loss:1.5075801610946655, Epoch_accuracy 0.43378907442092896, train_loss 1.4565353393554688
epoch: 0, last_lr 0.002080209713956803, Epoch_loss:1.5075801610946655, Epoch_accuracy 0.43378907442092896, train_loss 1.4565353393554688
epoch: 1, last_lr 0.005591753915665328, Epoch_loss:8.766005516052246, Epoch_accuracy 0.09824218600988388, train_loss 1.5436667203903198
epoch: 1, last_lr 0.005591753915665328, Epoch_loss:8.766005516052246, Epoch_accuracy 0.09824218600988388, train_loss 1.5436667203903198
epoch: 2, last_lr 0.010392856635080213, Epoch_loss:15.781758308410645, Epoch_accuracy 0.12871094048023224, train_loss 1.6487261056900024
epoch: 2, last_lr 0.010392856635080213, Epoch_loss:15.781758308410645, Epoch_accuracy 0.12871094048023224, train_loss 1.6487261056900024
epoch: 3, last_lr 0.015195875185531161, Epoch_loss:4.121547698974609, Epoch_accuracy 0.15214844048023224, train_loss 1.7063096761703491
epoch: 3, last_lr 0.015195875185531161, Epoch_loss:4.121547698974609, Epoch_accuracy 0.15214844048023224, train_loss 1.7063096761703491
epoch: 4, last_lr 0.01871265305967653, Epoch_loss:2.113895893096924, Epoch_accuracy 0.21621093153953552, train_loss 1.776288390159607
epoch: 4, last_lr 0.01871265305967653, Epoch_loss:2.113895893096924, Epoch_accuracy 0.21621093153953552, train_loss 1.776288390159607
