wehub-resource-sync
73 lines
2.4 KiB
Python
73 lines
2.4 KiB
Python
import torch
|
|
import torch.nn.functional as F
|
|
import numpy as np
|
|
from itertools import product
|
|
|
|
|
|
def compute_accuracy(model, data_loader, device):
|
|
model.eval()
|
|
with torch.no_grad():
|
|
correct_pred, num_examples = 0, 0
|
|
for i, (features, targets) in enumerate(data_loader):
|
|
|
|
features = features.to(device)
|
|
targets = targets.to(device)
|
|
|
|
logits = model(features)
|
|
if isinstance(logits, torch.distributed.rpc.api.RRef):
|
|
logits = logits.local_value()
|
|
_, predicted_labels = torch.max(logits, 1)
|
|
num_examples += targets.size(0)
|
|
correct_pred += (predicted_labels == targets).sum()
|
|
return correct_pred.float()/num_examples * 100
|
|
|
|
|
|
def compute_epoch_loss(model, data_loader, device):
|
|
model.eval()
|
|
curr_loss, num_examples = 0., 0
|
|
with torch.no_grad():
|
|
for features, targets in data_loader:
|
|
features = features.to(device)
|
|
targets = targets.to(device)
|
|
logits = model(features)
|
|
if isinstance(logits, torch.distributed.rpc.api.RRef):
|
|
logits = logits.local_value()
|
|
loss = F.cross_entropy(logits, targets, reduction='sum')
|
|
num_examples += targets.size(0)
|
|
curr_loss += loss
|
|
|
|
curr_loss = curr_loss / num_examples
|
|
return curr_loss
|
|
|
|
|
|
def compute_confusion_matrix(model, data_loader, device):
|
|
|
|
all_targets, all_predictions = [], []
|
|
with torch.no_grad():
|
|
|
|
for i, (features, targets) in enumerate(data_loader):
|
|
|
|
features = features.to(device)
|
|
targets = targets
|
|
logits = model(features)
|
|
_, predicted_labels = torch.max(logits, 1)
|
|
all_targets.extend(targets.to('cpu'))
|
|
all_predictions.extend(predicted_labels.to('cpu'))
|
|
|
|
all_predictions = all_predictions
|
|
all_predictions = np.array(all_predictions)
|
|
all_targets = np.array(all_targets)
|
|
|
|
class_labels = np.unique(np.concatenate((all_targets, all_predictions)))
|
|
if class_labels.shape[0] == 1:
|
|
if class_labels[0] != 0:
|
|
class_labels = np.array([0, class_labels[0]])
|
|
else:
|
|
class_labels = np.array([class_labels[0], 1])
|
|
n_labels = class_labels.shape[0]
|
|
lst = []
|
|
z = list(zip(all_targets, all_predictions))
|
|
for combi in product(class_labels, repeat=2):
|
|
lst.append(z.count(combi))
|
|
mat = np.asarray(lst)[:, None].reshape(n_labels, n_labels)
|
|
return mat |