code stringlengths 3 6.57k |
|---|
self.mlp1(x) |
x.sparse() |
self.conv1(y) |
self.pool(y) |
self.conv2(y1) |
self.pool(y) |
self.conv3(y2) |
self.pool(y) |
self.conv4(y3) |
self.pool(y) |
y1.slice(x) |
y2.slice(x) |
y3.slice(x) |
y4.slice(x) |
ME.cat(x1, x2, x3, x4) |
self.conv5(x.sparse() |
self.global_max_pool(y) |
self.global_avg_pool(y) |
self.final(ME.cat(x1, x2) |
GlobalMaxAvgPool(torch.nn.Module) |
__init__(self) |
torch.nn.Module.__init__(self) |
ME.MinkowskiGlobalMaxPooling() |
ME.MinkowskiGlobalAvgPooling() |
forward(self, tensor) |
self.global_max_pool(tensor) |
self.global_avg_pool(tensor) |
ME.cat(x, y) |
MinkowskiSplatFCNN(MinkowskiFCNN) |
forward(self, x: ME.TensorField) |
self.mlp1(x) |
x.splat() |
self.conv1(y) |
self.pool(y) |
self.conv2(y1) |
self.pool(y) |
self.conv3(y2) |
self.pool(y) |
self.conv4(y3) |
self.pool(y) |
y1.interpolate(x) |
y2.interpolate(x) |
y3.interpolate(x) |
y4.interpolate(x) |
ME.cat(x1, x2, x3, x4) |
self.conv5(x.sparse() |
self.global_max_pool(y) |
self.global_avg_pool(y) |
self.final(ME.cat(x1, x2) |
create_input_batch(batch, is_minknet, device="cuda", quantization_size=0.05) |
permute(0, 2, 1) |
to(device) |
__init__(self, translation) |
__call__(self, coords) |
np.random.uniform(low=-self.trans, high=self.trans, size=[1, 3]) |
make_data_loader(phase, is_minknet, config) |
CoordinateTransformation(trans=config.translation) |
CoordinateTranslation(config.test_translation) |
test(net, device, config, phase="val") |
isinstance(net, ME.MinkowskiNetwork) |
net.eval() |
torch.no_grad() |
net(input) |
torch.argmax(logit, 1) |
labels.append(batch["labels"].cpu() |
numpy() |
preds.append(pred.cpu() |
numpy() |
torch.cuda.empty_cache() |
metrics.accuracy_score(np.concatenate(labels) |
np.concatenate(preds) |
criterion(pred, labels, smoothing=True) |
labels.contiguous() |
view(-1) |
pred.size(1) |
torch.zeros_like(pred) |
scatter(1, labels.view(-1, 1) |
F.log_softmax(pred, dim=1) |
sum(dim=1) |
mean() |
F.cross_entropy(pred, labels, reduction="mean") |
train(net, device, config) |
isinstance(net, ME.MinkowskiNetwork) |
net.parameters() |
print(optimizer) |
print(scheduler) |
iter(make_data_loader("train", is_minknet, config) |
net.train() |
range(config.max_steps) |
optimizer.zero_grad() |
train_iter.next() |
iter(make_data_loader("train", is_minknet, config) |
train_iter.next() |
net(input) |
criterion(logit, data_dict["labels"].to(device) |
loss.backward() |
optimizer.step() |
scheduler.step() |
torch.cuda.empty_cache() |
print(f"Iter: {i}, Loss: {loss.item() |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.