PyTorch torch.nn.TransformerEncoder Funktion
PyTorch torch.nn Referenzhandbuch
torch.nn.TransformerEncoderEs ist das Transformer-Encoder-Modul in PyTorch.
Es besteht aus mehreren TransformerEncoderLayer-Schichten und wird zur Verarbeitung von Eingabesequenzen verwendet.
Funktionsdefinition
torch.nn.TransformerEncoder(encoder_layer, num_layers, norm=None)
Parameters
encoder_layer: Encoder-Schichtnum_layers: Anzahl der Schichten
Anwendungsbeispiele
Beispiel 1: Grundlegende Verwendung
Beispiel
import torch
import torch.nn as nn
# Einzelne Encoder-Schicht
encoder_layer = nn.TransformerEncoderLayer(d_model=512, nhead=8, batch_first=True)
# Vollständiger Encoder
transformer_encoder = nn.TransformerEncoder(encoder_layer, num_layers=6)
# Eingabe
x = torch.randn(32, 100, 512) # batch, seq, d_model
output = transformer_encoder(x)
print("Eingabeform:", x.shape)
print("Ausgabeform:", output.shape)
import torch.nn as nn
# Einzelne Encoder-Schicht
encoder_layer = nn.TransformerEncoderLayer(d_model=512, nhead=8, batch_first=True)
# Vollständiger Encoder
transformer_encoder = nn.TransformerEncoder(encoder_layer, num_layers=6)
# Eingabe
x = torch.randn(32, 100, 512) # batch, seq, d_model
output = transformer_encoder(x)
print("Eingabeform:", x.shape)
print("Ausgabeform:", output.shape)
Beispiel 2: Textklassifikation
Beispiel
import torch
import torch.nn as nn
class TransformerClassifier(nn.Module):
def __init__(self, vocab_size, d_model=256, nhead=4, num_layers=4):
super(TransformerClassifier, self).__init__()
self.embedding = nn.Embedding(vocab_size, d_model)
self.pos_encoder = nn.Parameter(torch.randn(1, 512, d_model) * 0.1)
encoder_layer = nn.TransformerEncoderLayer(d_model, nhead, batch_first=True)
self.encoder = nn.TransformerEncoder(encoder_layer, num_layers)
self.fc = nn.Linear(d_model, 10)
def forward(self, x):
# Einfache Positionskodierung
x = self.embedding(x) + self.pos_encoder[:, :x.size(1), :]
x = self.encoder(x)
# Nimm das erste Token
return self.fc(x[:, 0, :])
model = TransformerClassifier(10000)
x = torch.randint(0, 10000, (32, 100))
output = model(x)
print("Ausgabeform:", output.shape)
import torch.nn as nn
class TransformerClassifier(nn.Module):
def __init__(self, vocab_size, d_model=256, nhead=4, num_layers=4):
super(TransformerClassifier, self).__init__()
self.embedding = nn.Embedding(vocab_size, d_model)
self.pos_encoder = nn.Parameter(torch.randn(1, 512, d_model) * 0.1)
encoder_layer = nn.TransformerEncoderLayer(d_model, nhead, batch_first=True)
self.encoder = nn.TransformerEncoder(encoder_layer, num_layers)
self.fc = nn.Linear(d_model, 10)
def forward(self, x):
# Einfache Positionskodierung
x = self.embedding(x) + self.pos_encoder[:, :x.size(1), :]
x = self.encoder(x)
# Nimm das erste Token
return self.fc(x[:, 0, :])
model = TransformerClassifier(10000)
x = torch.randint(0, 10000, (32, 100))
output = model(x)
print("Ausgabeform:", output.shape)
Anwendungsfälle
- BERT: Encoder-Grundlagen
- Text-Klassifizierung
- Merkmalsextraktion
Andere Erweiterungen