What are the most common built-in layers in torch.nn and what do they do?

import torch, torch.nn as nn

# Linear layer internals
fc = nn.Linear(4, 8)
print(fc.weight.shape)   # (8, 4) — note: output × input
print(fc.bias.shape)     # (8,)

# Embedding
emb = nn.Embedding(num_embeddings=10000, embedding_dim=128,
                   padding_idx=0)   # index 0 gets a zero vector
tokens = torch.tensor([1, 42, 7])   # shape (3,)
out = emb(tokens)                   # shape (3, 128)

# BatchNorm vs LayerNorm
bn  = nn.BatchNorm1d(64)   # input (N, 64) — normalises across N
ln  = nn.LayerNorm(64)     # input (N, 64) — normalises across 64 features

x = torch.randn(16, 64)
print(bn(x).shape)   # (16, 64)
print(ln(x).shape)   # (16, 64)

# Dropout only active during training
drop = nn.Dropout(p=0.5)
model = nn.Sequential(nn.Linear(32,32), drop, nn.ReLU())
model.train();  x_tr = model(torch.randn(4,32))  # 50% zeros
model.eval();   x_ev = model(torch.randn(4,32))  # all active