pytorch
TLDR
Install PyTorch
$ pip install torch torchvision
Check CUDA availability$ python -c "import torch; print(torch.cuda.is_available())"
Check version$ python -c "import torch; print(torch.__version__)"
Run with specific GPU$ CUDA_VISIBLE_DEVICES=[0] python [script.py]
SYNOPSIS
pytorch Python library for deep learning
DESCRIPTION
PyTorch is an open-source deep learning framework with dynamic computational graphs. It provides tensor computation with GPU acceleration and automatic differentiation for neural networks.
BASIC EXAMPLE
$ import torch
import torch.nn as nn
# Create tensor
x = torch.tensor([[1., 2.], [3., 4.]])
# GPU transfer
if torch.cuda.is_available():
x = x.cuda()
# Simple model
model = nn.Sequential(
nn.Linear(10, 64),
nn.ReLU(),
nn.Linear(64, 1)
)
# Training loop
optimizer = torch.optim.Adam(model.parameters())
criterion = nn.MSELoss()
for epoch in range(100):
output = model(input)
loss = criterion(output, target)
optimizer.zero_grad()
loss.backward()
optimizer.step()
import torch.nn as nn
# Create tensor
x = torch.tensor([[1., 2.], [3., 4.]])
# GPU transfer
if torch.cuda.is_available():
x = x.cuda()
# Simple model
model = nn.Sequential(
nn.Linear(10, 64),
nn.ReLU(),
nn.Linear(64, 1)
)
# Training loop
optimizer = torch.optim.Adam(model.parameters())
criterion = nn.MSELoss()
for epoch in range(100):
output = model(input)
loss = criterion(output, target)
optimizer.zero_grad()
loss.backward()
optimizer.step()
ENVIRONMENT
$ # Use specific GPU
CUDA_VISIBLE_DEVICES=0,1 python train.py
# Check GPU memory
python -c "import torch; print(torch.cuda.memory_summary())"
CUDA_VISIBLE_DEVICES=0,1 python train.py
# Check GPU memory
python -c "import torch; print(torch.cuda.memory_summary())"
CAVEATS
GPU version requires CUDA. Large library (~2GB). Install from pytorch.org for CUDA version matching.
HISTORY
PyTorch was developed by Facebook AI Research (FAIR), released in 2016 as a successor to Torch (Lua).
SEE ALSO
tensorflow(1), python(1), nvidia-smi(1)
