1. Pytorch Operations

Numpy + AutoGrad

Numpy

Tensor

• 다차원 array를 표현하는 PyTorch 클래스
• 사실상 numpy 의 ndarray와 동일
• Tensor를 생성하는 함수도 거의 동일

Numpy like operations

n_array= np.arange(10).reshape(2,5)
t_array=torch.FloatTensor(n_array)
print(t_array)
print(t_array.ndim, t_array.shape)

GPU에 tensor 올려서 사용하기

t_data.device   #device(type='cpu')
if torch.cuda.is_available():
    t_data_cuda=t_data.to("cuda")
t_data_cuda.device   #device(type='cuda')

Tensor Handling

• view
• squeeze
• unsqueeze

Tensor operations

→ mostly same as numpy operations

• torch.mm → for matrix operations ( broadcasting X)
• torch.matmul →for matrix operations ( broadcasting O)