[Python] Are JAX operations already vectorized?

Discussão em 'Python' iniciado por Stack, Julho 28, 2025.

  1. Stack

    Stack Membro Participativo

    In the documentation, JAX provides vectorization. However, aren't JAX operations already vectorized? For example, to add two vectors, I thought that the element-wise additions were vectorized internally already.

    My guess is that vectorization is useful when: it's hard for us to add a dimension for broadcasting, so we resort to a more explicit vectorization.

    EDIT: for example, instead of vectorizing convolution2d with different kernels, I simply stack the kernels, copy and stack the channel, then perform the convolution2d with this stack of kernels.

    Continue reading...

Compartilhe esta Página