docs.rs failed to build g-nn-0.1.0
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Neural primitives and a small AdamW/SGD.
Building blocks for training: linear layers, activations, softmax and
losses, embeddings, and two optimizers. All differentiable functions build
on [g_ad], so their results carry reverse-mode autodiff nodes when their
inputs require gradients.
Optimizers
[Sgd] and [AdamW] both return updated parameter tensors rather than
mutating them, matching the library's functional style. Feed the returned
tensors back as the parameters of the next step.