Caten Documentation

  • Home
  • Quickstart
  • Development
  • API Reference
    • caten/air
    • caten/aasm
    • caten/codegen
    • caten/api
      • Overview
      • Tensor
      • Func
      • Module
      • Model
      • Initializers
      • ShapeTracker
      • Facet API
      • StateDict
    • caten/nn
      • Activation
      • Convolution
      • Criterion
      • Embedding
      • Linear
      • Normalization
      • Padding
      • Pooling
      • Encoding
      • Optimizers
  • Ready to use packages
    • Overview
    • caten/apps.gpt2
  • External Packages
    • caten/gguf
    • caten/oonx
    • caten/llm
In this article
  • Normalizations
    • [class] BatchNorm
    • [function] !batch-norm
    • [class] LayerNorm
    • [function] !layer-norm
    • [class] RMSNorm
    • [function] !rms-norm

Normalization

  1. Caten Documentation
  2. API Reference
  3. caten/nn
  4. Normalization
|
  • Share via

  •  Edit this article

Normalizations

[class] BatchNorm

(BatchNorm dims &key (eps 1e-5) (affine t) (bias t))
Batch Normalization Layer

[function] !batch-norm

(!batch-norm x &optional weight bias mean invstd (axis 1) (eps 1e-5))
Computes the batch normalization of a tensor x with optional weight and bias tensors.

[class] LayerNorm

(LayerNorm dims &key (eps 1e-5) (affine t) (bias t))
Layer Normalization Layer

[function] !layer-norm

(!layer-norm x normalized-shape &key (weight) (bias) (eps 1e-5))
Computes the layer normalization of a tensor x with optional weight and bias tensors.

[class] RMSNorm

(RMSNorm dims &key (eps 1e-5) (weight t))
Root Mean Square Normalization Layer

[function] !rms-norm

(!rms-norm x normalized-shape &key (weight) (eps 1e-5))
Computes the root mean square normalization of a tensor x with optional weight tensor.

Search
Enter a keyword to search.