Expand description
Core types and parameter structures for embedded-nn.
Structs§
- Activation
- Quantized activation clamping range (min, max).
- Context
- Context for scratchpad buffers (optional).
- Conv
Params - Parameters for Convolution layer operations.
- Dims
- Dimensions for 4D Tensors (Batch, Height, Width, Channels / Output Channels).
- DwConv
Params - Parameters for Depthwise Convolution layer operations.
- FcParams
- Parameters for Fully Connected (Linear) layer operations.
- PerChannel
Quant Params - Per-channel quantization parameters.
- PerTensor
Quant Params - Per-tensor quantization parameters.
- Pool
Params - Parameters for Pooling operations.
- Softmax
Params - Parameters for Softmax operation.
- Tile
- Tile or kernel spatial dimensions (Width, Height).
Enums§
- Error
- Error types for
embedded-nnoperations. - Quant
Params - Unified quantization parameters (either per-tensor or per-channel).
Type Aliases§
- Result
- Result type alias for
embedded-nn.