Class LlamaLayer

Index

Constructors

Properties

attn inputNorm mlp postAttnNorm

Methods

forward forwardTraining

Constructors

constructor

new LlamaLayer(
    weights: Weights,
    config: ModelConfig,
    prefix: string,
    Mlp?: MlpBlock,
): LlamaLayer
Parameters
- weights: Weights
- config: ModelConfig
- prefix: string
- Mlp: MlpBlock = LlamaMLP
Returns LlamaLayer
- Defined in packages/inference/src/models/minicpm5/model.ts:219

Properties

`Readonly`attn

attn: LlamaAttention

`Readonly`inputNorm

inputNorm: RMSNorm

`Readonly`mlp

mlp: MiniCpmMlp

`Readonly`postAttnNorm

postAttnNorm: RMSNorm

Methods

forward

forward(x: MlxArray, cache: Cache, phase: ReadPhase): MlxArray
Parameters
- x: MlxArray
- cache: Cache
- phase: ReadPhase
Returns MlxArray
- Defined in packages/inference/src/models/minicpm5/model.ts:226

forwardTraining

forwardTraining(x: MlxArray, cache: TrainingAttentionCache): MlxArray
The training forward (LlamaAttention.forwardTraining).
Parameters
- x: MlxArray
- cache: TrainingAttentionCache
Returns MlxArray
- Defined in packages/inference/src/models/minicpm5/model.ts:234