mixtral

mixtral

Mixtral on vLLM (vllm.model_executor.models.mixtral).

Llama's names and Llama's fused block, except that the block's feed-forward is block_sparse_moe, a mixture of experts, aliased to mlp; its output is what the block adds. The routing is inside vLLM's fused MoE kernel and has no values here.

RENAME module-attribute

RENAME = {'model.embed_tokens': 'embed_tokens', 'model.layers': 'layers', 'model.norm': 'norm', 'block_sparse_moe': 'mlp'}

ENVOYS module-attribute

ENVOYS = {MixtralDecoderLayer: Layer, MixtralAttention: Attention, MixtralMoE: Mlp}

Layer

Layer(*args: Any, **kwargs: Any)

Bases: FusedLayer

The decoder block: forward(positions, hidden_states, residual) -> (hidden_states, residual).