Implementation:Ollama Ollama Llama Model BailingMoE

Knowledge Sources	Ollama
Domains	LLM Inference, Model Architecture
Last Updated	2025-02-15 00:00 GMT

Overview

Implements the ggml computation graph builder for the BailingMoE model architecture.

Description

The llm_build_bailingmoe constructor builds a transformer with RoPE-based positional encoding (with rope factor support), RMS-normalized self-attention with Q/K/V projections including optional bias terms, and Mixture-of-Experts feed-forward layers across all transformer blocks. Uses build_moe_ffn for the expert routing and gating.

Usage

Enables Ollama to run BailingMoE models through the llama.cpp inference engine.

Code Reference

Source Location

• Repository: Ollama
• File: llama/llama.cpp/src/models/bailingmoe.cpp
• Lines: 1-144

Signature

llm_build_bailingmoe::llm_build_bailingmoe(
    const llama_model & model,
    const llm_graph_params & params) : llm_graph_context(params);

Import

#include "models.h"

I/O Contract

Inputs

Name	Type	Required	Description
model	const llama_model &	Yes	Loaded model with BailingMoE weights
params	const llm_graph_params &	Yes	Graph construction parameters

Outputs

Name	Type	Description
ggml graph	ggml_cgraph	Complete BailingMoE computation graph

Usage Examples

auto builder = llm_build_bailingmoe(model, params);

Related Pages

• Principle:Ollama_Ollama_LLM_Inference_Pipeline