qwen2_moe

Model type

qwen2_moe

Class

Qwen2MoECausalLMModel

Task

text-generation

Source

models/moe.py

Description

Qwen2-MoE causal language model.

Differs from the generic MoECausalLMModel by adding a per-layer shared expert (always active) and a sigmoid gate that scales its contribution, matching Qwen2MoeForCausalLM.

Usage

mobius build --model <MODEL_ID> output_dir/
from mobius import build

model = build("<MODEL_ID>")