starcoder2

Model type	starcoder2
Class	StarCoder2CausalLMModel
Task	text-generation
Source	models/starcoder2.py

Description

StarCoder2 causal language model.

Differences from the base CausalLMModel:

• Uses LayerNorm (mean-centering + std-normalisation) instead of RMSNorm.

• Uses a non-gated two-layer MLP (c_fc / c_proj in HF naming, mapped to up_proj / down_proj in ONNX).

Replicates HuggingFace’s Starcoder2ForCausalLM.

Usage

mobius build --model <MODEL_ID> output_dir/

from mobius import build

model = build("<MODEL_ID>")