StarCoder2#

StarCoder2 is BigCode’s second-generation code language model, available in 3B, 7B, and 15B sizes, trained on 600+ programming languages from The Stack v2.

Task

Code Generation

Architecture

Starcoder2ForCausalLM

Parameters

3B – 15B

HF Org

bigcode

Available Models#

  • starcoder2-3b: 3B

  • starcoder2-7b: 7B

  • starcoder2-15b: 15B

Architecture#

  • Starcoder2ForCausalLM

Example HF Models#

Model

HF ID

StarCoder2 3B

bigcode/starcoder2-3b

StarCoder2 7B

bigcode/starcoder2-7b

StarCoder2 15B

bigcode/starcoder2-15b

Example Recipes#

Recipe

Description

starcoder_2_7b_squad.yaml

SFT — StarCoder2 7B on SQuAD

starcoder_2_7b_hellaswag_fp8.yaml

SFT — StarCoder2 7B on HellaSwag with FP8

Try with NeMo AutoModel#

1. Install (full instructions):

pip install nemo-automodel

2. Clone the repo to get the example recipes:

git clone https://github.com/NVIDIA-NeMo/Automodel.git
cd Automodel

3. Run the recipe from inside the repo:

automodel --nproc-per-node=8 examples/llm_finetune/starcoder/starcoder_2_7b_squad.yaml
Run with Docker

1. Pull the container and mount a checkpoint directory:

docker run --gpus all -it --rm \
  --shm-size=8g \
  -v $(pwd)/checkpoints:/opt/Automodel/checkpoints \
  nvcr.io/nvidia/nemo-automodel:26.02.00

2. Navigate to the AutoModel directory (where the recipes are):

cd /opt/Automodel

3. Run the recipe:

automodel --nproc-per-node=8 examples/llm_finetune/starcoder/starcoder_2_7b_squad.yaml

See the Installation Guide and LLM Fine-Tuning Guide.

Fine-Tuning#

See the LLM Fine-Tuning Guide.

Hugging Face Model Cards#