Granite#

IBM Granite is IBM’s family of enterprise-focused language models. Granite 3.x models are trained on a mix of code and language data and are optimized for enterprise tasks including summarization, classification, and RAG. PowerLM (IBM Research) also uses this architecture.

Task

Text Generation

Architecture

GraniteForCausalLM

Parameters

2B – 8B

HF Org

ibm-granite

Available Models#

  • Granite 3.3 2B Instruct

  • Granite 3.1 8B Instruct

  • Granite 3.0 2B Base

  • PowerLM-3B (IBM Research)

Architecture#

  • GraniteForCausalLM

Example HF Models#

Model

HF ID

Granite 3.0 2B Base

ibm-granite/granite-3.0-2b-base

Granite 3.1 8B Instruct

ibm-granite/granite-3.1-8b-instruct

PowerLM 3B

ibm/PowerLM-3b

Example Recipes#

Recipe

Description

granite_3_3_2b_instruct_squad.yaml

SFT — Granite 3.3 2B Instruct on SQuAD

granite_3_3_2b_instruct_squad_peft.yaml

LoRA — Granite 3.3 2B Instruct on SQuAD

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/granite/granite_3_3_2b_instruct_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/granite/granite_3_3_2b_instruct_squad.yaml

See the Installation Guide and LLM Fine-Tuning Guide.

Fine-Tuning#

See the LLM Fine-Tuning Guide.

Hugging Face Model Cards#