Add example for Text2SQL LoRA on Nemotron 3 Nano 30B with NeMo Automodel (#76)

Change adds an end to end example for Text2SQL finetuning of Nemotron 3 Nano 30BA3B
using BIRD SQL dataset with NeMo Automodel.

Signed-off-by: PJ Loury <ploury@nvidia.com>

Author: pjloury

use-case-examples/sql-lora-finetuning-and-deployment/README.md

@@ -1,17 +1,20 @@
-# LoRA Fine-Tuning & Deployment of Nemotron for SQL using [Megatron Bridge](https://github.com/NVIDIA-NeMo/Megatron-Bridge)
+# LoRA Fine-Tuning & Deployment of Nemotron 3 Nano for Text2SQL
 
+End-to-end guide for LoRA fine-tuning **NVIDIA Nemotron-3-Nano-30B** on a text-to-SQL task (BIRD SQL) and deploying with NVIDIA NIM or vLLM. Two notebook options:
 
-End-to-end guide for LoRA fine-tuning **NVIDIA Nemotron-3-Nano-30B** on a text-to-SQL task (BIRD SQL) and deploying with NVIDIA NIM or vLLM.
+- **AutoModel** – Uses the [NeMo AutoModel](https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/containers/nemo-automodel) container for training and checkpointing.
+- **Megatron Bridge** – Uses [Megatron Bridge](https://github.com/NVIDIA-NeMo/Megatron-Bridge) for training.
 
 ## What's Inside
 
 | File | Description |
 |------|-------------|
-| `finetuning_deployment_guide_mbridge.ipynb` | Full walkthrough notebook |
-| `bird_sql/` | Dataset preparation helpers (BIRD SQL → chat-templated JSONL) |a
+| `finetuning_deployment_guide_automodel.ipynb` | Full walkthrough (NeMo AutoModel) |
+| `finetuning_deployment_guide_mbridge.ipynb` | Full walkthrough (Megatron Bridge) |
+| `bird_sql/` | Dataset preparation helpers (BIRD SQL → chat-templated JSONL) |
 
 ## Quick Start
 
-[![Deploy on Brev](https://brev-assets.s3.us-west-1.amazonaws.com/nv-lb-dark.svg)](https://brev.nvidia.com/launchable/deploy?launchableID=env-39PnUMhHmxbMcHKO61iQ8O5F7ZL)
+Open either notebook to get started. A **Deploy on Brev** one-click option is also available for the Megatron Bridge example:
 
-Or open the notebook and follow along step by step.
+[![Deploy on Brev](https://brev-assets.s3.us-west-1.amazonaws.com/nv-lb-dark.svg)](https://brev.nvidia.com/launchable/deploy?launchableID=env-39PnUMhHmxbMcHKO61iQ8O5F7ZL)

use-case-examples/sql-lora-finetuning-and-deployment/finetune.py

@@ -0,0 +1,40 @@
+# Copyright (c) 2025, NVIDIA CORPORATION. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import annotations
+
+import os
+
+from nemo_automodel.components.config._arg_parser import parse_args_and_load_config
+from nemo_automodel.recipes.llm.train_ft import TrainFinetuneRecipeForNextTokenPrediction
+
+# Default config: YAML in the same directory as this script (BIRD SQL PEFT for Nemotron-3-Nano, 2 GPUs)
+_THIS_DIR = os.path.dirname(os.path.abspath(__file__))
+_DEFAULT_CONFIG = os.path.join(_THIS_DIR, "bird_peft_nemotron_nano.yaml")
+
+
+def main(default_config_path: str = _DEFAULT_CONFIG) -> None:
+    """Main entry point for the fine-tuning recipe.
+
+    Loads the configuration, sets up the recipe, and initiates the training loop.
+    Override the config path via CLI: --config /path/to/other.yaml
+    """
+    cfg = parse_args_and_load_config(default_config_path)
+    recipe = TrainFinetuneRecipeForNextTokenPrediction(cfg)
+    recipe.setup()
+    recipe.run_train_validation_loop()
+
+
+if __name__ == "__main__":
+    main()