Instructions to use yashcse21/text2sql-llama1b-adapters-caution with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use yashcse21/text2sql-llama1b-adapters-caution with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir text2sql-llama1b-adapters-caution yashcse21/text2sql-llama1b-adapters-caution
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
| license: llama3.2 | |
| base_model: mlx-community/Llama-3.2-1B-Instruct-bf16 | |
| tags: [text-to-sql, lora, mlx, reasoning] | |
| library_name: mlx | |
| # Text-to-SQL · adapters_caution | |
| MLX LoRA adapter fine-tuned from | |
| `mlx-community/Llama-3.2-1B-Instruct-bf16` to turn a natural-language question + a `CREATE TABLE` schema into a | |
| schema-faithful SQLite `SELECT` (reasoning models emit `Reasoning: … / SQL: …`). | |
| ```python | |
| # download the adapter, then: | |
| from mlx_lm import load | |
| model, tok = load("mlx-community/Llama-3.2-1B-Instruct-bf16", adapter_path="<downloaded-adapter-dir>") | |
| ``` | |
| Pair generation with a deterministic AST guardrail (single read-only SELECT). | |
| See the project repo for the data pipeline, study, and evaluation. | |