Instructions to use sahilsangwan/qwen35-4b-text-to-sql-data-forge-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use sahilsangwan/qwen35-4b-text-to-sql-data-forge-lora with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # if on a CUDA device, also pip install mlx[cuda] # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("sahilsangwan/qwen35-4b-text-to-sql-data-forge-lora") prompt = "Once upon a time in" text = generate(model, tokenizer, prompt=prompt, verbose=True) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- MLX LM
How to use sahilsangwan/qwen35-4b-text-to-sql-data-forge-lora with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Generate some text mlx_lm.generate --model "sahilsangwan/qwen35-4b-text-to-sql-data-forge-lora" --prompt "Once upon a time"
- Atomic Chat
| { | |
| "adapter": "sahilsangwan/qwen35-4b-text-to-sql-data-forge-lora", | |
| "base_model": "Qwen/Qwen3.5-4B", | |
| "benchmark": "Spider dev", | |
| "conclusion": "The adapter improves over the local base model on full Spider dev after deterministic SQL extraction, but does not beat the cited public Sonnet-level Spider numbers.", | |
| "database_dir": "external/spider_download/spider_data/database", | |
| "dataset_source": "official Spider dev set", | |
| "dialect": "SQLite", | |
| "generation": { | |
| "max_tokens": 256, | |
| "temperature": 0.0 | |
| }, | |
| "public_reference_numbers": [ | |
| { | |
| "benchmark": "Spider test", | |
| "execution_accuracy": 0.6991, | |
| "model": "Claude 3.5 Sonnet", | |
| "note": "Different split/harness; included only as public reference.", | |
| "source": "Optimizing Reasoning for Text-to-SQL with Execution Feedback, ACL Findings 2025" | |
| }, | |
| { | |
| "benchmark": "Spider", | |
| "execution_accuracy": 0.857, | |
| "model": "Claude 3.5 Sonnet", | |
| "note": "Different harness; included only as public reference.", | |
| "source": "Dataworkers text-to-SQL benchmark article" | |
| } | |
| ], | |
| "raw_execution": { | |
| "base": { | |
| "correct": 428, | |
| "execution_accuracy": 0.4139 | |
| }, | |
| "delta_points": -4.06, | |
| "fine_tuned": { | |
| "correct": 386, | |
| "execution_accuracy": 0.3733 | |
| }, | |
| "note": "Raw model output was evaluated without SQL extraction; many generations contained reasoning text before SQL." | |
| }, | |
| "sample_count": 1034, | |
| "sql_extracted_execution": { | |
| "base": { | |
| "correct": 422, | |
| "execution_accuracy": 0.4081, | |
| "total": 1034, | |
| "valid_sql": 573, | |
| "valid_sql_rate": 0.5542 | |
| }, | |
| "base_extraction": { | |
| "changed": 840, | |
| "input": "generation/niches/text-to-sql/evals/spider_dev_qwen35_lora_800/base_predictions_full.jsonl", | |
| "no_sql_found": 194, | |
| "out": "generation/niches/text-to-sql/evals/spider_dev_qwen35_lora_800/base_predictions_full_sql_extracted.jsonl", | |
| "total": 1034, | |
| "unchanged": 0 | |
| }, | |
| "execution_accuracy_delta_points": 17.02, | |
| "fine_tuned": { | |
| "correct": 598, | |
| "execution_accuracy": 0.5783, | |
| "total": 1034, | |
| "valid_sql": 809, | |
| "valid_sql_rate": 0.7824 | |
| }, | |
| "fine_tuned_extraction": { | |
| "changed": 1014, | |
| "input": "generation/niches/text-to-sql/evals/spider_dev_qwen35_lora_800/fine_tuned_predictions_full.jsonl", | |
| "no_sql_found": 20, | |
| "out": "generation/niches/text-to-sql/evals/spider_dev_qwen35_lora_800/fine_tuned_predictions_full_sql_extracted.jsonl", | |
| "total": 1034, | |
| "unchanged": 0 | |
| }, | |
| "valid_sql_delta_points": 22.82 | |
| } | |
| } | |