modelgpt 0.2.0

LLM-powered AutoML — model selection, tuning, and training via any LLM

ModelGPT

LLM-powered AutoML. Describe your dataset, pick a task — ModelGPT asks a local LLM to design, tune, and fit the best model for you.

from modelgpt import ModelGPT

mg = ModelGPT()
model = mg.fit(X_train, y_train, task="regression")
predictions = model.predict(X_test)

---

How it works

Your data  ──▶  Dataset summary  ──▶  LLM prompt
                                          │
                                          ▼
                                 LLM generates Python
                                 (ensemble + CV tuning)
                                          │
                                          ▼
                                   exec() in sandbox
                                          │
                             ┌────────────┴────────────┐
                             ▼                         ▼
                       Fitted model           Error? retry (→ LLM
                       returned               sees error, self-corrects)
                                                       │
                                              Max retries hit?
                                                       ▼
                                              Safe fallback model

ModelGPT feeds the LLM a structured dataset summary (shape, dtypes, missing values, target statistics, class distribution). The LLM returns raw Python that uses cross-validated hyperparameter search and an ensemble method. That code runs in a controlled namespace with X and y already bound. If it fails, the error is sent back to the LLM for self-correction — up to max_retries times.

---

Installation

# 1. Clone
git clone https://github.com/yourname/modelGPT.git
cd modelGPT

# 2. Install Python dependencies
pip install pandas scikit-learn xgboost lightgbm catboost ollama

# 3. Install and start Ollama  (https://ollama.com)
ollama pull qwen3-vl:235b-cloud       # or any model you prefer
ollama serve

---

Project structure

modelGPT/
├── modelgpt/
│   └── modelgpt.py          # Core ModelGPT class
└── example/
    └── example_usage.py     # Quickstart demo (diabetes dataset)

---

API reference

ModelGPT(llm_model, max_retries, verbose)

Parameter	Type	Default	Description
llm_model	str	qwen3-vl:235b-cloud	Ollama model tag to use
max_retries	int	3	How many times to retry on code execution failure
verbose	bool	True	Print generated code and progress to stdout

.fit(X, y, task, metric)

Parameter	Type	Default	Description
X	pd.DataFrame	—	Feature matrix
y	pd.Series	—	Target vector
task	str	"regression"	"regression" or "classification"
metric	str | None	None	Metric hint passed to the LLM (e.g. "RMSE", "AUC")

Returns a fitted sklearn-compatible model with a .predict() method.

---

Choosing a model

ModelGPT works with any model available in Ollama. Larger models produce better code:

mg = ModelGPT(llm_model="qwen2.5-coder:32b")

---

Tips for better results

Tell the LLM which metric matters. The metric argument is forwarded directly into the prompt:

model = mg.fit(X_train, y_train, task="regression", metric="RMSE")
model = mg.fit(X_train, y_train, task="classification", metric="AUC")

Increase retries for hard tasks. If the LLM frequently generates broken code, raise max_retries:

mg = ModelGPT(max_retries=5)

Inspect what was generated. With verbose=True (the default) the full generated code is printed. You can copy it, tweak it, and re-run it manually.

Fallback is always safe. If every retry fails, ModelGPT automatically falls back to a well-tuned GradientBoostingRegressor / GradientBoostingClassifier, so .fit() never crashes your pipeline.

---

Requirements

• Python ≥ 3.10
• Ollama running locally (ollama serve)
• pandas, scikit-learn, ollama
• Optional but recommended: xgboost, lightgbm, catboost

---

License

MIT