| --- |
| license: apache-2.0 |
| base_model: mistralai/Mistral-7B-Instruct-v0.2 |
| tags: |
| - programmatic-advertising |
| - ad-tech |
| - fine-tuned |
| - lora |
| - mistral |
| language: |
| - en |
| --- |
| |
| # Programmatic AdTech LLM — Mistral 7B |
|
|
| A fine-tuned version of Mistral-7B-Instruct-v0.2 trained on programmatic advertising domain knowledge. |
|
|
| ## Model Details |
|
|
| - **Base model:** mistralai/Mistral-7B-Instruct-v0.2 |
| - **Fine-tuning method:** QLoRA (4-bit quantization + LoRA) |
| - **Developed by:** muradAgain |
| - **Language:** English |
| - **License:** Apache 2.0 |
|
|
| ## Training Details |
|
|
| - **Dataset:** 96 ad tech Q&A pairs across 10 categories |
| - **Categories:** Fundamentals, Supply Path, Ad Fraud, Targeting, CTV, DOOH, Measurement, Privacy, Political Advertising, Brand Safety, Campaign Operations |
| - **Epochs:** 3 |
| - **Learning rate:** 2e-4 |
| - **Hardware:** Kaggle T4 x2 GPU (free tier) |
| - **Final training loss:** 0.19 |
|
|
| ## Use Cases |
|
|
| - Ad tech education and onboarding |
| - Programmatic advertising Q&A assistant |
| - Internal tooling for AdOps teams |
| - Embedded chatbot for ad tech platforms |
|
|
| ## Limitations |
|
|
| - Trained on a small dataset (96 examples) |
| - Knowledge limited to training data categories |
| - Not suitable for real-time bidding decisions or financial advice |
