|
|
--- |
|
|
library_name: transformers |
|
|
tags: [phi3, fine-tuning, code-generation, matplotlib, seaborn, text-to-code] |
|
|
--- |
|
|
|
|
|
# Model Card for `ph3-FineTunned-matplotlib-seaborn-10k` |
|
|
|
|
|
This is a fine-tuned version of the **Phi-3** language model designed to generate Python data visualization code (using `matplotlib` and `seaborn`) from natural language prompts. It has been trained on 10,000 high-quality prompt–completion pairs focused on data plotting. |
|
|
|
|
|
--- |
|
|
|
|
|
## Model Details |
|
|
|
|
|
### Model Description |
|
|
|
|
|
- **Developed by:** [Prashant Suresh Shirgave](https://huggingface.co/prashantss1404) |
|
|
- **Shared by:** prashantss1404 |
|
|
- **Model type:** Text-to-Code Generation (Instruction-based) |
|
|
- **Language(s):** English (data viz-related queries) |
|
|
- **License:** Apache 2.0 |
|
|
- **Finetuned from model:** [Phi-3 Mini](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct) |
|
|
|
|
|
### Model Sources |
|
|
|
|
|
- **Model repository:** https://huggingface.co/prashantss1404/ph3-FineTunned-matplotlib-seaborn-10k |
|
|
- **Training dataset:** https://huggingface.co/datasets/prashantss1404/Matplotlib_Seaborn_merged_prompt_completion_10k |
|
|
- **Training Colab:** https://huggingface.co/prashantss1404/ph3-FineTunned-matplotlib-seaborn-10k/blob/main/Fine_Tunned_Phi3_on_matplotlib_and_seaborn.ipynb |
|
|
|
|
|
--- |
|
|
|
|
|
## Uses |
|
|
|
|
|
### Direct Use |
|
|
This model is designed to: |
|
|
- Generate Python visualization code (`matplotlib`, `seaborn`) from natural language queries. |
|
|
- Help automate plotting tasks in notebooks, dashboards, or LLM-based assistants. |
|
|
|
|
|
### Out-of-Scope Use |
|
|
- Not suitable for general-purpose coding outside of data visualization. |
|
|
- Not optimized for plotly or non-Python frameworks. |
|
|
|
|
|
--- |
|
|
|
|
|
## Bias, Risks, and Limitations |
|
|
|
|
|
### Limitations |
|
|
- The model was fine-tuned on text-based examples |
|
|
- It may generate incorrect or non-working code for complex queries. |
|
|
- No built-in code execution or error checking is included. |
|
|
- Meant for educational and experimental use only. |
|
|
|
|
|
### Recommendations |
|
|
Always **validate generated code** before executing. Combine with an execution sandbox (e.g., Streamlit, Jupyter) for best results. |
|
|
|
|
|
--- |
|
|
|
|
|
## How to Get Started with the Model |
|
|
|
|
|
```python |
|
|
from transformers import AutoTokenizer, AutoModelForCausalLM |
|
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained("prashantss1404/ph3-FineTunned-matplotlib-seaborn-10k") |
|
|
model = AutoModelForCausalLM.from_pretrained("prashantss1404/ph3-FineTunned-matplotlib-seaborn-10k") |
|
|
|
|
|
prompt = "Plot a bar chart of sales by region using seaborn" |
|
|
inputs = tokenizer(prompt, return_tensors="pt") |
|
|
outputs = model.generate(**inputs, max_new_tokens=200) |
|
|
print(tokenizer.decode(outputs[0])) |
|
|
|
