Uploaded model
- Finetuned to generate pandas codes given a dataframe and user query.
- ~100 datasets were taken from kaggle https://www.kaggle.com/datasets?search=Tabular+data
- These dataset were used to generate 390 sets of data queries and pandas code answers via llama3-70b https://www.kaggle.com/code/samsonleegh/sampling-data-qns-and-pandas-ans-from-dataset
- Finetuned llama3-8b-4bit with LoRA 16 adapters on 350 queries and answers pair https://colab.research.google.com/drive/1UkqjHIq-mP22AfHZCWz4kiU7hcWaXfgi?usp=sharing
- Compare ROUGE score of original vs finetuned model on 40 queries and answers pair
ROUGE Score Comparison
| Metric | llama3-8b | llama3-8b finetuned |
|---|---|---|
| ROUGE-1 | 0.4415 | 0.6585 |
| ROUGE-2 | 0.2480 | 0.4810 |
| ROUGE-L | 0.3155 | 0.5552 |
| ROUGE-Lsum | 0.3013 | 0.5570 |
- Developed by: samsonleegh
- License: apache-2.0
- Finetuned from model : unsloth/llama-3-8b-bnb-4bit
This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.
Inference Providers
NEW
This model is not currently available via any of the supported third-party Inference Providers, and
HF Inference API was unable to determine this model’s pipeline type.
Model tree for samsonleegh/lora_pandas_model
Base model
meta-llama/Meta-Llama-3-8B
Quantized
unsloth/llama-3-8b-bnb-4bit