README.md

---
base_model: Haleshot/Mathmate-7B-DELLA-ORPO
tags:
- finetuned
- orpo
- everyday-conversations
- adapter
datasets:
- HuggingFaceTB/everyday-conversations-llama3.1-2k
license: apache-2.0
language:
- en
library_name: transformers
pipeline_tag: text-generation
---

# Mathmate-7B-DELLA-ORPO-C

Mathmate-7B-DELLA-ORPO-C is a LoRA adapter for [Haleshot/Mathmate-7B-DELLA-ORPO](https://huggingface.co/Haleshot/Mathmate-7B-DELLA-ORPO), finetuned to improve performance on everyday conversations.

## Model Details

- **Base Model:** [Haleshot/Mathmate-7B-DELLA](https://huggingface.co/Haleshot/Mathmate-7B-DELLA-ORPO)
- **Training Dataset:** [HuggingFaceTB/everyday-conversations-llama3.1-2k](https://huggingface.co/datasets/HuggingFaceTB/everyday-conversations-llama3.1-2k)

## Dataset

The model was finetuned on the [HuggingFaceTB/everyday-conversations-llama3.1-2k](https://huggingface.co/datasets/HuggingFaceTB/everyday-conversations-llama3.1-2k) dataset, which focuses on everyday conversations and small talk.

## Usage

To use this LoRA adapter, you need to load both the base model and the adapter. Here's an example:

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel, PeftConfig
import torch

base_model_name = "Haleshot/Mathmate-7B-DELLA"
adapter_name = "Haleshot/Mathmate-7B-DELLA-ORPO-C"

base_model = AutoModelForCausalLM.from_pretrained(base_model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(base_model_name)
model = PeftModel.from_pretrained(base_model, adapter_name)

def generate_response(prompt, max_length=512):
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(**inputs, max_length=max_length, num_return_sequences=1, do_sample=True, temperature=0.7)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

prompt = "Let's have a casual conversation about the weather today."
response = generate_response(prompt)
print(response)
```

## Acknowledgements

Thanks to the HuggingFaceTB team for providing the everyday conversations dataset used in this finetuning process.