---
license: apache-2.0
base_model: teknium/OpenHermes-2.5-Mistral-7B
tags:
- generated_from_trainer
- audiobook
- fine-tuned
- text-generation
- lora
- axolotl
language:
- en
pipeline_tag: text-generation
widget:
- text: "Who is the main character in the story?"
- text: "Describe the setting of the book."
- text: "What are the main themes explored?"
---

# the-silver-pigs-finetuned

This model was fine-tuned on audiobook content using the AudioBook Visualizer application.

## Model Details

- **Base Model**: teknium/OpenHermes-2.5-Mistral-7B
- **Fine-tuning Method**: QLoRA (4-bit quantization with LoRA adapters)
- **Training Framework**: Axolotl
- **Training Infrastructure**: RunPod Serverless
- **Upload Date**: 2025-08-11T05:51:10.991Z

## Training Configuration

- **LoRA Rank**: 32
- **LoRA Alpha**: 16
- **LoRA Dropout**: 0.05
- **Target Modules**: q_proj, v_proj, k_proj, o_proj
- **Learning Rate**: 2e-4
- **Batch Size**: 2
- **Gradient Accumulation**: 4
- **Training Duration**: ~3 minutes

## Usage with vLLM

Deploy on RunPod serverless with these environment variables:

```json
{
  "MODEL_NAME": "the-silver-pigs-finetuned",
  "TRUST_REMOTE_CODE": "true",
  "MAX_MODEL_LEN": "2048",
  "DTYPE": "float16",
  "ENABLE_LORA": "true"
}
```

## Usage with Transformers

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

base_model = "teknium/OpenHermes-2.5-Mistral-7B"
model = AutoModelForCausalLM.from_pretrained(
    base_model,
    torch_dtype=torch.float16,
    device_map="auto"
)
model = PeftModel.from_pretrained(model, "the-silver-pigs-finetuned")

tokenizer = AutoTokenizer.from_pretrained("the-silver-pigs-finetuned")

# Query the model
prompt = "Tell me about the main character."
inputs = tokenizer(prompt, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=200, temperature=0.7)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)
```

## Training Data

This model was fine-tuned on audiobook transcript data, processed into ~490 question-answer pairs covering:
- Character descriptions and relationships
- Plot events and summaries
- Dialogue and memorable quotes
- Settings and world-building details

## Limitations

- Model knowledge is limited to the specific audiobook content
- May generate plausible but incorrect details
- Performance on general tasks may differ from base model

## Created With

[AudioBook Visualizer](https://github.com/yourusername/audiobook-visualizer) - An application for fine-tuning LLMs on audiobook content.

---
*Original model path: /runpod-volume/fine-tuning/training-1754891374667*