README.md

metadata

base_model: Qwen/Qwen2.5-7B-Instruct
library_name: peft
pipeline_tag: text-generation
language: en
license: mit
tags:
  - lora
  - sft
  - transformers
  - trl
  - unsloth
  - fine-tuned
datasets:
  - theprint/Advocate-9.4k

DevilsAdvocate-7B

A fine-tuned Qwen 2.5 7B model, fine tuned for more engaging conversation, encouraging the user to think about different aspects.

Model Details

This model is a fine-tuned version of Qwen/Qwen2.5-7B-Instruct using the Unsloth framework with LoRA (Low-Rank Adaptation) for efficient training.

• Developed by: theprint
• Model type: Causal Language Model (Fine-tuned with LoRA)
• Language: en
• License: mit
• Base model: Qwen/Qwen2.5-7B-Instruct
• Fine-tuning method: LoRA with rank 128

Intended Use

Brainstorming, idea development, general conversation

GGUF Quantized Versions

Quantized GGUF versions are available in the theprint/DevilsAdvocate-7B-GGUF repo.

• DevilsAdvocate-7B-f16.gguf (14531.9 MB) - 16-bit float (original precision, largest file)
• DevilsAdvocate-7B-q3_k_m.gguf (3632.0 MB) - 3-bit quantization (medium quality)
• DevilsAdvocate-7B-q4_k_m.gguf (4466.1 MB) - 4-bit quantization (medium, recommended for most use cases)
• DevilsAdvocate-7B-q5_k_m.gguf (5192.6 MB) - 5-bit quantization (medium, good quality)
• DevilsAdvocate-7B-q6_k.gguf (5964.5 MB) - 6-bit quantization (high quality)
• DevilsAdvocate-7B-q8_0.gguf (7723.4 MB) - 8-bit quantization (very high quality)

Training Details

Training Data

This data set was created to limit sycofancy and encourage gentle pushback in language models.

• Dataset: theprint/Advocate-9.4k
• Format: alpaca

Training Procedure

• Training epochs: 2
• LoRA rank: 128
• Learning rate: 0.0001
• Batch size: 1
• Framework: Unsloth + transformers + PEFT
• Hardware: NVIDIA RTX 5090

Usage

from unsloth import FastLanguageModel
import torch

# Load model and tokenizer
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="theprint/DevilsAdvocate-7B",
    max_seq_length=4096,
    dtype=None,
    load_in_4bit=True,
)

# Enable inference mode
FastLanguageModel.for_inference(model)

# Example usage
inputs = tokenizer(["Your prompt here"], return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.7)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

Alternative Usage (Standard Transformers)

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained(
    "theprint/DevilsAdvocate-7B",
    torch_dtype=torch.float16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("theprint/DevilsAdvocate-7B")

# Example usage
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Your question here"}
]

inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True)
outputs = model.generate(inputs, max_new_tokens=256, temperature=0.7, do_sample=True)
response = tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True)
print(response)

Using with llama.cpp

# Download a quantized version (q4_k_m recommended for most use cases)
wget https://huggingface.co/theprint/DevilsAdvocate-7B/resolve/main/gguf/DevilsAdvocate-7B-q4_k_m.gguf

# Run with llama.cpp
./llama.cpp/main -m DevilsAdvocate-7B-q4_k_m.gguf -p "Your prompt here" -n 256

Limitations

May provide incorrect information.

Citation

If you use this model, please cite:

@misc{devilsadvocate_7b,
  title={DevilsAdvocate-7B: Fine-tuned Qwen/Qwen2.5-7B-Instruct},
  author={theprint},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/theprint/DevilsAdvocate-7B}
}

Acknowledgments

• Base model: Qwen/Qwen2.5-7B-Instruct
• Training dataset: theprint/Advocate-9.4k
• Fine-tuning framework: Unsloth
• Quantization: llama.cpp