You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

SainiON Hacks Bug Hunting

A fine-tuned Qwen 3B language model optimized for improved conversational responses using LoRA (Low-Rank Adaptation). The LoRA adapter has been merged into the base model, making this a standalone model ready for inference.


Base Model

  • Architecture: Qwen 3B
  • Fine-tuning Method: LoRA
  • Merged: Yes
  • Framework: Transformers
  • Format: Hugging Face Transformers

Features

  • Improved conversational responses
  • Better greeting and small-talk handling
  • Improved instruction following
  • Friendly assistant behavior
  • Standalone merged model (no LoRA adapter required)

Training Details

Item Value
Method LoRA Fine-tuning
Epochs 3
Optimizer AdamW
Learning Rate 2e-4
Batch Size 2
Gradient Accumulation 4

Intended Use

This model is intended for:

  • General conversation
  • Question answering
  • Writing assistance
  • Programming assistance
  • Educational purposes
  • Research and experimentation

Limitations

  • Knowledge is limited to the base model's training cutoff.
  • May produce incorrect or hallucinated information.
  • Responses should be verified before use in critical environments.
  • Not intended for high-risk or safety-critical decisions.

Loading the Model

from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "Nitinsaini077/SainiON-BugHunt"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto"
)

Example

messages = [
    {"role": "user", "content": "Hi"}
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

inputs = tokenizer(text, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=128
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Training Pipeline

  1. Loaded the base Qwen 3B model
  2. Applied LoRA fine-tuning
  3. Trained on conversational instruction-response pairs
  4. Merged LoRA weights into the base model
  5. Exported as a standalone Hugging Face model

Repository Structure

config.json
generation_config.json
model-00001-of-00002.safetensors
model-00002-of-00002.safetensors
model.safetensors.index.json
tokenizer.json
tokenizer_config.json
chat_template.jinja

Future Improvements

  • Larger instruction datasets
  • Cybersecurity and VAPT fine-tuning
  • AI Security datasets
  • Retrieval-Augmented Generation (RAG)
  • Better reasoning and code generation
  • GGUF export for local inference

License

Please refer to the license of the original Qwen base model. This repository contains a fine-tuned derivative and should be used in accordance with the original model's license.


Acknowledgements

  • Alibaba Qwen Team
  • Hugging Face Transformers
  • PEFT (LoRA)
  • TRL
Downloads last month
-
Safetensors
Model size
3B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 1 Ask for provider support