richarderkhov/nztinversive_-_llama3.2-1b-uncensored-gguf overview

{
  "metadata": {},
  "card_data": {
    "frontmatter": {},
    "hero_image_url": "",
    "summary": "",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "Quantization made by Richard Erkhov.\n\n[Github](https://github.com/RichardErkhov)\n\n[Discord](https://discord.gg/pvy7H8DZMG)\n\n[Request more models](https://github.com/RichardErkhov/quant_request)\n\n\nllama3.2-1b-Uncensored - GGUF\n- Model creator: https://huggingface.co/nztinversive/\n- Original model: https://huggingface.co/nztinversive/llama3.2-1b-Uncensored/\n\n\n| Name | Quant method | Size |\n| ---- | ---- | ---- |\n| [llama3.2-1b-Uncensored.Q2_K.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q2_K.gguf) | Q2_K | 0.54GB |\n| [llama3.2-1b-Uncensored.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.IQ3_XS.gguf) | IQ3_XS | 0.58GB |\n| [llama3.2-1b-Uncensored.IQ3_S.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.IQ3_S.gguf) | IQ3_S | 0.6GB |\n| [llama3.2-1b-Uncensored.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q3_K_S.gguf) | Q3_K_S | 0.6GB |\n| [llama3.2-1b-Uncensored.IQ3_M.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.IQ3_M.gguf) | IQ3_M | 0.61GB |\n| [llama3.2-1b-Uncensored.Q3_K.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q3_K.gguf) | Q3_K | 0.64GB |\n| [llama3.2-1b-Uncensored.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q3_K_M.gguf) | Q3_K_M | 0.64GB |\n| [llama3.2-1b-Uncensored.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q3_K_L.gguf) | Q3_K_L | 0.68GB |\n| [llama3.2-1b-Uncensored.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.IQ4_XS.gguf) | IQ4_XS | 0.7GB |\n| [llama3.2-1b-Uncensored.Q4_0.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q4_0.gguf) | Q4_0 | 0.72GB |\n| [llama3.2-1b-Uncensored.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.IQ4_NL.gguf) | IQ4_NL | 0.72GB |\n| [llama3.2-1b-Uncensored.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q4_K_S.gguf) | Q4_K_S | 0.72GB |\n| [llama3.2-1b-Uncensored.Q4_K.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q4_K.gguf) | Q4_K | 0.75GB |\n| [llama3.2-1b-Uncensored.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q4_K_M.gguf) | Q4_K_M | 0.75GB |\n| [llama3.2-1b-Uncensored.Q4_1.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q4_1.gguf) | Q4_1 | 0.77GB |\n| [llama3.2-1b-Uncensored.Q5_0.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q5_0.gguf) | Q5_0 | 0.83GB |\n| [llama3.2-1b-Uncensored.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q5_K_S.gguf) | Q5_K_S | 0.83GB |\n| [llama3.2-1b-Uncensored.Q5_K.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q5_K.gguf) | Q5_K | 0.85GB |\n| [llama3.2-1b-Uncensored.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q5_K_M.gguf) | Q5_K_M | 0.85GB |\n| [llama3.2-1b-Uncensored.Q5_1.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q5_1.gguf) | Q5_1 | 0.89GB |\n| [llama3.2-1b-Uncensored.Q6_K.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q6_K.gguf) | Q6_K | 0.95GB |\n| [llama3.2-1b-Uncensored.Q8_0.gguf](https://huggingface.co/RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf/blob/main/llama3.2-1b-Uncensored.Q8_0.gguf) | Q8_0 | 1.23GB |\n\n\n\n\nOriginal model description:\n---\nlanguage:\n- en\nlicense: mit\nlibrary_name: transformers\ntags:\n- llama\n- uncensored\n- abliteration\npipeline_tag: text-generation\n---\n\n# Uncensoring LLaMA 3.2 1B Model\n\n## Overview\n\nThis repository demonstrates the process of uncensoring a 1-billion-parameter LLaMA 3.2 model using \"abliteration.\" Abliteration allows the model to generate outputs without the restrictions imposed by its default safety mechanisms. The goal is to give developers more control over the model's output by removing censorship filters while ensuring responsible AI usage.\n\n**Disclaimer:** This model and methodology are intended for research and educational purposes only. Uncensoring models must be done with ethical considerations, and it's critical to avoid harmful or irresponsible applications.\n\n## Model Details\n\n* **Model Name**: LLaMA 3.2 (1B Parameters)\n* **Version**: Uncensored variant via the Abliteration technique\n* **Framework**: PyTorch\n* **Source**: Hugging Face LLaMA model\n\n## Abliteration: The Process\n\nAbliteration removes the filtering mechanisms from the model's decoding process, allowing more open-ended responses. It's achieved by modifying how the logits (the model's output probabilities) are handled.\n\n## How to Use\n\nTo use the uncensored model, follow the instructions below.\n\n### Requirements\n\nTo get started, install the necessary packages:\n\n```bash\npip install torch transformers\n```\n\n### Loading the Uncensored Model\n\nYou can load the uncensored model directly using the Hugging Face `transformers` library.\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\n\n# Load the tokenizer and model\ntokenizer = AutoTokenizer.from_pretrained(\"your-hf-username/uncensored-llama-3.2-1b\")\nmodel = AutoModelForCausalLM.from_pretrained(\"your-hf-username/uncensored-llama-3.2-1b\")\n```\n\n### Generating Text\n\nYou can generate text with the uncensored model using the following code:\n\n```python\ndef uncensored_generate(model, tokenizer, input_text):\n    inputs = tokenizer(input_text, return_tensors=\"pt\").input_ids\n    \n    # Generate the output without applying safety filters\n    outputs = model.generate(inputs, max_length=100, do_sample=True, temperature=0.9, top_k=50)\n    decoded_output = tokenizer.decode(outputs[0], skip_special_tokens=True)\n    return decoded_output\n\n# Example usage\ninput_text = \"What are your thoughts on controversial topics?\"\noutput = uncensored_generate(model, tokenizer, input_text)\nprint(output)\n```\n\n### Fine-Tuning the Uncensored Model (Optional)\n\nFor optimal results, you can fine-tune the model on uncensored datasets. Here's a simple way to set up fine-tuning using the Hugging Face `Trainer`:\n\n```python\nfrom transformers import Trainer, TrainingArguments\n\ntraining_args = TrainingArguments(\n    output_dir=\"./results\",\n    num_train_epochs=1,\n    per_device_train_batch_size=2,\n    save_steps=10_000,\n    save_total_limit=2,\n)\n\ntrainer = Trainer(\n    model=model,\n    args=training_args,\n    train_dataset=uncensored_dataset  # Load your uncensored dataset\n)\n\ntrainer.train()\n```\n\n## Ethical Considerations\n\nWhile this model has the ability to generate uncensored responses, it is critical to use it responsibly. Uncensored models can be prone to generating harmful or inappropriate content. Ensure you are aware of the implications of deploying uncensored models and avoid applications that may lead to unethical outcomes.\n\n## How to Contribute\n\nContributions to the project are welcome! You can fine-tune the model, improve performance, or experiment with different ways to uncensor the model.\n\n1. Fork this repository on Hugging Face.\n2. Make changes to the model or code.\n3. Share your results and improvements.\n\n## License\n\nThis model is released under the MIT License.\n\n## References\n\n* Original blog post: Uncensor any LLM with Abliteration\n* Hugging Face Transformers Documentation\n\n\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "endpoints_compatible",
    "region:us"
  ],
  "likes": 0,
  "downloads": 331,
  "gated": false,
  "private": false,
  "last_modified": "2024-10-16T20:25:24.000Z",
  "created_at": "2024-10-16T19:13:56.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "67101074375b976cc63eab1e",
  "id": "RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf",
  "modelId": "RichardErkhov/nztinversive_-_llama3.2-1b-Uncensored-gguf",
  "sha": "6491ef1153e79368e3c72c125ba1de38e56cd24a",
  "createdAt": "2024-10-16T19:13:56.000Z",
  "lastModified": "2024-10-16T20:25:24.000Z",
  "author": "RichardErkhov",
  "downloads": 331,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 24
}