LLaMaPaca
Model Details Model Name: LLaMaPaca Base Model: unsloth/llama-3.2-1b-instruct-unsloth-bnb-4bit Adapter Type: LoRA (Low-Rank Adaptation) Library: PEFT (Parameter-Efficient Fine-Tuning) Pipeline Tag: text-generation
Description
LLaMaPaca is a LoRA adapter fine-tuned on the LLaMA 3.2 1B Instruct model using Unsloth's optimized training framework. This adapter enables parameter-efficient customization of the base model for specific tasks or domains while maintaining the core capabilities of LLaMA 3.2. The adapter was trained using 4-bit quantization via bitsandbytes, making it memory-efficient and suitable for deployment on consumer-grade hardware.
Technical Specifications
Architecture: LLaMA 3.2 with LoRA adapters Base Model Size: ~1B parameters Quantization: 4-bit (bitsandbytes) Training Framework: Unsloth + PEFT Adapter Format: PEFT LoRA
Usage
Installation
bash
pip install transformers peft accelerate bitsandbytes
Loading the Model
python
from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel# Load base modelbase_model_name = "unsloth/llama-3.2-1b-instruct-unsloth-bnb-4bit"adapter_name = "your-username/LLaMaPaca" # Replace with actual repotokenizer = AutoTokenizer.from_pretrained(base_model_name)model = AutoModelForCausalLM.from_pretrained( base_model_name, load_in_4bit=True, device_map="auto")# Load LoRA adaptermodel = PeftModel.from_pretrained(model, adapter_name)# Generate textprompt = "Your instruction here..."inputs = tokenizer(prompt, return_tensors="pt").to("cuda")outputs = model.generate(**inputs, max_new_tokens=256)print(tokenizer.decode(outputs[0], skip_special_tokens=True)) Using with Text Generation Pipeline
python
from transformers import pipelinepipe = pipeline( "text-generation", model=base_model_name, model_kwargs={"load_in_4bit": True}, adapter_name=adapter_name)result = pipe("Your prompt here...", max_new_tokens=256)
Training Details
Method: LoRA (Low-Rank Adaptation) Optimization: Unsloth acceleration Quantization: 4-bit precision with bitsandbytes Framework: PEFT + Transformers Intended Use Cases Instruction following and conversational AI Domain-specific text generation Custom task adaptation with minimal resource requirements Edge deployment scenarios requiring efficient models Limitations Performance depends on the quality and quantity of fine-tuning data May inherit biases from the base LLaMA 3.2 model 4-bit quantization may result in slight accuracy trade-offs Adapter is specific to the base model architecture Citation If you use this model in your research, please cite: bibtex & TensorVizion
@misc{llamapaca, title={LLaMaPaca: LoRA Adapter for LLaMA 3.2 1B Instruct}, author={Tensorizion}, year={2026}, publisher={Hugging Face}, howpublished={\url{https://huggingface.co/TensorVizion/LLaMaPaca}}}
License
Please refer to the base model license (LLaMA 3.2 Community License) and specify any additional licensing terms for your adapter.
- Downloads last month
- 41