abubakaraabi786/tinyllama-peft-merged

TEXT GENERATIONConcurrency Cost:1Model Size:1.1BQuant:BF16Ctx Length:2kPublished:Apr 26, 2026License:apache-2.0Architecture:Transformer Open Weights Cold

abubakaraabi786/tinyllama-peft-merged is a 1.1 billion parameter TinyLlama model, fine-tuned using PEFT LoRA and fully merged for direct inference. Developed by abubakaraabi786, this model is designed for text generation tasks with a 2048-token context length. It offers a production-ready solution for applications requiring a compact yet capable language model without the need for PEFT adapters during inference.

Loading preview...

TinyLlama PEFT Merged: A Production-Ready Fine-Tuned Model

This model, abubakaraabi786/tinyllama-peft-merged, is a 1.1 billion parameter TinyLlama variant that has been fine-tuned using PEFT (Parameter-Efficient Fine-Tuning) LoRA and subsequently fully merged. This merging process means the model is production-ready and does not require PEFT adapters for inference, simplifying deployment.

Key Capabilities & Features

  • Compact Size: With 1.1 billion parameters and a 2.2 GB footprint, it's suitable for resource-constrained environments.
  • Direct Inference: No PEFT framework is needed for inference; simply load and generate text.
  • PyTorch Safetensors Format: Provided in FP16 precision for efficient loading and execution.
  • Context Length: Supports a context window of 2048 tokens, allowing for moderately long inputs and outputs.
  • Text Generation: Optimized for various text generation tasks, as demonstrated by its prompt format.

Good For

  • Rapid Deployment: Ideal for developers looking for a fine-tuned model that can be used immediately without complex setup.
  • Resource-Efficient Applications: Its smaller size makes it suitable for scenarios where computational resources are limited.
  • Custom Question Answering: The prompt format (Q: ...\nA:) suggests suitability for instruction-following and question-answering tasks, particularly those aligned with its training data, which includes specific educational content.