anjelinejeline/Qwen2.5-14B-Instruct-epi

TEXT GENERATIONConcurrent Unit Cost:1Model Size:14.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 17, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Qwen2.5-14B-Instruct-Epi is a 14 billion parameter instruction-tuned language model developed by anjelinejeline, fine-tuned from Qwen2.5-14B-Instruct. This model is specifically domain-adapted for classifying epidemiological news articles related to West Nile virus (WNV). It excels at extracting structured outbreak information and generating a JSON object containing classification, outbreak status, affected species, countries, publication metadata, and event dates.

Loading preview...

Model Overview

anjelinejeline/Qwen2.5-14B-Instruct-epi is a specialized language model derived from Qwen2.5-14B-Instruct, developed by anjelinejeline. It has been meticulously fine-tuned to address a specific domain: the classification of epidemiological news articles, particularly those concerning West Nile virus (WNV).

Key Capabilities

  • Domain Adaptation: Specifically trained for epidemiological text analysis related to WNV.
  • Information Extraction: Designed to extract structured outbreak information from news articles.
  • Structured Output: Generates a JSON object containing critical details such as:
    • Article classification
    • Outbreak status
    • Affected species
    • Relevant countries
    • Publication metadata
    • Event dates

Training Details

The model was fine-tuned using the Unsloth framework, leveraging Low-Rank Adaptation (LoRA) for efficient parameter-efficient fine-tuning. The final LoRA adapters were merged into the base model, resulting in a standalone FP16 model optimized for inference. This training approach, utilizing Unsloth and Huggingface's TRL library, enabled a 2x faster training process.