willianr/llama-3.1-8b-attribute-extractor-qlora

TEXT GENERATIONPricing:Input $0.37 / Cached $0.074 / Output $0.38Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:8kTool Calling:SupportedPublished:Sep 9, 2026Architecture:Transformer Featherless Exclusive Cold

The willianr/llama-3.1-8b-attribute-extractor-qlora is an 8.11 billion parameter Llama 3.1 model fine-tuned using QLoRA for attribute extraction from product descriptions. It specializes in transforming unstructured text, particularly for smartphones, into structured JSON outputs like brand, model, and storage. This model is designed to facilitate product identification and clustering by providing a semantic layer for entity resolution.

Loading preview...

Overview

This model, willianr/llama-3.1-8b-attribute-extractor-qlora, is an 8.11 billion parameter Llama 3.1 model fine-tuned with QLoRA (4-bit quantization) to perform attribute extraction from textual product descriptions. Developed by willianr, it transforms unstructured text into structured JSON, identifying attributes such as brand, model, RAM, storage, processor, and operating system. The initial training focused on smartphone data, with a long-term vision to expand to other product categories and serve as a generalizable semantic layer for entity identification and clustering.

Key Capabilities

  • Structured Attribute Extraction: Converts product descriptions into JSON objects with identified attributes.
  • Specialized for Smartphones: Currently optimized for extracting details from smartphone titles and descriptions.
  • QLoRA Fine-tuning: Utilizes QLoRA with a base model of unsloth/meta-llama-3.1-8b-unsloth-bnb-4bit, making it efficient with only ~1.03% trainable parameters.
  • Data Pipeline: Incorporates a robust data preparation pipeline including web scraping, cleaning, normalization, deduplication, attribute construction, and data augmentation.

Good For

  • E-commerce Platforms: Ideal for organizing and categorizing product listings from diverse sources.
  • Product Data Management: Structuring unstructured product information for databases or search engines.
  • Entity Resolution: As an intermediate step to identify and cluster similar products based on extracted attributes.
  • Research in LLM Applications: Investigating the use of LLMs for information extraction and entity resolution problems.

Finetunes
1 models