shadicopty/llama3.2-entity
TEXT GENERATIONConcurrency Cost:1Model Size:3.2BQuant:BF16Ctx Length:32kPublished:Dec 23, 2024License:apache-2.0Architecture:Transformer0.0K Open Weights Warm
shadicopty/llama3.2-entity is a 3.2 billion parameter language model fine-tuned from Llama3.2-3b with a 32768 token context length. It is specifically trained to extract and format names of companies, projects, and people into a JSON structure. This model is optimized for on-device text anonymization before data is sent to cloud services.
Loading preview...
Model Overview
shadicopty/llama3.2-entity is a specialized 3.2 billion parameter language model, fine-tuned from the Llama3.2-3b architecture. It boasts a substantial context length of 32768 tokens, enabling it to process longer texts for entity extraction tasks.
Key Capabilities
- Entity Extraction: Proficiently identifies and extracts names of companies, projects, and people from text.
- Structured Output: Formats extracted entities into a specific JSON structure, making it easy for downstream processing.
- On-Device Anonymization: Designed for use in scenarios where text needs to be anonymized locally (e.g., on a desktop) before being transmitted to cloud-based services, enhancing privacy and data security.
Training Details
The model was fine-tuned using the unsloth library on Google Colab, indicating an efficient and accessible training methodology.
Ideal Use Cases
- Privacy-Preserving Applications: Excellent for applications requiring the removal or obfuscation of sensitive entity information from text before cloud processing.
- Data Pre-processing: Useful for preparing datasets by extracting and structuring key entities for analysis or further machine learning tasks.
- Local Data Handling: Suited for environments where data residency or privacy regulations necessitate local processing of sensitive information.