Smart Vet
LoRA adapter for veterinary text analysis
1 min read
This LoRA adapter (a small add-on built on top of a base model) is based on unsloth/gemma-4-E2B-it using the QLoRA method for working with veterinary medical records. The model specializes in ICD-11 syndrome classification, as well as disease name extraction and personal data anonymization from text. This solution is useful for developers and researchers building tools for automated veterinary documentation processing. Note that the adapter is not an independent diagnostic tool and is used exclusively for NLP tasks.
Key features
- Multi-label ICD-11 syndrome classification
- Disease entity extraction (Disease NER) accounting for differential diagnoses
- Personal data search and masking (Anonymisation NER)
- Trained on the PetEVAL dataset using QLoRA optimization
What you can do
- Apply the adapter alongside the base model to parse electronic patient records in a veterinary clinic
- Set up automatic classification of symptoms and diagnoses in medical reports
- Use the NER component to clean and anonymize texts before further processing
Example
To run and use the adapter with the base model unsloth/gemma-4-E2B-it, the standard Hugging Face pipeline is used with 4-bit loading mode.
Read more
DetailsBase model: unsloth/gemma-4-E2B-it
Training method: QLoRA (Rank: 16, alpha: 32, dropout: 0.05)
Limitations: Not a veterinary diagnostic system, requires the base model to run










Comments