pdarleyjr/iplc-t5-clinical

Clinical Report Generator API

FastAPI-based API for generating clinical report summaries using a fine-tuned T5 model optimized for clinical documentation.

Model Information

This API uses a T5 model fine-tuned specifically for clinical report generation. The model has been optimized with enhanced generation parameters for better quality summaries.

API Endpoints

POST /predict

Generate a clinical report summary from input text.

Request body:

{
    "text": "string"
}

Example:

{
    "text": "evaluation type: initial. primary diagnosis: F84.0. severity: mild. primary language: english"
}

Response:

{
    "success": true,
    "data": "string",
    "error": null,
    "metrics": {
        "process_time": 1.23
    }
}

GET /health

Health check endpoint that provides system and model status.

Response:

{
    "status": "healthy",
    "model_loaded": true,
    "last_load_time": 1234567890.123,
    "version": "1.0.0",
    "gpu_available": true,
    "gpu_name": "NVIDIA GeForce RTX 3080"
}

Input Format Guidelines

For best results with the T5 model:

1. Structure your input data clearly using key-value pairs
2. Separate different fields with periods or commas
3. Keep inputs concise and focused on relevant clinical information
4. The API automatically prefixes inputs with "summarize:" for optimal model performance

Technical Details

• Model: pdarleyjr/iplc-t5-clinical
• Max Input Length: 512 tokens
• Max Output Length: 512 tokens (increased for more comprehensive summaries)
• Generation Parameters:
  • Number of beams: 5 (increased for better coherence)
  • Temperature: 0.7 (added for more natural generation)
  • No repeat ngram size: 3
  • Length penalty: 2.0
  • Early stopping: Enabled