README.md

---
language: en
license: apache-2.0
library_name: peft
tags:
- h2oai
- causal-lm
- text-generation
- adhd
- cpt-ii
- clinical-assistant
base_model: h2oai/h2o-danube3-500m-chat
---

# ADHD CPT Analyst

## Model Description

This model is a fine-tuned version of `h2oai/h2o-danube3-500m-chat`, specifically adapted for analyzing and interpreting textual reports from the Conners' Continuous Performance Test II (CPT-II). It has been trained using Low-Rank Adaptation (LoRA) on a dataset of CPT-II results to identify patterns relevant to the assessment of ADHD.

The model takes a textual summary of a patient's CPT-II scores as input and can provide analysis, explanations of the metrics, and potential interpretations.

## Intended Uses & Limitations

This model is intended as a research and educational tool. It can be used to:
- Assist researchers in analyzing patterns across large datasets of CPT-II reports.
- Help students and trainees learn about the different metrics in a CPT-II report and their potential clinical significance.
- Provide a preliminary interpretation of a CPT-II report.

**Crucial Disclaimer:** This model is **not a medical device** and should **not** be used for self-diagnosis or as a substitute for professional medical advice, diagnosis, or treatment. Always consult with a qualified healthcare provider for any health-related concerns.

## How to Use

To use this model, you need to load the base model (`h2oai/h2o-danube3-500m-chat`) and then apply the LoRA adapter.

```python
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

# Model and repository parameters
base_model_id = "h2oai/h2o-danube3-500m-chat"
adapter_id = "monkwarrior08/adhd-cpt-analyst"

# Load tokenizer and base model
tokenizer = AutoTokenizer.from_pretrained(base_model_id)
base_model = AutoModelForCausalLM.from_pretrained(
    base_model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

# Load the LoRA adapter
model = PeftModel.from_pretrained(base_model, adapter_id)
model.eval()

# Create a prompt with a patient's data
patient_report = """
Patient ID: 3.0
Assessment Status: 3.0
Assessment Duration: 839999.0 seconds

CPT II Summary Report:
- Omissions:
  - General T-Score: 78.75
  - ADHD T-Score: 70.25
  - Raw Score: 11.0
- Commissions:
  - General T-Score: 65.98
  - ADHD T-Score: 70.89
  - Raw Score: 28.0
- Hit Reaction Time (HitRT):
  - General T-Score: 36.57
  - Mean Reaction Time: 325.20 ms
ADHD Confidence Index: 86.87
"""

prompt = f"<|prompt|>Analyze this CPT-II report and summarize the findings for potential indicators of ADHD.:\\n{patient_report}<|end|><|answer|>"

# Generate a response
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(
    **inputs, 
    max_new_tokens=256, 
    eos_token_id=tokenizer.eos_token_id,
    do_sample=True,
    temperature=0.6,
    top_p=0.9,
)

response = tokenizer.decode(outputs[0], skip_special_tokens=True)

# Extract only the model's answer
answer = response.split('<|answer|>')[1].strip()
print(answer)
```

## Training Data

The model was fine-tuned on a private dataset derived from the `ADHD Diagnosis CPT II Data.csv` file. Each record was converted into a textual summary containing the following key metrics:
- Omissions (T-Scores, Raw Score)
- Commissions (T-Scores, Raw Score)
- Hit Reaction Time (T-Score, Mean)
- Variability of SE
- d'
- Beta
- Perseverations
- ADHD Confidence Index

## Training Procedure

The model was trained using the `trl` library's `SFTTrainer` with a LoRA configuration. The primary goal was to teach the model to understand the relationship between the various CPT-II metrics and their relevance in ADHD assessment.

### BibTeX Citation

If you use this model in your research, please consider citing it:

```bibtex
@software{monkwarrior08_2024_adhd_cpt_analyst,
  author = {monkwarrior08},
  title = {ADHD CPT Analyst: A Fine-tuned Language Model for CPT-II Report Interpretation},
  month = {8},
  year = {2024},
  url = {https://huggingface.co/monkwarrior08/adhd-cpt-analyst}
}
```