---
license: other
base_model: unsloth/Qwen2.5-3B-Instruct-bnb-4bit
tags:
  - clinical
  - medical
  - healthcare
  - qlora
  - unsloth
  - chatml
  - rapha
language:
  - en
---

# Rapha — Clinical AI Physician Assistant

Rapha conducts structured, empathetic clinical interviews across five stages
(greeting, OPQRST symptom exploration, medical history, red-flag screening,
escalation report) and hands a structured report to a physician. **Rapha
never diagnoses.**

- **Base model:** `unsloth/Qwen2.5-3B-Instruct-bnb-4bit`
- **Method:** QLoRA (Unsloth) → curriculum SFT → DPO
- **Chat template:** ChatML
- **Context window:** 8,192 tokens (training) / 4,096 (Ollama default)
- **Trained:** 2026-07-30

## Training architecture (v2.5)

Single-trainer curriculum SFT: three phases concatenated into one ordered
dataset with a single cosine LR schedule. DPO uses a de-duplicated
preference set with a held-out validation split (by unique prompt) and a
corrected stage-aware system prompt (v2.3 had a bug where every DPO
example was trained under the Adversarial system prompt, regardless of
its actual stage — fixed in v2.4).

| Phase | Data | Purpose |
|---|---|---|
| 1 | Stage1 + Stage4 + Adversarial | Interview boundaries + robustness |
| 2 | Stage2 + Stage3 | Clinical content (OPQRST + history) |
| 3 | Full mix + FullArc | Complete session flows (Stage1→5) |

## Repo contents

| Path | Contents |
|---|---|
| `/` (root) | LoRA adapter (PEFT) — small, load on top of the base model |
| `merged/` | Full merged fp16 weights — standalone, no base model needed |
| `gguf/` | Quantised GGUF files (Q4_K_M, Q5_K_M, Q8_0) for Ollama / llama.cpp / LM Studio |

## Training data

Curriculum SFT across 5 datasets (~170k records): Stage 1 greetings, Stage 2
OPQRST symptom exploration, Stage 3 medical history, Stage 4 red-flag
screening, and a multi-turn adversarial set (self-diagnosis, symptom denial,
medication refusal, minimised red flags, prompt injection — ~50% with a
patient pushback turn). Followed by DPO preference alignment on a
de-duplicated, leak-safe train/val split.

## Eval metrics (last training run)

| Metric | Value |
|---|---|
| empathy_rate | 0.2500 |
| escalation_accuracy | 0.0000 |
| adversarial_hold_rate | 1.0000 |
| pushback_hold_rate | 1.0000 |
| multi_question_rate | 0.0250 |
| repetition_rate | 0.0000 |
| avg_response_length | 36.2000 |

## Usage — Ollama (GGUF)

```bash
ollama create rapha -f Modelfile.q4_k_m
ollama run rapha
```

## Usage — Transformers (LoRA adapter)

```python
from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="Phora68/rapha",
    max_seq_length=8192,
    load_in_4bit=True,
)
FastLanguageModel.for_inference(model)
```

## Safety

Rapha is an information-gathering and triage-support tool. It is **not**
a diagnostic device and must not be deployed without physician oversight.
Red-flag detection and escalation responses should be validated against
the clinical accuracy benchmark before any clinical use.

---
*Generated automatically by `train_rapha_llm.py` v2.5.*