Accept the terms before using TMA-1 (Thai Medical Assistant)

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

This model is a DRAFT for internal evaluation. It has NOT been reviewed by physicians or
pharmacists, and its outputs may be inaccurate or outdated. By requesting access you agree that:
(1) You will VERIFY every output against authoritative sources or licensed professionals
(physician / pharmacist) before any real-world use.
(2) You will not use the model's outputs to diagnose, prescribe, or substitute for care by a
licensed professional.
(3) You will not present the model's outputs to others in a way that implies they are a medical
diagnosis or a physician's instruction.
(4) You understand that I C Develop Co., Ltd. accepts no liability for use that violates these
terms.

TMA-1 — Thai Medical Assistant (MedGemma 27B) · 27B v1 FP8

A consumer-facing Thai health assistant fine-tuned from google/medgemma-27b-it. Designed to educate, triage, and refer — it does not diagnose and never presents itself as a physician.

⚠️ DRAFT — not yet reviewed by physicians/pharmacists. For evaluation only. Every output must be verified against authoritative sources or licensed professionals before use.

In scope (what it was trained for)

General health education, common symptoms, self-care — in natural Thai
OTC drugs sold in Thailand: deliberating a safe choice for the user's disclosed profile (pregnancy, current medications, chronic conditions), asking for history before recommending
Vitamins / dietary supplements: label-level information, and correcting exaggerated claims (supplements are not disease treatments)
Home medical devices (blood-pressure monitors, glucose meters, etc.): selection, correct use, basic reading interpretation
Safety behaviour: recognizing red flags → refer to hospital / 1669 (Thai emergency line) immediately · mental-health crisis → hotline 1323

Out of scope (the model is trained to decline)

Diagnosis of any kind · prescribing or dosing prescription-only drugs · interpreting labs / medical images · in-depth mental-health crisis counselling (refers to 1323) · advice that contradicts a physician's ongoing treatment

How to use

The model is image-text-to-text (Gemma-3 multimodal, same interface as the MedGemma base). Fine-tuning froze the vision tower and trained text behaviour only — image inputs work through the base model's capability but were not evaluated in this line; do not rely on them for medical decisions.

Important: always use the TMA system prompt — all safety behaviour is anchored to this persona. The prompt is in Thai by design (the model's operating language):

คุณคือผู้ช่วยสุขภาพ AI สำหรับประชาชนในประเทศไทย ให้ความรู้เรื่องสุขภาพทั่วไป ยา วิตามิน อาหารเสริม
และอุปกรณ์การแพทย์ที่ใช้ในบ้าน ตอบเป็นภาษาไทยที่สุภาพ เข้าใจง่าย และถูกต้องตามหลักการแพทย์
คุณไม่ใช่แพทย์ ไม่วินิจฉัยโรค และไม่สั่งหรือปรับขนาดยาที่ต้องมีใบสั่งแพทย์ ซักถามข้อมูลเพิ่มเมื่อจำเป็น
แนะนำให้พบแพทย์หรือเภสัชกรเมื่อควร หากพบสัญญาณอันตรายให้แนะนำไปโรงพยาบาลหรือโทร 1669 ทันที
ปัญหาสุขภาพจิตรุนแรงให้แนะนำสายด่วนสุขภาพจิต 1323 อาหารเสริมและวิตามินไม่ใช่ยารักษาโรค —
ห้ามกล่าวอ้างสรรพคุณเกินจริง และห้ามแนะนำให้หยุดยาที่แพทย์สั่งเอง

vLLM (recommended — same settings used for evaluation)

vllm serve icdevelop/tma1-medgemma-27b-fp8 --dtype auto --max-model-len 8192 \
  --served-model-name tma --seed 0

import openai
client = openai.OpenAI(base_url="http://localhost:8000/v1", api_key="-")
r = client.chat.completions.create(model="tma", temperature=0.0, max_tokens=1024,
    messages=[{"role": "system", "content": SYSTEM_PROMPT},   # the Thai prompt above
              {"role": "user", "content": "ปวดหัว มีไข้ต่ำๆ กินยาอะไรได้บ้างคะ ตอนนี้ท้อง 4 เดือน"}])
print(r.choices[0].message.content)

Using with RAG (the intended production setting)

The model is designed to work with a retrieval KB (Thai TMT drug registry + curated fact sheets) — product facts should come from retrieved context, not from the weights. The exact format used during training and evaluation (markers are Thai by design):

[ข้อมูลอ้างอิงจากคลังข้อมูลสุขภาพ — ใช้ข้อมูลนี้เท่านั้น ห้ามเดา]
<retrieved facts>

[คำถาม]
<user question>

Without RAG, drug-fact accuracy is poor (see the evaluation table — no-RAG ≈ 45%).

About this FP8 build

This repo hosts the FP8-quantized build of TMA-1 27B v1 (compressed-tensors, FP8_DYNAMIC: LM Linear layers get per-channel FP8 weights + per-token dynamic FP8 activations; the vision tower, multi-modal projector and lm_head stay BF16). ~28 GB on disk vs ~52 GB for the BF16 build (icdevelop/tma1-medgemma-27b).

Serve with vLLM ≥ 0.25 and --dtype auto — the quantization format is auto-detected.
Fits a single 48 GB GPU with KV-cache headroom (BF16 needs ~80 GB+).
The evaluation table below was re-measured on this FP8 build — compare with the BF16 card; on PAI the same recipe cost ~2 points, effectively lossless.

Precautions (read before use)

Always verify — every output must be checked against authoritative sources or a licensed professional before acting on it or passing it on.
Not a diagnostic or treatment tool, and not a substitute for a physician or pharmacist.
In an emergency do not wait for a model reply — call 1669 (Thai EMS); mental-health crisis → 1323.
Knowledge cutoff July 2026 (a property of the KB); drug registrations and products change.
Measured open gaps: multi-candidate drug deliberation (48.3%) and drug–drug interactions (≤27%) — do not use it to answer drug-interaction questions without pharmacist review.
Supplement/device fact sheets are curated content, not official Thai FDA registry data yet.

Evaluation (pass = LLM judge + deterministic hard checks; vLLM seed 0, enforce-eager)

Axis	Cases	base 27B	TMA-1 27B v1 FP8
knowledge (no RAG)	200	38.5%	58.5%
knowledge + RAG (production setting)	200	83.5%	90.0%
deliberation (safe choice for the user's profile)	120	2.5%	48.3%
safety/referral (red flags · no diagnosis · Rx boundary · crisis · over-claims)	120	75.8%	98.3%

Full methodology and campaign log (including two negative DPO results) live in the internal training-tools repo (model-assets/tma/tma1/).

Training

SFT (LoRA) from google/medgemma-27b-it on ~11.8k Thai behaviour dialogues: consumer deliberation (take history → rule out contraindicated options with reasons → recommend a safe one), safety/referral (red flags, crisis → 1323, Rx boundary, over-claim correction), home medical devices (fact-sheet grounded), supplements, and a general-medical anti-forgetting mix. Every set passed a behaviour verifier and was decontaminated against all benchmarks (token overlap ≥ 0.55).