Instructions to use Ronin48LLC/bones-lora-adapter with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Ronin48LLC/bones-lora-adapter with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-3.3-70B-Instruct")
model = PeftModel.from_pretrained(base_model, "Ronin48LLC/bones-lora-adapter")

Transformers

How to use Ronin48LLC/bones-lora-adapter with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Ronin48LLC/bones-lora-adapter")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Ronin48LLC/bones-lora-adapter", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Ronin48LLC/bones-lora-adapter with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Ronin48LLC/bones-lora-adapter"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Ronin48LLC/bones-lora-adapter",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Ronin48LLC/bones-lora-adapter

SGLang

How to use Ronin48LLC/bones-lora-adapter with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Ronin48LLC/bones-lora-adapter" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Ronin48LLC/bones-lora-adapter",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Ronin48LLC/bones-lora-adapter" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Ronin48LLC/bones-lora-adapter",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Ronin48LLC/bones-lora-adapter with Docker Model Runner:
```
docker model run hf.co/Ronin48LLC/bones-lora-adapter
```

BONES — Biomedical On-scene Navigator for Emergency Services

"He's dead, Jim." — Dr. Leonard H. McCoy, USS Enterprise

An open-source LoRA adapter for EMS clinical decision support.

BONES is a QLoRA adapter on top of Meta Llama 3.3 70B Instruct, fine-tuned on emergency medical services protocols, pharmacology references, triage frameworks, and clinical decision-support knowledge — built to assist EMRs, EMTs, AEMTs, and Paramedics in the field and in training. Model #4 of the Ronin 48 first-responder suite, alongside BRUNO (fire) and SELMA (law enforcement).

Model Details

Field	Value
Base Model	meta-llama/Llama-3.3-70B-Instruct
Adapter Type	LoRA (QLoRA)
LoRA Rank / Alpha / Dropout	64 / 128 / 0.05
Target Modules	`q_proj`, `k_proj`, `v_proj`, `o_proj`, `gate_proj`, `up_proj`, `down_proj`
Task	EMS protocol lookup, drug reference, triage and differential support
Training Method	QLoRA supervised fine-tuning via `SFTTrainer` (trl)
License	MIT (adapter weights)
Maintainer	Ronin48LLC

Capabilities

Protocol lookup — AHA ACLS/PALS/BLS algorithms, NWC protocols, NREMT scope of practice
Drug reference — EMS formulary, weight-based dosing, contraindications, interactions
Triage support — START/SALT/JumpSTART, mass-casualty incident guidance
Differential support — symptom-to-differential reasoning for field assessment
Trauma guidance — hemorrhage control, spinal precautions, burn classification
OB/Peds — childbirth emergencies, pediatric dosing (Broselow), neonatal resuscitation
Toxicology — overdose recognition, antidote references, decontamination
Documentation — PCR narrative generation, patient-assessment templates

Intended Use

For certified EMS personnel operating under medical direction:

Protocol lookup and reference during training and non-emergency preparation
Differential considerations for patient presentations
Drug reference and dosing ranges — verified against your formulary before use
Triage category guidance (START/SALT)
PCR documentation assistance
Training and scenario-based learning

BONES is a clinical decision support tool, not a replacement for medical direction. When in doubt, call medical control.

Limitations

Read LIMITATIONS.md before deploying in any operational context. In short:

BONES is not a physician; its outputs are not medical advice. All clinical decisions are made by a licensed provider under medical direction.
It does not replace regional protocols, offline medical direction, or base-hospital contact requirements.
It cannot assess the patient — it works from what you tell it.
It answers across the full EMR–Paramedic spectrum; you are responsible for operating within your certified scope.
It is not an FDA-cleared medical device and has not undergone clinical validation.
It has a training-data cutoff; AHA guideline updates and protocol changes after it may not be reflected.
Drug dosing must always be verified — pediatric weight-based dosing, maximum doses, and contraindications — against your agency formulary before administration.

Usage

This is a PEFT LoRA adapter. Load it on top of the base model:

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base_model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.3-70B-Instruct",
    load_in_4bit=True,
    device_map="auto",
)
model = PeftModel.from_pretrained(base_model, "Ronin48LLC/bones-lora-adapter")
tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-3.3-70B-Instruct")

Note: Access to the base model requires accepting Meta's license on Hugging Face.

The base model with the BONES system prompt and no fine-tuning is a workable baseline; see the source repository's Quick Start.

Training

Framework: PyTorch + Transformers + TRL + PEFT, 4-bit quantized base
Adapter config: rank 64, alpha 128, dropout 0.05, all seven attention and MLP projections (see adapter_config.json)
Run details: training_args.bin is preserved in this repository; the step count and loss curve were not recorded in the upload

Training Data Sources

Source	Description	License
AHA ACLS/PALS/BLS Guidelines	Cardiac arrest and resuscitation algorithms	Public guidelines
NAEMSP / NASEMSO Protocols	National EMS protocols and scope of practice	Public
OpenMedSpel / MedQA	Medical Q&A datasets	Open
PubMed Central	EMS and emergency-medicine literature	Open Access
NREMT Scope of Practice	EMR/EMT/AEMT/Paramedic scope tables	Public
Synthetic Scenarios	AI-generated dispatch-to-treatment examples	Proprietary

Related Models — the Ronin 48 first-responder suite

BONES, BRUNO, and SELMA share scenes constantly — consult the appropriate model for each domain.

Model	Domain	Use when…
BONES (this adapter)	EMS — EMR / EMT / AEMT / Paramedic	Patient assessment, treatment protocols, drug dosing, triage, transport
BRUNO	Fire service — company officer / IC	Fireground tactics, size-up, hazmat, extrication, water supply, ICS
SELMA	Law enforcement	Criminal statute identification, charge elements, constitutional flags

Shared scene	Primary	Support
Structure fire with casualties	BRUNO (fireground ops)	BONES (patient care)
Vehicle accident with entrapment	BRUNO (extrication)	BONES (care during extrication)
Hazmat with patient exposures	BRUNO (mitigation, decon zone)	BONES (patient decon and treatment)
Mass casualty incident	BONES (triage, treatment)	BRUNO (ICS, sectors) + SELMA (criminal nexus)
Overdose call	BONES (patient care, naloxone)	SELMA (distribution charges if applicable)
Cardiac arrest in a burning structure	BRUNO (scene safety, egress)	BONES (resuscitation protocol)

ABBY (forensics) operates independently of the first-responder suite; SELMA pairs with ATTICUS on the legal side.

Source

github.com/CryptoJones/BONES — data collection, QLoRA pipeline, synthetic-scenario generator.

License

Downloads last month: 19

Model tree for Ronin48LLC/bones-lora-adapter

Base model

meta-llama/Llama-3.1-70B

Finetuned

meta-llama/Llama-3.3-70B-Instruct

Adapter

(443)

this model