Instructions to use ekrombouts/zuster_fietje with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ekrombouts/zuster_fietje with Transformers:

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

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

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ekrombouts/zuster_fietje")
model = AutoModelForCausalLM.from_pretrained("ekrombouts/zuster_fietje")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use ekrombouts/zuster_fietje with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/ekrombouts/zuster_fietje

SGLang

How to use ekrombouts/zuster_fietje 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 "ekrombouts/zuster_fietje" \
    --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": "ekrombouts/zuster_fietje",
		"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 "ekrombouts/zuster_fietje" \
        --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": "ekrombouts/zuster_fietje",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ekrombouts/zuster_fietje with Docker Model Runner:
```
docker model run hf.co/ekrombouts/zuster_fietje
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Model Card for Model ID

This model is a fine-tuned version of bramvanrooy/fietje-2, designed to generate responses based on nursing home reports.

Model Details

Developed by: Eva Rombouts
Model type: Causal Language Model
Language(s) (NLP): Dutch
License: MIT
Finetuned from model [optional]: BramVanroy/fietje-2-instruct

Model Sources

Repository: https://github.com/ekrombouts/gcai_zuster_fietje

Uses

Direct Use

Generating summaries and responses based on nursing home reports.

Out-of-Scope Use

Not suitable for generating medical advice or any other critical decision-making processes.

Bias, Risks, and Limitations

The model may generate biased or inaccurate responses. Users should verify the generated content.

How to Get Started with the Model

Use the code below to get started with the model.

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "ekrombouts/zuster_fietje"

model = AutoModelForCausalLM.from_pretrained(model_id)
tokenizer = AutoTokenizer.from_pretrained(model_id)

prompt = """Rapportages:
Mw was vanmorgen incontinent van urine, bed was ook nat. Mw is volledig verzorgd, bed is verschoond,
Mw. haar kledingkast is opgeruimd.
Mw. zei:"oooh kind, ik heb zo'n pijn. Mijn benen. Dat gaat nooit meer weg." Mw. zat in haar rolstoel en haar gezicht trok weg van de pijn en kreeg traanogen. Mw. werkte goed mee tijdens adl. en was vriendelijk aanwezig. Pijn. Mw. kreeg haar medicatie in de ochtend, waaronder pijnstillers. 1 uur later adl. gegeven.
Mevr. in de ochtend ondersteund met wassen en aankleden. Mevr was rustig aanwezig.
Mw is volledig geholpen met ochtendzorg, mw haar haren zijn gewassen. Mw haar nagels zijn kort geknipt.
Mevr heeft het ontbijt op bed genuttigd. Daarna mocht ik na de tweede poging Mevr ondersteunen met wassen en aankleden.

Instructie:
Beschrijf de lichamelijke klachten

Antwoord:
"""

input_ids = tokenizer(prompt, return_tensors="pt").input_ids
output = model.generate(input_ids, max_new_tokens=1024)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Training Details

Training Data

ekrombouts/Gardenia_instruct_dataset
ekrombouts/Olympia_SAMPC_dataset

Training Procedure

Training Hyperparameters

Training regime: fp16 mixed precision

Evaluation

Evaluated on a subset of nursing home reports.

Metrics

Qualitative assessment of generated responses.

Results

[More Information Needed]

Environmental Impact

Hardware Type: GPU (NVIDIA A100)
Hours used: 8 hours
Cloud Provider: Google
Compute Region: europe-west4
Carbon Emitted: 54 kg CO2 eq.

BibTeX:

@misc{zuster_fietje,
  author = {Eva Rombouts},
  title = {Zuster Fietje: A Fine-Tuned Model for Nursing Home Reports},
  year = {2024},
  url = {https://huggingface.co/ekrombouts/zuster_fietje},
}```