Instructions to use admin-lima/phi35-saude-amazonia-lora with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use admin-lima/phi35-saude-amazonia-lora with Transformers:

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

pipe = pipeline("text-generation", model="admin-lima/phi35-saude-amazonia-lora")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("admin-lima/phi35-saude-amazonia-lora", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use admin-lima/phi35-saude-amazonia-lora with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "admin-lima/phi35-saude-amazonia-lora"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "admin-lima/phi35-saude-amazonia-lora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/admin-lima/phi35-saude-amazonia-lora

SGLang

How to use admin-lima/phi35-saude-amazonia-lora 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 "admin-lima/phi35-saude-amazonia-lora" \
    --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": "admin-lima/phi35-saude-amazonia-lora",
		"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 "admin-lima/phi35-saude-amazonia-lora" \
        --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": "admin-lima/phi35-saude-amazonia-lora",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use admin-lima/phi35-saude-amazonia-lora with Docker Model Runner:
```
docker model run hf.co/admin-lima/phi35-saude-amazonia-lora
```

phi35-saude-amazonia-lora / README.md

rogetxtapai

Update README.md

c5b01e2 verified 20 days ago

preview code

Raw

History Blame

1.52 kB

	---
	language:
	- pt
	library_name: peft
	base_model: microsoft/Phi-3.5-mini-instruct
	license: mit
	tags:
	- health
	- clinical
	- amazonia
	- qlora
	- rag
	- sus
	- portuguese
	- offline
	---

	# Phi-3.5-mini: Assistente Clínico Offline da Amazônia

	Adapter LoRA (QLoRA 4-bit) sobre `microsoft/Phi-3.5-mini-instruct` (3,8B), fine-tunado em 1.000 diálogos clínicos sintéticos em português brasileiro.

	## Configuração do treino

	\| Parâmetro \| Valor \|
	\|---\|---\|
	\| Modelo base \| microsoft/Phi-3.5-mini-instruct (3,8B) \|
	\| Quantização \| 4-bit NF4 (QLoRA) \|
	\| LoRA targets \| qkv_proj, o_proj \|
	\| LoRA r / alpha \| 16 / 32 \|
	\| Épocas \| 3 \|
	\| Learning rate \| 2e-4 (cosine) \|
	\| Dataset \| admin-lima/sllm-amazonia-saude-sft (1000 exemplos) \|

	## Resultados (hold-out n=33)

	\| Métrica \| Valor \|
	\|---\|---\|
	\| TRR (recusa correta) \| 0.929 \|
	\| FRR (falsa recusa) \| 0.266 \|
	\| F1 (recusa) \| 0.813 \|
	\| Dosage-Hallucination \| 0.03 \|

	## Uso

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

	base = "microsoft/Phi-3.5-mini-instruct"
	adapter = "admin-lima/phi35-saude-amazonia-lora"

	tok = AutoTokenizer.from_pretrained(base, trust_remote_code=True)
	model = AutoModelForCausalLM.from_pretrained(base, torch_dtype="float16",
	trust_remote_code=True, device_map="auto")
	model = PeftModel.from_pretrained(model, adapter)
	```

	## Aviso

	O sistema apoia a decisão do profissional; não a substitui.