Instructions to use Interact-ai/mindcare with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use Interact-ai/mindcare with PEFT:
from peft import PeftModel from transformers import AutoModelForCausalLM base_model = AutoModelForCausalLM.from_pretrained("unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit") model = PeftModel.from_pretrained(base_model, "Interact-ai/mindcare") - Transformers
How to use Interact-ai/mindcare with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Interact-ai/mindcare") 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("Interact-ai/mindcare", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Interact-ai/mindcare with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Interact-ai/mindcare" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Interact-ai/mindcare", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Interact-ai/mindcare
- SGLang
How to use Interact-ai/mindcare 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 "Interact-ai/mindcare" \ --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": "Interact-ai/mindcare", "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 "Interact-ai/mindcare" \ --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": "Interact-ai/mindcare", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Unsloth Desktop
- Docker Model Runner
How to use Interact-ai/mindcare with Docker Model Runner:
docker model run hf.co/Interact-ai/mindcare
MindCare
MindCare is a parameter-efficient LoRA adapter for
unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit.
It is intended for supportive, empathetic mental-health conversations and
well-being guidance. This repository contains adapter weights, not a complete
standalone base model.
Important: MindCare is not a therapist, doctor, crisis service, or diagnostic tool. Do not use it to diagnose conditions, prescribe treatment, or make emergency decisions. For immediate danger or a mental-health crisis, contact local emergency services or a qualified crisis service.
Intended use
MindCare can be used as a research or application component for:
- supportive conversational interfaces;
- journaling and reflection prompts;
- general well-being and coping suggestions;
- experimenting with LoRA adapters for Llama 3.1.
Responses should be reviewed by a responsible human before being used in health-related products. Applications should provide clear escalation paths to qualified professionals and emergency resources.
Limitations and risks
- The adapter may produce incorrect, incomplete, insensitive, or overconfident responses.
- It may reflect biases in the base model and training data.
- It has not been validated for clinical safety or diagnostic accuracy.
- It may fail to recognize every crisis, self-harm signal, cultural context, or language nuance.
- Do not send private or identifying health information to an untrusted deployment.
Quick start
Install the required packages:
pip install -U transformers peft accelerate bitsandbytes torch
Load the base model and adapter:
import torch
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
base_id = "unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit"
adapter_id = "Interact-ai/mindcare"
tokenizer = AutoTokenizer.from_pretrained(adapter_id)
model = AutoModelForCausalLM.from_pretrained(
base_id,
device_map="auto",
torch_dtype=torch.float16,
)
model = PeftModel.from_pretrained(model, adapter_id)
messages = [
{
"role": "system",
"content": (
"You are a supportive assistant. Be empathetic and clear. "
"Do not diagnose or provide emergency medical advice."
),
},
{
"role": "user",
"content": "I have been feeling overwhelmed lately. Can you help me reflect?",
},
]
inputs = tokenizer.apply_chat_template(
messages,
add_generation_prompt=True,
return_tensors="pt",
).to(model.device)
with torch.no_grad():
output = model.generate(
inputs,
max_new_tokens=256,
do_sample=True,
temperature=0.7,
top_p=0.9,
)
print(tokenizer.decode(output[0][inputs.shape[-1]:], skip_special_tokens=True))
The base model may be gated and may require accepting its license and authenticating with Hugging Face before loading.
Technical details
| Property | Value |
|---|---|
| Base model | unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit |
| Adapter type | LoRA / PEFT |
| Task | Causal language modeling |
LoRA rank (r) |
16 |
| LoRA alpha | 32 |
| LoRA dropout | 0.05 |
| Target modules | q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj |
| PEFT version | 0.21.2 |
The adapter is distributed in safetensors format. The tokenizer and chat
template are included in this repository.
Training and evaluation
This release does not include a public training-dataset card, training hyperparameters, or benchmark results. Do not infer clinical performance from the model name or from anecdotal outputs. Add reproducible dataset, provenance, safety-review, and evaluation information before using the adapter in a production or health-related setting.
License and attribution
The base model is subject to the license and acceptable-use requirements of
unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit.
Users must review and comply with the base model's terms as well as any
requirements applicable to their data and deployment. This adapter should not
be presented as an official Meta or Unsloth model.
Repository contents
adapter_model.safetensors- LoRA adapter weightsadapter_config.json- PEFT adapter configurationtokenizer.jsonandtokenizer_config.json- tokenizer fileschat_template.jinja- Llama 3.1 chat templateREADME.md- model card and usage documentation
Citation
If you use this adapter, cite the repository:
@misc{mindcare_lora,
title = {MindCare: Mental Health LoRA Adapter},
author = {Interact-ai},
year = {2026},
url = {https://huggingface.co/Interact-ai/mindcare}
}
- Downloads last month
- 38
Model tree for Interact-ai/mindcare
Base model
meta-llama/Llama-3.1-8B