MindCare
an adapter for empathetic, supportive conversations
2 min readupdated
MindCare is a LoRA (a lightweight add-on for a base model) built on unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bit for empathetic conversation, self-reflection, and general well-being tips. It doesn't replace a doctor or therapist, but helps conversational interfaces maintain a gentle, supportive tone. Built on the 8B version of Llama 3.1, the adapter ships in safetensors format with a ready-to-use chat template, making it a solid fit for developers building mental well-being chatbots and researchers studying text generation.
Key features
- Built on an efficient 4-bit version of Llama 3.1 8B.
- Fine-tuned for gentle, empathetic, and supportive conversations.
- Uses PEFT with a LoRA adapter configured at rank r=16 and alpha=32.
- Bundles a configured tokenizer and a
chat_template.jinjachat template.
What you can do
- Integrate the adapter into an experimental chatbot for emotional support.
- Build an assistant for daily self-reflection and journaling.
- Use it in research projects exploring tone and sentiment in LLM conversations.
Example
Here is how to load the base model and the LoRA adapter in Python:
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)
Read more
DetailsBase model:
unsloth/Meta-Llama-3.1-8B-Instruct-bnb-4bitArchitecture type: LoRA / PEFT (targeting modules q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj)
PEFT version: 0.21.2
Limitations: The model is not clinically validated and does not replace professional help in crisis situations.










Comments