Thera-Space: Llama-3-8B for Psycho-Counseling

Model Description and Motivation

This repository hosts Thera-Space, a specialized variant of Llama-3-8B, fine-tuned to act as a conversational therapist. The goal is to provide empathetic, supportive, and reflective responses in a counseling-style context.

The base model, PsychoCounsel-Llama3-8B, was already aligned using a preference dataset grounded in professional psycho-counseling principles, achieving strong results against GPT-4o. On top of this, we fine-tuned it further with additional high-quality counseling conversations to strengthen empathy, consistency, and therapeutic tone.

This model is intended for experimental and role-playing purposes only — not as a substitute for professional mental health care.

Available Formats

  • Full Model: suitable for fine-tuning or research.
  • Quantized GGUF (q4_k_m): optimized for efficient inference on local devices.

Training Data

Thera-Space was fine-tuned using a curated mix of publicly available mental health and counseling datasets, including:

These datasets helped shape the model’s empathetic, reflective, and therapist-like conversational style.

Use Cases

  • Role-playing therapy-style conversations.
  • Exploring empathetic conversational AI.
  • Research in counseling-aligned LLMs.

⚠️ Disclaimer: This model is for research and role-play only. It is not a replacement for professional therapy.

Downloads last month
93
Safetensors
Model size
8B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for omar-salama/PsychoCounsel-Theraspace-Llama3-8B

Quantized
(4)
this model
Quantizations
2 models

Datasets used to train omar-salama/PsychoCounsel-Theraspace-Llama3-8B