Someet24/empathetic-qwen3-8b-Jan
The Someet24/empathetic-qwen3-8b-Jan is an 8 billion parameter Qwen3-based conversational AI model, fine-tuned for generating empathetic and supportive responses. It leverages multi-task supervised fine-tuning on datasets like EmpatheticDialogues and ESConv to excel in emotional support conversations and mental wellness chatbot applications. This merged model offers a 32768 token context length and is designed for direct loading without a base model.
Loading preview...
Empathetic Qwen3-8B for Supportive Conversations
This model, empathetic-qwen3-8b-Jan by Someet24, is an 8 billion parameter Qwen3-based language model specifically fine-tuned to generate empathetic and supportive conversational responses. It is a standalone merged model, meaning it can be loaded directly using the transformers library without requiring a separate base model.
Key Capabilities & Training:
- Empathetic Dialogue Generation: Optimized for understanding and responding with empathy, making it suitable for emotional support contexts.
- Multi-task Supervised Fine-tuning (SFT): Trained on a combination of specialized datasets including:
- EmpatheticDialogues: Focused on emotional conversations.
- ESConv: Incorporates emotional support conversations with strategic labeling.
- GoEmotions: Utilized for multi-label emotion classification.
- Efficient Training: Employed QLoRA with Unsloth optimization on a Kaggle T4 GPU, followed by merging the adapter into the base model.
- Context Length: Supports a context window of 32768 tokens.
Intended Use Cases:
- Emotional Support Chatbots: Ideal for applications requiring understanding and supportive dialogue.
- Mental Wellness Applications: Can be integrated into tools designed to assist with mental well-being.
- Empathetic Dialogue Systems: Useful for creating AI agents that can engage in more human-like, understanding conversations.
Limitations:
It is important to note that this model is not a substitute for professional mental health support and may not be appropriate for handling crisis situations. It is currently limited to English language interactions.