empathetic-qwen3-8b-Jan

🤗 Standalone merged model - No base model needed! Load directly with transformers.

An empathetic conversational AI fine-tuned for supportive, understanding responses.

Model Description

This is a merged model (not a LoRA adapter) based on Qwen3-8B, trained on:

• EmpatheticDialogues: Emotional conversation dataset
• ESConv: Emotional support conversations with strategy labels
• GoEmotions: Multi-label emotion classification

Training Details

• Base Model: Qwen3-8B
• Method: Multi-task SFT with auxiliary emotion & strategy heads
• Training: QLoRA with Unsloth optimization, then merged
• Hardware: Kaggle T4 GPU

Usage (Simple - No Unsloth Needed!)

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("Someet24/empathetic-qwen3-8b-Jan", torch_dtype="auto", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("Someet24/empathetic-qwen3-8b-Jan")

# Generate response
messages = [
    {"role": "system", "content": "You are an empathetic, supportive friend."},
    {"role": "user", "content": "I'm feeling really anxious about tomorrow."}
]

text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to("cuda")

outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.7, do_sample=True)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Deploy with VLLM (Recommended for Production)

pip install vllm
python -m vllm.entrypoints.openai.api_server --model Someet24/empathetic-qwen3-8b-Jan --dtype float16

Intended Use

• Emotional support conversations
• Mental wellness chatbots
• Empathetic dialogue systems

Limitations

• Not a replacement for professional mental health support
• May not handle crisis situations appropriately
• English only

License

Apache 2.0