GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
arnabwithab
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
arnabwithab
Model Tree
Qwen/Qwen3-0.6B
Input Modalities
Output Modalities
Supported Functionality
Psychotherapy-LLM/PsyCoPref).<think> traces despite the no-think header — strip them before scoring.python
from transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-0.6B", device_map="auto")tok = AutoTokenizer.from_pretrained("arnabwithab/ai-therapist-qwen3-0.6b-dpo-lora")model = PeftModel.from_pretrained(base, "arnabwithab/ai-therapist-qwen3-0.6b-dpo-lora") msgs = [{"role": "user", "content": "I've been feeling anxious about work and can't sleep."}]prompt = tok.apply_chat_template(msgs, tokenize=False, add_generation_prompt=True)inputs = tok(prompt, return_tensors="pt").to(model.device)out = model.generate(**inputs, max_new_tokens=512, temperature=0.7, top_p=0.8)print(tok.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))vLLM:
bash
vllm serve Qwen/Qwen3-0.6B --enable-lora \ --lora-modules ai-therapist-qwen3-0.6b-dpo-lora=arnabwithab/ai-therapist-qwen3-0.6b-dpo-lora \ --served-model-name ai-therapist-qwen3-0.6b-dpo-lora --reasoning-parser qwen3Not a therapist — output is not to be treated as verifiable legal advice.
Code + eval harness: https://github.com/arnabwithab/ai-therapist
vllm serve Qwen/Qwen3-0.6B --enable-lora \ --lora-modules ai-therapist-qwen3-0.6b-dpo-lora=arnabwithab/ai-therapist-qwen3-0.6b-dpo-lora \ --served-model-name ai-therapist-qwen3-0.6b-dpo-lora --reasoning-parser qwen3