GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
mahdi2020
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
mahdi2020
Model Tree
Qwen/Qwen2.5-1.5B-Instruct
Input Modalities
Output Modalities
Supported Functionality
This adapter is designed to convert raw, natural language customer support queries into structured action intents (such as order cancellations, shipping modifications, or account closures). It is optimal for integration with backend databases, webhooks, and automated customer service routing workflows.
q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj (1.18% of total model parameters)You can load this model along with its base weights using peft and transformers as follows:
python
import torchimport jsonfrom transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfigfrom peft import PeftModel MODEL_ID = "Qwen/Qwen2.5-1.5B-Instruct"ADAPTER_ID = "mahdi2020/qwen2.5-1.5B-ecommerce-lora-fine-tuned" # Configure 4-bit quantization for resource efficiencybnb_config = BitsAndBytesConfig( load_in_4bit=True, bnb_4bit_use_double_quant=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16,) # Load Tokenizertokenizer = AutoTokenizer.from_pretrained(ADAPTER_ID) # Load Base Model + PEFT Adapterbase_model = AutoModelForCausalLM.from_pretrained( MODEL_ID, quantization_config=bnb_config, device_map="auto", trust_remote_code=True,)model = PeftModel.from_pretrained(base_model, ADAPTER_ID)model.eval() def get_action(user_message: str): prompt = ( f"<|im_start|>user\n{user_message}<|im_end|>\n" f"<|im_start|>assistant\n" ) inputs = tokenizer(prompt, return_tensors="pt").to(model.device) with torch.no_grad(): output_ids = model.generate( **inputs, max_new_tokens=128, do_sample=False, pad_token_id=tokenizer.eos_token_id, eos_token_id=tokenizer.convert_tokens_to_ids("<|im_end|>"), ) new_tokens = output_ids[0][inputs["input_ids"].shape[1]:] raw_output = tokenizer.decode(new_tokens, skip_special_tokens=True).strip() return json.loads(raw_output) # Example executionprint(get_action("I want to cancel my order #789456"))# Output:# {'intent': 'cancel_order', 'order_id': '789456'}import torchimport jsonfrom transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfigfrom peft import PeftModel MODEL_ID = "Qwen/Qwen2.5-1.5B-Instruct"ADAPTER_ID = "mahdi2020/qwen2.5-1.5B-ecommerce-lora-fine-tuned" # Configure 4-bit quantization for resource efficiencybnb_config = BitsAndBytesConfig( load_in_4bit=True, bnb_4bit_use_double_quant=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16,) # Load Tokenizertokenizer = AutoTokenizer.from_pretrained(ADAPTER_ID) # Load Base Model + PEFT Adapterbase_model = AutoModelForCausalLM.from_pretrained( MODEL_ID, quantization_config=bnb_config, device_map="auto", trust_remote_code=True,)model = PeftModel.from_pretrained(base_model, ADAPTER_ID)model.eval() def get_action(user_message: str): prompt = ( f"<|im_start|>user\n{user_message}<|im_end|>\n" f"<|im_start|>assistant\n" ) inputs = tokenizer(prompt, return_tensors="pt").to(model.device) with torch.no_grad(): output_ids = model.generate( **inputs, max_new_tokens=128, do_sample=False, pad_token_id=tokenizer.eos_token_id, eos_token_id=tokenizer.convert_tokens_to_ids("<|im_end|>"), ) new_tokens = output_ids[0][inputs["input_ids"].shape[1]:] raw_output = tokenizer.decode(new_tokens, skip_special_tokens=True).strip() return json.loads(raw_output) # Example executionprint(get_action("I want to cancel my order #789456"))# Output:# {'intent': 'cancel_order', 'order_id': '789456'}