GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Erevanox
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
Erevanox
Model Tree
Qwen/Qwen2.5-1.5B-Instruct
Input Modalities
Output Modalities
Supported Functionality
This repository provides an open-weight LoRA adapter trained on Qwen/Qwen2.5-1.5B-Instruct specifically fine-tuned to act as a grounded outdoor activity advisor. It evaluates live meteorological, air quality, and OpenStreetMap geospatial evidence to recommend when and where to engage in outdoor activities, identify safety precautions, and state when conditions are hazardous.
schemas/recommendation.schema.json.missing_information_identified.The model is explicitly trained on structured profiles for:
Qwen/Qwen2.5-1.5B-Instruct["q_proj", "v_proj"]python
import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base_model_id = "Qwen/Qwen2.5-1.5B-Instruct"adapter_id = "YOUR_HF_USERNAME/geoai-qwen2.5-1.5b-outdoor-lora" # 1. Load Tokenizertokenizer = AutoTokenizer.from_pretrained(base_model_id) # 2. Load Base Modelbase_model = AutoModelForCausalLM.from_pretrained( base_model_id, torch_dtype=torch.float32, device_map="auto") # 3. Attach LoRA Adaptermodel = PeftModel.from_pretrained(base_model, adapter_id)model.eval() # 4. Formulate Promptmessages = [ {"role": "system", "content": "You are a grounded GeoAI outdoor safety assistant..."}, {"role": "user", "content": "Activity: Running | Location: Marine Drive | Temp: 34C | UV: 11.2"}]formatted = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer(formatted, return_tensors="pt") with torch.no_grad(): outputs = model.generate(**inputs, max_new_tokens=400, do_sample=False) response = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)print(response)app/safety/rules.py) that strictly enforce WHO UV ≥11, EPA AQI >300, and NWS lightning rules.import torchfrom transformers import AutoModelForCausalLM, AutoTokenizerfrom peft import PeftModel base_model_id = "Qwen/Qwen2.5-1.5B-Instruct"adapter_id = "YOUR_HF_USERNAME/geoai-qwen2.5-1.5b-outdoor-lora" # 1. Load Tokenizertokenizer = AutoTokenizer.from_pretrained(base_model_id) # 2. Load Base Modelbase_model = AutoModelForCausalLM.from_pretrained( base_model_id, torch_dtype=torch.float32, device_map="auto") # 3. Attach LoRA Adaptermodel = PeftModel.from_pretrained(base_model, adapter_id)model.eval() # 4. Formulate Promptmessages = [ {"role": "system", "content": "You are a grounded GeoAI outdoor safety assistant..."}, {"role": "user", "content": "Activity: Running | Location: Marine Drive | Temp: 34C | UV: 11.2"}]formatted = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)inputs = tokenizer(formatted, return_tensors="pt") with torch.no_grad(): outputs = model.generate(**inputs, max_new_tokens=400, do_sample=False) response = tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True)print(response)