New and Improved reasoning traces. Better ui generation. Smarter decisions. Better code generation! Trained on a 700+ dataset.
USE BUDGET FORCING (putting the word answer or think at the end of the assistant generation to keep generationg more thinking and use 'answer' to write code.)
SFT on 1 x H100 for 1 hour.
1<|im_start|>user
2{question}<|im_end|>
3<|im_start|>assistant
4<|im_start|>think
5{reasoning}<|im_end|>
6<|im_start|>answer
1from transformers import AutoModelForCausalLM, AutoTokenizer
2
3model_name = "qingy2024/UIGEN-T1.1-Qwen-32B"
4tokenizer = AutoTokenizer.from_pretrained(model_name)
5model = AutoModelForCausalLM.from_pretrained(model_name).to("cuda")
6
7prompt = """<|im_start|>user
8Make a dark-themed dashboard for an oil rig.<|im_end|>
9<|im_start|>assistant
10<|im_start|>think
11"""
12
13inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
14outputs = model.generate(**inputs, max_new_tokens=12012, do_sample=True, temperature=0.7) #max tokens has to be greater than 12k
15
16print(tokenizer.decode(outputs[0], skip_special_tokens=True))
1@misc{smirki_UIGEN-T1.1,
2 title={UIGEN-T1.1.1: Chain-of-Thought UI Generation Model},
3 author={smirki},
4 year={2025},
5 publisher={Hugging Face},
6 url={https://huggingface.co/smirki/UIGEN-T1.11}
7}
