| import torch | |
| import os | |
| from Ming_Uni.MingUniInference import Ming_Uni_Inference | |
| from Ming_Uni.process import MyProcessor | |
| device = torch.cuda.current_device() | |
| device = torch.device(device) | |
| model_path='../Ming-Lite-Uni/' | |
| model = Ming_Uni_Inference(model_path) | |
| model.to(torch.bfloat16) | |
| model.to(device) | |
| model.eval() | |
| llm_model=os.path.join(model_path, 'qwen2_5_llm') | |
| my_proc=MyProcessor(llm_model) | |
| image_file = "tests/cake.jpg" | |
| prompt = "add a candle on top of the cake" | |
| inputs = my_proc.process(image_file=image_file, prompt=prompt, device=device) | |
| result = model.image_gen_generate(inputs, steps=30, seed=42, cfg=5.0, height=512, width=512)[1] | |
| result.save("result.png") |