| 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") |
