|
Download docs/zh/Research_Tutorial/inference_time_scaling.md from ymyy307/diffsynth: direct link, hf CLI and curl.
- Browser
- Download file 7.89 kB
-
https://huggingface.co/ymyy307/diffsynth/resolve/main/docs/zh/Research_Tutorial/inference_time_scaling.md
- Command line
-
hf download hf://ymyy307/diffsynth/docs/zh/Research_Tutorial/inference_time_scaling.md
-
curl -L -o inference_time_scaling.md https://huggingface.co/ymyy307/diffsynth/resolve/main/docs/zh/Research_Tutorial/inference_time_scaling.md
7.89 kB
| # 推理改进优化技术 | |
| DiffSynth-Studio 旨在以基础框架驱动技术创新。本文以 Inference-time scaling 为例,展示如何基于 DiffSynth-Studio 构建免训练(Training-free)的图像生成增强方案。 | |
| Notebook: https://github.com/modelscope/DiffSynth-Studio/blob/main/docs/zh/Research_Tutorial/inference_time_scaling.ipynb | |
| ## 1. 图像质量量化 | |
| 首先,我们需要找到一个指标来量化图像生成模型生成的图像质量。最简单直接的方案是人工打分,但这样做的成本太高,无法大规模使用。不过,收集人工打分后,训练一个图像分类模型来预测人类的打分结果,是完全可行的。PickScore [[1]](https://arxiv.org/abs/2305.01569) 就是这样一个模型,运行下面的代码,将会自动下载并加载 [PickScore 模型](https://modelscope.cn/models/AI-ModelScope/PickScore_v1)。 | |
| ```python | |
| from modelscope import AutoProcessor, AutoModel | |
| import torch | |
| class PickScore(torch.nn.Module): | |
| def __init__(self): | |
| super().__init__() | |
| self.processor = AutoProcessor.from_pretrained("laion/CLIP-ViT-H-14-laion2B-s32B-b79K") | |
| self.model = AutoModel.from_pretrained("AI-ModelScope/PickScore_v1").eval().to("cuda") | |
| def forward(self, image, prompt): | |
| image_inputs = self.processor(images=image, padding=True, truncation=True, max_length=77, return_tensors="pt").to("cuda") | |
| text_inputs = self.processor(text=prompt, padding=True, truncation=True, max_length=77, return_tensors="pt").to("cuda") | |
| with torch.inference_mode(): | |
| image_embs = self.model.get_image_features(**image_inputs).pooler_output | |
| image_embs = image_embs / torch.norm(image_embs, dim=-1, keepdim=True) | |
| text_embs = self.model.get_text_features(**text_inputs).pooler_output | |
| text_embs = text_embs / torch.norm(text_embs, dim=-1, keepdim=True) | |
| score = (text_embs @ image_embs.T).flatten().item() | |
| return score | |
| reward_model = PickScore() | |
| ``` | |
| ## 2. Inference-time Scaling 技术 | |
| Inference-time Scaling [[2]](https://arxiv.org/abs/2504.00294) 是一类有趣的技术,旨在通过增加推理时的计算量来提升生成结果的质量。例如,在语言模型中,[Qwen/Qwen3.5-27B](https://modelscope.cn/models/Qwen/Qwen3.5-27B)、[deepseek-ai/DeepSeek-R1](deepseek-ai/DeepSeek-R1) 等模型通过“思考模式”引导模型花更多时间仔细思考,让回答结果更准确。接下来我们以模型 [black-forest-labs/FLUX.2-klein-4B](https://modelscope.cn/models/black-forest-labs/FLUX.2-klein-4B) 为例,探讨如何为图像生成模型设计 Inference-time Scaling 方案。 | |
| > 在开始前,我们稍微改造了 `Flux2ImagePipeline` 的代码,使其能够根据输入的特定高斯噪声矩阵进行初始化,便于复现结果,详见 [diffsynth/pipelines/flux2_image.py](https://github.com/modelscope/DiffSynth-Studio/blob/main/diffsynth/pipelines/flux2_image.py) 中的 `Flux2Unit_NoiseInitializer`。 | |
| 运行以下代码,加载模型 [black-forest-labs/FLUX.2-klein-4B](https://modelscope.cn/models/black-forest-labs/FLUX.2-klein-4B)。 | |
| ```python | |
| from diffsynth.pipelines.flux2_image import Flux2ImagePipeline, ModelConfig | |
| pipe = Flux2ImagePipeline.from_pretrained( | |
| torch_dtype=torch.bfloat16, | |
| device="cuda", | |
| model_configs=[ | |
| ModelConfig(model_id="black-forest-labs/FLUX.2-klein-4B", origin_file_pattern="text_encoder/*.safetensors"), | |
| ModelConfig(model_id="black-forest-labs/FLUX.2-klein-4B", origin_file_pattern="transformer/*.safetensors"), | |
| ModelConfig(model_id="black-forest-labs/FLUX.2-klein-4B", origin_file_pattern="vae/diffusion_pytorch_model.safetensors"), | |
| ], | |
| tokenizer_config=ModelConfig(model_id="black-forest-labs/FLUX.2-klein-4B", origin_file_pattern="tokenizer/"), | |
| ) | |
| ``` | |
| 用提示词 `"sketch, a cat"` 生成一只素描猫猫,并用 PickScore 模型打分。 | |
| ```python | |
| def evaluate_noise(noise, pipe, reward_model, prompt): | |
| # Generate an image and compute the score. | |
| image = pipe( | |
| prompt=prompt, | |
| num_inference_steps=4, | |
| initial_noise=noise, | |
| progress_bar_cmd=lambda x: x, | |
| ) | |
| score = reward_model(image, prompt) | |
| return score | |
| torch.manual_seed(1) | |
| prompt = "sketch, a cat" | |
| noise = pipe.generate_noise((1, 128, 64, 64), rand_device="cuda", rand_torch_dtype=pipe.torch_dtype) | |
| image_1 = pipe(prompt, num_inference_steps=4, initial_noise=noise) | |
| print("Score:", reward_model(image_1, prompt)) | |
| image_1 | |
| ``` | |
|  | |
| ### 2.1 Best-of-N 随机搜索 | |
| 模型的生成结果具有一定的随机性,如果用不同的随机种子,生成的图像结果也是不同的,有时图像质量高,有时图像质量低。那么,我们有一个简单的 Inference-time scaling 方案:使用多个不同的随机种子分别生成图像,然后利用 PickScore 进行打分,只保留分数最高的那一张。 | |
| ```python | |
| from tqdm import tqdm | |
| def random_search(base_latents, objective_reward_fn, total_eval_budget): | |
| # Search for the noise randomly. | |
| best_noise = base_latents | |
| best_score = objective_reward_fn(base_latents) | |
| for it in tqdm(range(total_eval_budget - 1)): | |
| noise = pipe.generate_noise((1, 128, 64, 64), seed=None) | |
| score = objective_reward_fn(noise) | |
| if score > best_score: | |
| best_score, best_noise = score, noise | |
| return best_noise | |
| best_noise = random_search( | |
| base_latents=noise, | |
| objective_reward_fn=lambda noise: evaluate_noise(noise, pipe, reward_model, prompt), | |
| total_eval_budget=50, | |
| ) | |
| image_2 = pipe(prompt, num_inference_steps=4, initial_noise=best_noise) | |
| print("Score:", reward_model(image_2, prompt)) | |
| image_2 | |
| ``` | |
|  | |
| 我们可以清晰地看到,经过多次随机搜索后,最终选出的猫猫毛发细节更加丰富,PickScore 分数也有明显提升。但这种暴力的随机搜索效率极低,生成时间成倍增长,且很容易触及质量上限。因此,我们希望能够找到一种更高效的搜索方法,在同等计算预算下达到更高的分数。 | |
| ### 2.2 SES 搜索 | |
| 为了突破随机搜索的瓶颈,我们引入了 SES (Spectral Evolution Search) 算法 [[3]](https://arxiv.org/abs/2602.03208),详细的代码位于 [diffsynth/utils/ses](https://github.com/modelscope/DiffSynth-Studio/blob/main/diffsynth/utils/ses)。 | |
| 扩散模型生成的图像,很大程度上由初始噪声的低频分量决定。SES 算法通过小波变换将高斯噪声分解,固定高频细节,专门针对低频部分使用交叉熵方法进行演化搜索,能以更高的效率找到优质的初始噪声。 | |
| 运行下面的代码,即可使用 SES 更高效地搜索最佳的高斯噪声矩阵。 | |
| ```python | |
| from diffsynth.utils.ses import ses_search | |
| best_noise = ses_search( | |
| base_latents=noise, | |
| objective_reward_fn=lambda noise: evaluate_noise(noise, pipe, reward_model, prompt), | |
| total_eval_budget=50, | |
| ) | |
| image_3 = pipe(prompt, num_inference_steps=4, initial_noise=best_noise) | |
| print("Score:", reward_model(image_3, prompt)) | |
| image_3 | |
| ``` | |
|  | |
| 可以观察到,在同样的计算预算下,相比于随机搜索,SES 的结果在 PickScore 得分上取得了显著的提升。“素描猫猫”展现出了更精致的整体构图以及更具层次感的明暗对比。 | |
| Inference-time scaling 能够以更长推理时间为代价获得更高的图像质量,那么它生成的图像数据也可以用 DPO [[4]](https://arxiv.org/abs/2311.12908)、差分训练 [[5]](https://arxiv.org/abs/2412.12888) 等方式赋予模型自身,那就是另外一个有趣的探索方向了。 | |