File size: 7,461 Bytes
e1eae4e
392b502
e1eae4e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
"""
    modified based on diffusion library from Huggingface: https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
"""
import inspect
import warnings
from typing import List, Optional, Union

import torch

from tqdm.auto import tqdm
from transformers import CLIPFeatureExtractor, CLIPTextModel, CLIPTokenizer

from diffusers.models import AutoencoderKL, UNet2DConditionModel
from diffusers.pipeline_utils import DiffusionPipeline
from diffusers.schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
from safety_checker import StableDiffusionSafetyChecker


class ComposableStableDiffusionPipeline(DiffusionPipeline):
    def __init__(
        self,
        vae: AutoencoderKL,
        text_encoder: CLIPTextModel,
        tokenizer: CLIPTokenizer,
        unet: UNet2DConditionModel,
        scheduler: Union[DDIMScheduler, PNDMScheduler, LMSDiscreteScheduler],
        safety_checker: StableDiffusionSafetyChecker,
        feature_extractor: CLIPFeatureExtractor,
    ):
        super().__init__()
        scheduler = scheduler.set_format("pt")
        self.register_modules(
            vae=vae,
            text_encoder=text_encoder,
            tokenizer=tokenizer,
            unet=unet,
            scheduler=scheduler,
            safety_checker=safety_checker,
            feature_extractor=feature_extractor,
        )

    @torch.no_grad()
    def __call__(
        self,
        prompt: Union[str, List[str]],
        height: Optional[int] = 512,
        width: Optional[int] = 512,
        num_inference_steps: Optional[int] = 50,
        guidance_scale: Optional[float] = 7.5,
        eta: Optional[float] = 0.0,
        generator: Optional[torch.Generator] = None,
        output_type: Optional[str] = "pil",
        **kwargs,
    ):
        if "torch_device" in kwargs:
            device = kwargs.pop("torch_device")
            warnings.warn(
                "`torch_device` is deprecated as an input argument to `__call__` and will be removed in v0.3.0."
                " Consider using `pipe.to(torch_device)` instead."
            )

            # Set device as before (to be removed in 0.3.0)
            if device is None:
                device = "cuda" if torch.cuda.is_available() else "cpu"
            self.to(device)

        if isinstance(prompt, str):
            batch_size = 1
        elif isinstance(prompt, list):
            batch_size = len(prompt)
        else:
            raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}")

        if height % 8 != 0 or width % 8 != 0:
            raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")

        if '|' in prompt:
            prompt = [x.strip() for x in prompt.split('|')]
            print(prompt)

        # get prompt text embeddings
        text_input = self.tokenizer(
            prompt,
            padding="max_length",
            max_length=self.tokenizer.model_max_length,
            truncation=True,
            return_tensors="pt",
        )
        text_embeddings = self.text_encoder(text_input.input_ids.to(self.device))[0]

        # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
        # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
        # corresponds to doing no classifier free guidance.
        do_classifier_free_guidance = guidance_scale > 1.0
        # get unconditional embeddings for classifier free guidance
        if do_classifier_free_guidance:
            max_length = text_input.input_ids.shape[-1]
            uncond_input = self.tokenizer(
                [""] * batch_size, padding="max_length", max_length=max_length, return_tensors="pt"
            )
            uncond_embeddings = self.text_encoder(uncond_input.input_ids.to(self.device))[0]

            # For classifier free guidance, we need to do two forward passes.
            # Here we concatenate the unconditional and text embeddings into a single batch
            # to avoid doing two forward passes
            text_embeddings = torch.cat([uncond_embeddings, text_embeddings])

        # get the intial random noise
        latents = torch.randn(
            (batch_size, self.unet.in_channels, height // 8, width // 8),
            generator=generator,
            device=self.device,
        )

        # set timesteps
        accepts_offset = "offset" in set(inspect.signature(self.scheduler.set_timesteps).parameters.keys())
        extra_set_kwargs = {}
        if accepts_offset:
            extra_set_kwargs["offset"] = 1

        self.scheduler.set_timesteps(num_inference_steps, **extra_set_kwargs)

        # if we use LMSDiscreteScheduler, let's make sure latents are mulitplied by sigmas
        if isinstance(self.scheduler, LMSDiscreteScheduler):
            latents = latents * self.scheduler.sigmas[0]

        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
        # eta (Ξ·) is only used with the DDIMScheduler, it will be ignored for other schedulers.
        # eta corresponds to Ξ· in DDIM paper: https://arxiv.org/abs/2010.02502
        # and should be between [0, 1]
        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
        extra_step_kwargs = {}
        if accepts_eta:
            extra_step_kwargs["eta"] = eta

        for i, t in tqdm(enumerate(self.scheduler.timesteps)):
            # expand the latents if we are doing classifier free guidance
            latent_model_input = torch.cat([latents] * text_embeddings.shape[0]) if do_classifier_free_guidance else latents
            if isinstance(self.scheduler, LMSDiscreteScheduler):
                sigma = self.scheduler.sigmas[i]
                latent_model_input = latent_model_input / ((sigma**2 + 1) ** 0.5)

            # predict the noise residual
            noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=text_embeddings)["sample"]

            # perform guidance
            if do_classifier_free_guidance:
                pred_decomp = noise_pred.chunk(text_embeddings.shape[0])
                noise_pred_uncond, noise_pred_text = pred_decomp[0], torch.cat(pred_decomp[1:], dim=0).mean(dim=0, keepdim=True)
                noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)

            # compute the previous noisy sample x_t -> x_t-1
            if isinstance(self.scheduler, LMSDiscreteScheduler):
                latents = self.scheduler.step(noise_pred, i, latents, **extra_step_kwargs)["prev_sample"]
            else:
                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs)["prev_sample"]

        # scale and decode the image latents with vae
        latents = 1 / 0.18215 * latents
        image = self.vae.decode(latents)

        image = (image / 2 + 0.5).clamp(0, 1)
        image = image.cpu().permute(0, 2, 3, 1).numpy()

        # run safety checker
        safety_cheker_input = self.feature_extractor(self.numpy_to_pil(image), return_tensors="pt").to(self.device)
        image, has_nsfw_concept = self.safety_checker(images=image, clip_input=safety_cheker_input.pixel_values)

        if output_type == "pil":
            image = self.numpy_to_pil(image)

        return {"sample": image, "nsfw_content_detected": has_nsfw_concept}