Skip to content

GitLab

"...git@developer.sourcefind.cn:renzhc/diffusers_dcu.git" did not exist on "433cb3f801470feef0a6bab3c90c7b303926ec98"
Unverified Commit 694f9658 authored by hlky's avatar hlky Committed by GitHub
Browse files

Support IPAdapter for more Flux pipelines (#10708) 



* Support IPAdapter for more Flux pipelines

* -copied from

---------
Co-authored-by: default avatarSayak Paul <spsayakpaul@gmail.com>
parent 2d8a41ca
Hide whitespace changes
Inline Side-by-side
Showing with 531 additions and 22 deletions
src/diffusers/pipelines/flux/pipeline_flux_control_img2img.py
View file @ 694f9658
...@@ -438,7 +438,6 @@ class FluxControlImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSin ...@@ -438,7 +438,6 @@ class FluxControlImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSin
return timesteps, num_inference_steps - t_start return timesteps, num_inference_steps - t_start
# Copied from diffusers.pipelines.flux.pipeline_flux_img2img.FluxImg2ImgPipeline.check_inputs
def check_inputs( def check_inputs(
self, self,
prompt, prompt,
......
src/diffusers/pipelines/flux/pipeline_flux_control_inpaint.py
View file @ 694f9658
...@@ -477,7 +477,6 @@ class FluxControlInpaintPipeline( ...@@ -477,7 +477,6 @@ class FluxControlInpaintPipeline(
return timesteps, num_inference_steps - t_start return timesteps, num_inference_steps - t_start
# Copied from diffusers.pipelines.flux.pipeline_flux_img2img.FluxImg2ImgPipeline.check_inputs
def check_inputs( def check_inputs(
self, self,
prompt, prompt,
......
src/diffusers/pipelines/flux/pipeline_flux_controlnet.py
View file @ 694f9658
...@@ -18,14 +18,16 @@ from typing import Any, Callable, Dict, List, Optional, Tuple, Union ...@@ -18,14 +18,16 @@ from typing import Any, Callable, Dict, List, Optional, Tuple, Union
import numpy as np import numpy as np
import torch import torch
from transformers import ( from transformers import (
CLIPImageProcessor,
CLIPTextModel, CLIPTextModel,
CLIPTokenizer, CLIPTokenizer,
CLIPVisionModelWithProjection,
T5EncoderModel, T5EncoderModel,
T5TokenizerFast, T5TokenizerFast,
) )
from ...image_processor import PipelineImageInput, VaeImageProcessor from ...image_processor import PipelineImageInput, VaeImageProcessor
from ...loaders import FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin
from ...models.autoencoders import AutoencoderKL from ...models.autoencoders import AutoencoderKL
from ...models.controlnets.controlnet_flux import FluxControlNetModel, FluxMultiControlNetModel from ...models.controlnets.controlnet_flux import FluxControlNetModel, FluxMultiControlNetModel
from ...models.transformers import FluxTransformer2DModel from ...models.transformers import FluxTransformer2DModel
...@@ -171,7 +173,7 @@ def retrieve_timesteps( ...@@ -171,7 +173,7 @@ def retrieve_timesteps(
return timesteps, num_inference_steps return timesteps, num_inference_steps
class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin): class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin, FluxIPAdapterMixin):
r""" r"""
The Flux pipeline for text-to-image generation. The Flux pipeline for text-to-image generation.
...@@ -198,8 +200,8 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -198,8 +200,8 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
[T5TokenizerFast](https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5TokenizerFast). [T5TokenizerFast](https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5TokenizerFast).
""" """
model_cpu_offload_seq = "text_encoder->text_encoder_2->transformer->vae" model_cpu_offload_seq = "text_encoder->text_encoder_2->image_encoder->transformer->vae"
_optional_components = [] _optional_components = ["image_encoder", "feature_extractor"]
_callback_tensor_inputs = ["latents", "prompt_embeds"] _callback_tensor_inputs = ["latents", "prompt_embeds"]
def __init__( def __init__(
...@@ -214,6 +216,8 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -214,6 +216,8 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
controlnet: Union[ controlnet: Union[
FluxControlNetModel, List[FluxControlNetModel], Tuple[FluxControlNetModel], FluxMultiControlNetModel FluxControlNetModel, List[FluxControlNetModel], Tuple[FluxControlNetModel], FluxMultiControlNetModel
], ],
image_encoder: CLIPVisionModelWithProjection = None,
feature_extractor: CLIPImageProcessor = None,
): ):
super().__init__() super().__init__()
if isinstance(controlnet, (list, tuple)): if isinstance(controlnet, (list, tuple)):
...@@ -228,6 +232,8 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -228,6 +232,8 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
transformer=transformer, transformer=transformer,
scheduler=scheduler, scheduler=scheduler,
controlnet=controlnet, controlnet=controlnet,
image_encoder=image_encoder,
feature_extractor=feature_extractor,
) )
self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8 self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
# Flux latents are turned into 2x2 patches and packed. This means the latent width and height has to be divisible # Flux latents are turned into 2x2 patches and packed. This means the latent width and height has to be divisible
...@@ -413,14 +419,62 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -413,14 +419,62 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
return prompt_embeds, pooled_prompt_embeds, text_ids return prompt_embeds, pooled_prompt_embeds, text_ids
# Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_image
def encode_image(self, image, device, num_images_per_prompt):
dtype = next(self.image_encoder.parameters()).dtype
if not isinstance(image, torch.Tensor):
image = self.feature_extractor(image, return_tensors="pt").pixel_values
image = image.to(device=device, dtype=dtype)
image_embeds = self.image_encoder(image).image_embeds
image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
return image_embeds
# Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.prepare_ip_adapter_image_embeds
def prepare_ip_adapter_image_embeds(
self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
):
image_embeds = []
if ip_adapter_image_embeds is None:
if not isinstance(ip_adapter_image, list):
ip_adapter_image = [ip_adapter_image]
if len(ip_adapter_image) != len(self.transformer.encoder_hid_proj.image_projection_layers):
raise ValueError(
f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.transformer.encoder_hid_proj.image_projection_layers)} IP Adapters."
)
for single_ip_adapter_image, image_proj_layer in zip(
ip_adapter_image, self.transformer.encoder_hid_proj.image_projection_layers
):
single_image_embeds = self.encode_image(single_ip_adapter_image, device, 1)
image_embeds.append(single_image_embeds[None, :])
else:
for single_image_embeds in ip_adapter_image_embeds:
image_embeds.append(single_image_embeds)
ip_adapter_image_embeds = []
for i, single_image_embeds in enumerate(image_embeds):
single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
single_image_embeds = single_image_embeds.to(device=device)
ip_adapter_image_embeds.append(single_image_embeds)
return ip_adapter_image_embeds
def check_inputs( def check_inputs(
self, self,
prompt, prompt,
prompt_2, prompt_2,
height, height,
width, width,
negative_prompt=None,
negative_prompt_2=None,
prompt_embeds=None, prompt_embeds=None,
negative_prompt_embeds=None,
pooled_prompt_embeds=None, pooled_prompt_embeds=None,
negative_pooled_prompt_embeds=None,
callback_on_step_end_tensor_inputs=None, callback_on_step_end_tensor_inputs=None,
max_sequence_length=None, max_sequence_length=None,
): ):
...@@ -455,10 +509,33 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -455,10 +509,33 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
elif prompt_2 is not None and (not isinstance(prompt_2, str) and not isinstance(prompt_2, list)): elif prompt_2 is not None and (not isinstance(prompt_2, str) and not isinstance(prompt_2, list)):
raise ValueError(f"`prompt_2` has to be of type `str` or `list` but is {type(prompt_2)}") raise ValueError(f"`prompt_2` has to be of type `str` or `list` but is {type(prompt_2)}")
if negative_prompt is not None and negative_prompt_embeds is not None:
raise ValueError(
f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:"
f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
)
elif negative_prompt_2 is not None and negative_prompt_embeds is not None:
raise ValueError(
f"Cannot forward both `negative_prompt_2`: {negative_prompt_2} and `negative_prompt_embeds`:"
f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
)
if prompt_embeds is not None and negative_prompt_embeds is not None:
if prompt_embeds.shape != negative_prompt_embeds.shape:
raise ValueError(
"`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but"
f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`"
f" {negative_prompt_embeds.shape}."
)
if prompt_embeds is not None and pooled_prompt_embeds is None: if prompt_embeds is not None and pooled_prompt_embeds is None:
raise ValueError( raise ValueError(
"If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`." "If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`."
) )
if negative_prompt_embeds is not None and negative_pooled_prompt_embeds is None:
raise ValueError(
"If `negative_prompt_embeds` are provided, `negative_pooled_prompt_embeds` also have to be passed. Make sure to generate `negative_pooled_prompt_embeds` from the same text encoder that was used to generate `negative_prompt_embeds`."
)
if max_sequence_length is not None and max_sequence_length > 512: if max_sequence_length is not None and max_sequence_length > 512:
raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}") raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
...@@ -597,6 +674,9 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -597,6 +674,9 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
self, self,
prompt: Union[str, List[str]] = None, prompt: Union[str, List[str]] = None,
prompt_2: Optional[Union[str, List[str]]] = None, prompt_2: Optional[Union[str, List[str]]] = None,
negative_prompt: Union[str, List[str]] = None,
negative_prompt_2: Optional[Union[str, List[str]]] = None,
true_cfg_scale: float = 1.0,
height: Optional[int] = None, height: Optional[int] = None,
width: Optional[int] = None, width: Optional[int] = None,
num_inference_steps: int = 28, num_inference_steps: int = 28,
...@@ -612,6 +692,12 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -612,6 +692,12 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
latents: Optional[torch.FloatTensor] = None, latents: Optional[torch.FloatTensor] = None,
prompt_embeds: Optional[torch.FloatTensor] = None, prompt_embeds: Optional[torch.FloatTensor] = None,
pooled_prompt_embeds: Optional[torch.FloatTensor] = None, pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
ip_adapter_image: Optional[PipelineImageInput] = None,
ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
negative_ip_adapter_image: Optional[PipelineImageInput] = None,
negative_ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
negative_prompt_embeds: Optional[torch.FloatTensor] = None,
negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
output_type: Optional[str] = "pil", output_type: Optional[str] = "pil",
return_dict: bool = True, return_dict: bool = True,
joint_attention_kwargs: Optional[Dict[str, Any]] = None, joint_attention_kwargs: Optional[Dict[str, Any]] = None,
...@@ -679,6 +765,17 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -679,6 +765,17 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
pooled_prompt_embeds (`torch.FloatTensor`, *optional*): pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
If not provided, pooled text embeddings will be generated from `prompt` input argument. If not provided, pooled text embeddings will be generated from `prompt` input argument.
ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. If not
provided, embeddings are computed from the `ip_adapter_image` input argument.
negative_ip_adapter_image:
(`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
negative_ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. If not
provided, embeddings are computed from the `ip_adapter_image` input argument.
output_type (`str`, *optional*, defaults to `"pil"`): output_type (`str`, *optional*, defaults to `"pil"`):
The output format of the generate image. Choose between The output format of the generate image. Choose between
[PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`. [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
...@@ -727,8 +824,12 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -727,8 +824,12 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
prompt_2, prompt_2,
height, height,
width, width,
negative_prompt=negative_prompt,
negative_prompt_2=negative_prompt_2,
prompt_embeds=prompt_embeds, prompt_embeds=prompt_embeds,
negative_prompt_embeds=negative_prompt_embeds,
pooled_prompt_embeds=pooled_prompt_embeds, pooled_prompt_embeds=pooled_prompt_embeds,
negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs, callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs,
max_sequence_length=max_sequence_length, max_sequence_length=max_sequence_length,
) )
...@@ -752,6 +853,7 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -752,6 +853,7 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
lora_scale = ( lora_scale = (
self.joint_attention_kwargs.get("scale", None) if self.joint_attention_kwargs is not None else None self.joint_attention_kwargs.get("scale", None) if self.joint_attention_kwargs is not None else None
) )
do_true_cfg = true_cfg_scale > 1 and negative_prompt is not None
( (
prompt_embeds, prompt_embeds,
pooled_prompt_embeds, pooled_prompt_embeds,
...@@ -766,6 +868,21 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -766,6 +868,21 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
max_sequence_length=max_sequence_length, max_sequence_length=max_sequence_length,
lora_scale=lora_scale, lora_scale=lora_scale,
) )
if do_true_cfg:
(
negative_prompt_embeds,
negative_pooled_prompt_embeds,
_,
) = self.encode_prompt(
prompt=negative_prompt,
prompt_2=negative_prompt_2,
prompt_embeds=negative_prompt_embeds,
pooled_prompt_embeds=negative_pooled_prompt_embeds,
device=device,
num_images_per_prompt=num_images_per_prompt,
max_sequence_length=max_sequence_length,
lora_scale=lora_scale,
)
# 3. Prepare control image # 3. Prepare control image
num_channels_latents = self.transformer.config.in_channels // 4 num_channels_latents = self.transformer.config.in_channels // 4
...@@ -899,12 +1016,43 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -899,12 +1016,43 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
] ]
controlnet_keep.append(keeps[0] if isinstance(self.controlnet, FluxControlNetModel) else keeps) controlnet_keep.append(keeps[0] if isinstance(self.controlnet, FluxControlNetModel) else keeps)
if (ip_adapter_image is not None or ip_adapter_image_embeds is not None) and (
negative_ip_adapter_image is None and negative_ip_adapter_image_embeds is None
):
negative_ip_adapter_image = np.zeros((width, height, 3), dtype=np.uint8)
elif (ip_adapter_image is None and ip_adapter_image_embeds is None) and (
negative_ip_adapter_image is not None or negative_ip_adapter_image_embeds is not None
):
ip_adapter_image = np.zeros((width, height, 3), dtype=np.uint8)
if self.joint_attention_kwargs is None:
self._joint_attention_kwargs = {}
image_embeds = None
negative_image_embeds = None
if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
image_embeds = self.prepare_ip_adapter_image_embeds(
ip_adapter_image,
ip_adapter_image_embeds,
device,
batch_size * num_images_per_prompt,
)
if negative_ip_adapter_image is not None or negative_ip_adapter_image_embeds is not None:
negative_image_embeds = self.prepare_ip_adapter_image_embeds(
negative_ip_adapter_image,
negative_ip_adapter_image_embeds,
device,
batch_size * num_images_per_prompt,
)
# 7. Denoising loop # 7. Denoising loop
with self.progress_bar(total=num_inference_steps) as progress_bar: with self.progress_bar(total=num_inference_steps) as progress_bar:
for i, t in enumerate(timesteps): for i, t in enumerate(timesteps):
if self.interrupt: if self.interrupt:
continue continue
if image_embeds is not None:
self._joint_attention_kwargs["ip_adapter_image_embeds"] = image_embeds
# broadcast to batch dimension in a way that's compatible with ONNX/Core ML # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
timestep = t.expand(latents.shape[0]).to(latents.dtype) timestep = t.expand(latents.shape[0]).to(latents.dtype)
...@@ -960,6 +1108,25 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF ...@@ -960,6 +1108,25 @@ class FluxControlNetPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleF
controlnet_blocks_repeat=controlnet_blocks_repeat, controlnet_blocks_repeat=controlnet_blocks_repeat,
)[0] )[0]
if do_true_cfg:
if negative_image_embeds is not None:
self._joint_attention_kwargs["ip_adapter_image_embeds"] = negative_image_embeds
neg_noise_pred = self.transformer(
hidden_states=latents,
timestep=timestep / 1000,
guidance=guidance,
pooled_projections=negative_pooled_prompt_embeds,
encoder_hidden_states=negative_prompt_embeds,
controlnet_block_samples=controlnet_block_samples,
controlnet_single_block_samples=controlnet_single_block_samples,
txt_ids=text_ids,
img_ids=latent_image_ids,
joint_attention_kwargs=self.joint_attention_kwargs,
return_dict=False,
controlnet_blocks_repeat=controlnet_blocks_repeat,
)[0]
noise_pred = neg_noise_pred + true_cfg_scale * (noise_pred - neg_noise_pred)
# compute the previous noisy sample x_t -> x_t-1 # compute the previous noisy sample x_t -> x_t-1
latents_dtype = latents.dtype latents_dtype = latents.dtype
latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0] latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
......
src/diffusers/pipelines/flux/pipeline_flux_img2img.py
View file @ 694f9658
...@@ -17,10 +17,17 @@ from typing import Any, Callable, Dict, List, Optional, Union ...@@ -17,10 +17,17 @@ from typing import Any, Callable, Dict, List, Optional, Union
import numpy as np import numpy as np
import torch import torch
from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5TokenizerFast from transformers import (
CLIPImageProcessor,
CLIPTextModel,
CLIPTokenizer,
CLIPVisionModelWithProjection,
T5EncoderModel,
T5TokenizerFast,
)
from ...image_processor import PipelineImageInput, VaeImageProcessor from ...image_processor import PipelineImageInput, VaeImageProcessor
from ...loaders import FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, FromSingleFileMixin, TextualInversionLoaderMixin
from ...models.autoencoders import AutoencoderKL from ...models.autoencoders import AutoencoderKL
from ...models.transformers import FluxTransformer2DModel from ...models.transformers import FluxTransformer2DModel
from ...schedulers import FlowMatchEulerDiscreteScheduler from ...schedulers import FlowMatchEulerDiscreteScheduler
...@@ -159,7 +166,7 @@ def retrieve_timesteps( ...@@ -159,7 +166,7 @@ def retrieve_timesteps(
return timesteps, num_inference_steps return timesteps, num_inference_steps
class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin): class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFileMixin, FluxIPAdapterMixin):
r""" r"""
The Flux pipeline for image inpainting. The Flux pipeline for image inpainting.
...@@ -186,8 +193,8 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -186,8 +193,8 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
[T5TokenizerFast](https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5TokenizerFast). [T5TokenizerFast](https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5TokenizerFast).
""" """
model_cpu_offload_seq = "text_encoder->text_encoder_2->transformer->vae" model_cpu_offload_seq = "text_encoder->text_encoder_2->image_encoder->transformer->vae"
_optional_components = [] _optional_components = ["image_encoder", "feature_extractor"]
_callback_tensor_inputs = ["latents", "prompt_embeds"] _callback_tensor_inputs = ["latents", "prompt_embeds"]
def __init__( def __init__(
...@@ -199,6 +206,8 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -199,6 +206,8 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
text_encoder_2: T5EncoderModel, text_encoder_2: T5EncoderModel,
tokenizer_2: T5TokenizerFast, tokenizer_2: T5TokenizerFast,
transformer: FluxTransformer2DModel, transformer: FluxTransformer2DModel,
image_encoder: CLIPVisionModelWithProjection = None,
feature_extractor: CLIPImageProcessor = None,
): ):
super().__init__() super().__init__()
...@@ -210,6 +219,8 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -210,6 +219,8 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
tokenizer_2=tokenizer_2, tokenizer_2=tokenizer_2,
transformer=transformer, transformer=transformer,
scheduler=scheduler, scheduler=scheduler,
image_encoder=image_encoder,
feature_extractor=feature_extractor,
) )
self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8 self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
# Flux latents are turned into 2x2 patches and packed. This means the latent width and height has to be divisible # Flux latents are turned into 2x2 patches and packed. This means the latent width and height has to be divisible
...@@ -395,6 +406,50 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -395,6 +406,50 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
return prompt_embeds, pooled_prompt_embeds, text_ids return prompt_embeds, pooled_prompt_embeds, text_ids
# Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_image
def encode_image(self, image, device, num_images_per_prompt):
dtype = next(self.image_encoder.parameters()).dtype
if not isinstance(image, torch.Tensor):
image = self.feature_extractor(image, return_tensors="pt").pixel_values
image = image.to(device=device, dtype=dtype)
image_embeds = self.image_encoder(image).image_embeds
image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
return image_embeds
# Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.prepare_ip_adapter_image_embeds
def prepare_ip_adapter_image_embeds(
self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
):
image_embeds = []
if ip_adapter_image_embeds is None:
if not isinstance(ip_adapter_image, list):
ip_adapter_image = [ip_adapter_image]
if len(ip_adapter_image) != len(self.transformer.encoder_hid_proj.image_projection_layers):
raise ValueError(
f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.transformer.encoder_hid_proj.image_projection_layers)} IP Adapters."
)
for single_ip_adapter_image, image_proj_layer in zip(
ip_adapter_image, self.transformer.encoder_hid_proj.image_projection_layers
):
single_image_embeds = self.encode_image(single_ip_adapter_image, device, 1)
image_embeds.append(single_image_embeds[None, :])
else:
for single_image_embeds in ip_adapter_image_embeds:
image_embeds.append(single_image_embeds)
ip_adapter_image_embeds = []
for i, single_image_embeds in enumerate(image_embeds):
single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
single_image_embeds = single_image_embeds.to(device=device)
ip_adapter_image_embeds.append(single_image_embeds)
return ip_adapter_image_embeds
# Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image
def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator): def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
if isinstance(generator, list): if isinstance(generator, list):
...@@ -429,8 +484,12 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -429,8 +484,12 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
strength, strength,
height, height,
width, width,
negative_prompt=None,
negative_prompt_2=None,
prompt_embeds=None, prompt_embeds=None,
negative_prompt_embeds=None,
pooled_prompt_embeds=None, pooled_prompt_embeds=None,
negative_pooled_prompt_embeds=None,
callback_on_step_end_tensor_inputs=None, callback_on_step_end_tensor_inputs=None,
max_sequence_length=None, max_sequence_length=None,
): ):
...@@ -468,10 +527,33 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -468,10 +527,33 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
elif prompt_2 is not None and (not isinstance(prompt_2, str) and not isinstance(prompt_2, list)): elif prompt_2 is not None and (not isinstance(prompt_2, str) and not isinstance(prompt_2, list)):
raise ValueError(f"`prompt_2` has to be of type `str` or `list` but is {type(prompt_2)}") raise ValueError(f"`prompt_2` has to be of type `str` or `list` but is {type(prompt_2)}")
if negative_prompt is not None and negative_prompt_embeds is not None:
raise ValueError(
f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:"
f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
)
elif negative_prompt_2 is not None and negative_prompt_embeds is not None:
raise ValueError(
f"Cannot forward both `negative_prompt_2`: {negative_prompt_2} and `negative_prompt_embeds`:"
f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
)
if prompt_embeds is not None and negative_prompt_embeds is not None:
if prompt_embeds.shape != negative_prompt_embeds.shape:
raise ValueError(
"`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but"
f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`"
f" {negative_prompt_embeds.shape}."
)
if prompt_embeds is not None and pooled_prompt_embeds is None: if prompt_embeds is not None and pooled_prompt_embeds is None:
raise ValueError( raise ValueError(
"If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`." "If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`."
) )
if negative_prompt_embeds is not None and negative_pooled_prompt_embeds is None:
raise ValueError(
"If `negative_prompt_embeds` are provided, `negative_pooled_prompt_embeds` also have to be passed. Make sure to generate `negative_pooled_prompt_embeds` from the same text encoder that was used to generate `negative_prompt_embeds`."
)
if max_sequence_length is not None and max_sequence_length > 512: if max_sequence_length is not None and max_sequence_length > 512:
raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}") raise ValueError(f"`max_sequence_length` cannot be greater than 512 but is {max_sequence_length}")
...@@ -586,6 +668,9 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -586,6 +668,9 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
self, self,
prompt: Union[str, List[str]] = None, prompt: Union[str, List[str]] = None,
prompt_2: Optional[Union[str, List[str]]] = None, prompt_2: Optional[Union[str, List[str]]] = None,
negative_prompt: Union[str, List[str]] = None,
negative_prompt_2: Optional[Union[str, List[str]]] = None,
true_cfg_scale: float = 1.0,
image: PipelineImageInput = None, image: PipelineImageInput = None,
height: Optional[int] = None, height: Optional[int] = None,
width: Optional[int] = None, width: Optional[int] = None,
...@@ -598,6 +683,12 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -598,6 +683,12 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
latents: Optional[torch.FloatTensor] = None, latents: Optional[torch.FloatTensor] = None,
prompt_embeds: Optional[torch.FloatTensor] = None, prompt_embeds: Optional[torch.FloatTensor] = None,
pooled_prompt_embeds: Optional[torch.FloatTensor] = None, pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
ip_adapter_image: Optional[PipelineImageInput] = None,
ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
negative_ip_adapter_image: Optional[PipelineImageInput] = None,
negative_ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
negative_prompt_embeds: Optional[torch.FloatTensor] = None,
negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
output_type: Optional[str] = "pil", output_type: Optional[str] = "pil",
return_dict: bool = True, return_dict: bool = True,
joint_attention_kwargs: Optional[Dict[str, Any]] = None, joint_attention_kwargs: Optional[Dict[str, Any]] = None,
...@@ -659,6 +750,17 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -659,6 +750,17 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
pooled_prompt_embeds (`torch.FloatTensor`, *optional*): pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
If not provided, pooled text embeddings will be generated from `prompt` input argument. If not provided, pooled text embeddings will be generated from `prompt` input argument.
ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. If not
provided, embeddings are computed from the `ip_adapter_image` input argument.
negative_ip_adapter_image:
(`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
negative_ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. If not
provided, embeddings are computed from the `ip_adapter_image` input argument.
output_type (`str`, *optional*, defaults to `"pil"`): output_type (`str`, *optional*, defaults to `"pil"`):
The output format of the generate image. Choose between The output format of the generate image. Choose between
[PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`. [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
...@@ -697,8 +799,12 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -697,8 +799,12 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
strength, strength,
height, height,
width, width,
negative_prompt=negative_prompt,
negative_prompt_2=negative_prompt_2,
prompt_embeds=prompt_embeds, prompt_embeds=prompt_embeds,
negative_prompt_embeds=negative_prompt_embeds,
pooled_prompt_embeds=pooled_prompt_embeds, pooled_prompt_embeds=pooled_prompt_embeds,
negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs, callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs,
max_sequence_length=max_sequence_length, max_sequence_length=max_sequence_length,
) )
...@@ -724,6 +830,7 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -724,6 +830,7 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
lora_scale = ( lora_scale = (
self.joint_attention_kwargs.get("scale", None) if self.joint_attention_kwargs is not None else None self.joint_attention_kwargs.get("scale", None) if self.joint_attention_kwargs is not None else None
) )
do_true_cfg = true_cfg_scale > 1 and negative_prompt is not None
( (
prompt_embeds, prompt_embeds,
pooled_prompt_embeds, pooled_prompt_embeds,
...@@ -738,6 +845,21 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -738,6 +845,21 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
max_sequence_length=max_sequence_length, max_sequence_length=max_sequence_length,
lora_scale=lora_scale, lora_scale=lora_scale,
) )
if do_true_cfg:
(
negative_prompt_embeds,
negative_pooled_prompt_embeds,
_,
) = self.encode_prompt(
prompt=negative_prompt,
prompt_2=negative_prompt_2,
prompt_embeds=negative_prompt_embeds,
pooled_prompt_embeds=negative_pooled_prompt_embeds,
device=device,
num_images_per_prompt=num_images_per_prompt,
max_sequence_length=max_sequence_length,
lora_scale=lora_scale,
)
# 4.Prepare timesteps # 4.Prepare timesteps
sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
...@@ -791,12 +913,43 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -791,12 +913,43 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
else: else:
guidance = None guidance = None
if (ip_adapter_image is not None or ip_adapter_image_embeds is not None) and (
negative_ip_adapter_image is None and negative_ip_adapter_image_embeds is None
):
negative_ip_adapter_image = np.zeros((width, height, 3), dtype=np.uint8)
elif (ip_adapter_image is None and ip_adapter_image_embeds is None) and (
negative_ip_adapter_image is not None or negative_ip_adapter_image_embeds is not None
):
ip_adapter_image = np.zeros((width, height, 3), dtype=np.uint8)
if self.joint_attention_kwargs is None:
self._joint_attention_kwargs = {}
image_embeds = None
negative_image_embeds = None
if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
image_embeds = self.prepare_ip_adapter_image_embeds(
ip_adapter_image,
ip_adapter_image_embeds,
device,
batch_size * num_images_per_prompt,
)
if negative_ip_adapter_image is not None or negative_ip_adapter_image_embeds is not None:
negative_image_embeds = self.prepare_ip_adapter_image_embeds(
negative_ip_adapter_image,
negative_ip_adapter_image_embeds,
device,
batch_size * num_images_per_prompt,
)
# 6. Denoising loop # 6. Denoising loop
with self.progress_bar(total=num_inference_steps) as progress_bar: with self.progress_bar(total=num_inference_steps) as progress_bar:
for i, t in enumerate(timesteps): for i, t in enumerate(timesteps):
if self.interrupt: if self.interrupt:
continue continue
if image_embeds is not None:
self._joint_attention_kwargs["ip_adapter_image_embeds"] = image_embeds
# broadcast to batch dimension in a way that's compatible with ONNX/Core ML # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
timestep = t.expand(latents.shape[0]).to(latents.dtype) timestep = t.expand(latents.shape[0]).to(latents.dtype)
noise_pred = self.transformer( noise_pred = self.transformer(
...@@ -811,6 +964,22 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile ...@@ -811,6 +964,22 @@ class FluxImg2ImgPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FromSingleFile
return_dict=False, return_dict=False,
)[0] )[0]
if do_true_cfg:
if negative_image_embeds is not None:
self._joint_attention_kwargs["ip_adapter_image_embeds"] = negative_image_embeds
neg_noise_pred = self.transformer(
hidden_states=latents,
timestep=timestep / 1000,
guidance=guidance,
pooled_projections=negative_pooled_prompt_embeds,
encoder_hidden_states=negative_prompt_embeds,
txt_ids=text_ids,
img_ids=latent_image_ids,
joint_attention_kwargs=self.joint_attention_kwargs,
return_dict=False,
)[0]
noise_pred = neg_noise_pred + true_cfg_scale * (noise_pred - neg_noise_pred)
# compute the previous noisy sample x_t -> x_t-1 # compute the previous noisy sample x_t -> x_t-1
latents_dtype = latents.dtype latents_dtype = latents.dtype
latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0] latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
......
src/diffusers/pipelines/flux/pipeline_flux_inpaint.py
View file @ 694f9658
...@@ -18,10 +18,17 @@ from typing import Any, Callable, Dict, List, Optional, Union ...@@ -18,10 +18,17 @@ from typing import Any, Callable, Dict, List, Optional, Union
import numpy as np import numpy as np
import PIL.Image import PIL.Image
import torch import torch
from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5TokenizerFast from transformers import (
CLIPImageProcessor,
CLIPTextModel,
CLIPTokenizer,
CLIPVisionModelWithProjection,
T5EncoderModel,
T5TokenizerFast,
)
from ...image_processor import PipelineImageInput, VaeImageProcessor from ...image_processor import PipelineImageInput, VaeImageProcessor
from ...loaders import FluxLoraLoaderMixin, TextualInversionLoaderMixin from ...loaders import FluxIPAdapterMixin, FluxLoraLoaderMixin, TextualInversionLoaderMixin
from ...models.autoencoders import AutoencoderKL from ...models.autoencoders import AutoencoderKL
from ...models.transformers import FluxTransformer2DModel from ...models.transformers import FluxTransformer2DModel
from ...schedulers import FlowMatchEulerDiscreteScheduler from ...schedulers import FlowMatchEulerDiscreteScheduler
...@@ -156,7 +163,7 @@ def retrieve_timesteps( ...@@ -156,7 +163,7 @@ def retrieve_timesteps(
return timesteps, num_inference_steps return timesteps, num_inference_steps
class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin, FluxIPAdapterMixin):
r""" r"""
The Flux pipeline for image inpainting. The Flux pipeline for image inpainting.
...@@ -183,8 +190,8 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -183,8 +190,8 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
[T5TokenizerFast](https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5TokenizerFast). [T5TokenizerFast](https://huggingface.co/docs/transformers/en/model_doc/t5#transformers.T5TokenizerFast).
""" """
model_cpu_offload_seq = "text_encoder->text_encoder_2->transformer->vae" model_cpu_offload_seq = "text_encoder->text_encoder_2->image_encoder->transformer->vae"
_optional_components = [] _optional_components = ["image_encoder", "feature_extractor"]
_callback_tensor_inputs = ["latents", "prompt_embeds"] _callback_tensor_inputs = ["latents", "prompt_embeds"]
def __init__( def __init__(
...@@ -196,6 +203,8 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -196,6 +203,8 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
text_encoder_2: T5EncoderModel, text_encoder_2: T5EncoderModel,
tokenizer_2: T5TokenizerFast, tokenizer_2: T5TokenizerFast,
transformer: FluxTransformer2DModel, transformer: FluxTransformer2DModel,
image_encoder: CLIPVisionModelWithProjection = None,
feature_extractor: CLIPImageProcessor = None,
): ):
super().__init__() super().__init__()
...@@ -207,6 +216,8 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -207,6 +216,8 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
tokenizer_2=tokenizer_2, tokenizer_2=tokenizer_2,
transformer=transformer, transformer=transformer,
scheduler=scheduler, scheduler=scheduler,
image_encoder=image_encoder,
feature_extractor=feature_extractor,
) )
self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8 self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if getattr(self, "vae", None) else 8
# Flux latents are turned into 2x2 patches and packed. This means the latent width and height has to be divisible # Flux latents are turned into 2x2 patches and packed. This means the latent width and height has to be divisible
...@@ -400,6 +411,50 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -400,6 +411,50 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
return prompt_embeds, pooled_prompt_embeds, text_ids return prompt_embeds, pooled_prompt_embeds, text_ids
# Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.encode_image
def encode_image(self, image, device, num_images_per_prompt):
dtype = next(self.image_encoder.parameters()).dtype
if not isinstance(image, torch.Tensor):
image = self.feature_extractor(image, return_tensors="pt").pixel_values
image = image.to(device=device, dtype=dtype)
image_embeds = self.image_encoder(image).image_embeds
image_embeds = image_embeds.repeat_interleave(num_images_per_prompt, dim=0)
return image_embeds
# Copied from diffusers.pipelines.flux.pipeline_flux.FluxPipeline.prepare_ip_adapter_image_embeds
def prepare_ip_adapter_image_embeds(
self, ip_adapter_image, ip_adapter_image_embeds, device, num_images_per_prompt
):
image_embeds = []
if ip_adapter_image_embeds is None:
if not isinstance(ip_adapter_image, list):
ip_adapter_image = [ip_adapter_image]
if len(ip_adapter_image) != len(self.transformer.encoder_hid_proj.image_projection_layers):
raise ValueError(
f"`ip_adapter_image` must have same length as the number of IP Adapters. Got {len(ip_adapter_image)} images and {len(self.transformer.encoder_hid_proj.image_projection_layers)} IP Adapters."
)
for single_ip_adapter_image, image_proj_layer in zip(
ip_adapter_image, self.transformer.encoder_hid_proj.image_projection_layers
):
single_image_embeds = self.encode_image(single_ip_adapter_image, device, 1)
image_embeds.append(single_image_embeds[None, :])
else:
for single_image_embeds in ip_adapter_image_embeds:
image_embeds.append(single_image_embeds)
ip_adapter_image_embeds = []
for i, single_image_embeds in enumerate(image_embeds):
single_image_embeds = torch.cat([single_image_embeds] * num_images_per_prompt, dim=0)
single_image_embeds = single_image_embeds.to(device=device)
ip_adapter_image_embeds.append(single_image_embeds)
return ip_adapter_image_embeds
# Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image # Copied from diffusers.pipelines.stable_diffusion_3.pipeline_stable_diffusion_3_inpaint.StableDiffusion3InpaintPipeline._encode_vae_image
def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator): def _encode_vae_image(self, image: torch.Tensor, generator: torch.Generator):
if isinstance(generator, list): if isinstance(generator, list):
...@@ -437,8 +492,12 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -437,8 +492,12 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
height, height,
width, width,
output_type, output_type,
negative_prompt=None,
negative_prompt_2=None,
prompt_embeds=None, prompt_embeds=None,
negative_prompt_embeds=None,
pooled_prompt_embeds=None, pooled_prompt_embeds=None,
negative_pooled_prompt_embeds=None,
callback_on_step_end_tensor_inputs=None, callback_on_step_end_tensor_inputs=None,
padding_mask_crop=None, padding_mask_crop=None,
max_sequence_length=None, max_sequence_length=None,
...@@ -477,10 +536,33 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -477,10 +536,33 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
elif prompt_2 is not None and (not isinstance(prompt_2, str) and not isinstance(prompt_2, list)): elif prompt_2 is not None and (not isinstance(prompt_2, str) and not isinstance(prompt_2, list)):
raise ValueError(f"`prompt_2` has to be of type `str` or `list` but is {type(prompt_2)}") raise ValueError(f"`prompt_2` has to be of type `str` or `list` but is {type(prompt_2)}")
if negative_prompt is not None and negative_prompt_embeds is not None:
raise ValueError(
f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:"
f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
)
elif negative_prompt_2 is not None and negative_prompt_embeds is not None:
raise ValueError(
f"Cannot forward both `negative_prompt_2`: {negative_prompt_2} and `negative_prompt_embeds`:"
f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
)
if prompt_embeds is not None and negative_prompt_embeds is not None:
if prompt_embeds.shape != negative_prompt_embeds.shape:
raise ValueError(
"`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but"
f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`"
f" {negative_prompt_embeds.shape}."
)
if prompt_embeds is not None and pooled_prompt_embeds is None: if prompt_embeds is not None and pooled_prompt_embeds is None:
raise ValueError( raise ValueError(
"If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`." "If `prompt_embeds` are provided, `pooled_prompt_embeds` also have to be passed. Make sure to generate `pooled_prompt_embeds` from the same text encoder that was used to generate `prompt_embeds`."
) )
if negative_prompt_embeds is not None and negative_pooled_prompt_embeds is None:
raise ValueError(
"If `negative_prompt_embeds` are provided, `negative_pooled_prompt_embeds` also have to be passed. Make sure to generate `negative_pooled_prompt_embeds` from the same text encoder that was used to generate `negative_prompt_embeds`."
)
if padding_mask_crop is not None: if padding_mask_crop is not None:
if not isinstance(image, PIL.Image.Image): if not isinstance(image, PIL.Image.Image):
...@@ -684,6 +766,9 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -684,6 +766,9 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
self, self,
prompt: Union[str, List[str]] = None, prompt: Union[str, List[str]] = None,
prompt_2: Optional[Union[str, List[str]]] = None, prompt_2: Optional[Union[str, List[str]]] = None,
negative_prompt: Union[str, List[str]] = None,
negative_prompt_2: Optional[Union[str, List[str]]] = None,
true_cfg_scale: float = 1.0,
image: PipelineImageInput = None, image: PipelineImageInput = None,
mask_image: PipelineImageInput = None, mask_image: PipelineImageInput = None,
masked_image_latents: PipelineImageInput = None, masked_image_latents: PipelineImageInput = None,
...@@ -699,6 +784,12 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -699,6 +784,12 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
latents: Optional[torch.FloatTensor] = None, latents: Optional[torch.FloatTensor] = None,
prompt_embeds: Optional[torch.FloatTensor] = None, prompt_embeds: Optional[torch.FloatTensor] = None,
pooled_prompt_embeds: Optional[torch.FloatTensor] = None, pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
ip_adapter_image: Optional[PipelineImageInput] = None,
ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
negative_ip_adapter_image: Optional[PipelineImageInput] = None,
negative_ip_adapter_image_embeds: Optional[List[torch.Tensor]] = None,
negative_prompt_embeds: Optional[torch.FloatTensor] = None,
negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
output_type: Optional[str] = "pil", output_type: Optional[str] = "pil",
return_dict: bool = True, return_dict: bool = True,
joint_attention_kwargs: Optional[Dict[str, Any]] = None, joint_attention_kwargs: Optional[Dict[str, Any]] = None,
...@@ -777,6 +868,17 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -777,6 +868,17 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
pooled_prompt_embeds (`torch.FloatTensor`, *optional*): pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
If not provided, pooled text embeddings will be generated from `prompt` input argument. If not provided, pooled text embeddings will be generated from `prompt` input argument.
ip_adapter_image: (`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. If not
provided, embeddings are computed from the `ip_adapter_image` input argument.
negative_ip_adapter_image:
(`PipelineImageInput`, *optional*): Optional image input to work with IP Adapters.
negative_ip_adapter_image_embeds (`List[torch.Tensor]`, *optional*):
Pre-generated image embeddings for IP-Adapter. It should be a list of length same as number of
IP-adapters. Each element should be a tensor of shape `(batch_size, num_images, emb_dim)`. If not
provided, embeddings are computed from the `ip_adapter_image` input argument.
output_type (`str`, *optional*, defaults to `"pil"`): output_type (`str`, *optional*, defaults to `"pil"`):
The output format of the generate image. Choose between The output format of the generate image. Choose between
[PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`. [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
...@@ -818,8 +920,12 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -818,8 +920,12 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
height, height,
width, width,
output_type=output_type, output_type=output_type,
negative_prompt=negative_prompt,
negative_prompt_2=negative_prompt_2,
prompt_embeds=prompt_embeds, prompt_embeds=prompt_embeds,
negative_prompt_embeds=negative_prompt_embeds,
pooled_prompt_embeds=pooled_prompt_embeds, pooled_prompt_embeds=pooled_prompt_embeds,
negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs, callback_on_step_end_tensor_inputs=callback_on_step_end_tensor_inputs,
padding_mask_crop=padding_mask_crop, padding_mask_crop=padding_mask_crop,
max_sequence_length=max_sequence_length, max_sequence_length=max_sequence_length,
...@@ -856,6 +962,7 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -856,6 +962,7 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
lora_scale = ( lora_scale = (
self.joint_attention_kwargs.get("scale", None) if self.joint_attention_kwargs is not None else None self.joint_attention_kwargs.get("scale", None) if self.joint_attention_kwargs is not None else None
) )
do_true_cfg = true_cfg_scale > 1 and negative_prompt is not None
( (
prompt_embeds, prompt_embeds,
pooled_prompt_embeds, pooled_prompt_embeds,
...@@ -870,6 +977,21 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -870,6 +977,21 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
max_sequence_length=max_sequence_length, max_sequence_length=max_sequence_length,
lora_scale=lora_scale, lora_scale=lora_scale,
) )
if do_true_cfg:
(
negative_prompt_embeds,
negative_pooled_prompt_embeds,
_,
) = self.encode_prompt(
prompt=negative_prompt,
prompt_2=negative_prompt_2,
prompt_embeds=negative_prompt_embeds,
pooled_prompt_embeds=negative_pooled_prompt_embeds,
device=device,
num_images_per_prompt=num_images_per_prompt,
max_sequence_length=max_sequence_length,
lora_scale=lora_scale,
)
# 4.Prepare timesteps # 4.Prepare timesteps
sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas sigmas = np.linspace(1.0, 1 / num_inference_steps, num_inference_steps) if sigmas is None else sigmas
...@@ -946,12 +1068,43 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -946,12 +1068,43 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
else: else:
guidance = None guidance = None
if (ip_adapter_image is not None or ip_adapter_image_embeds is not None) and (
negative_ip_adapter_image is None and negative_ip_adapter_image_embeds is None
):
negative_ip_adapter_image = np.zeros((width, height, 3), dtype=np.uint8)
elif (ip_adapter_image is None and ip_adapter_image_embeds is None) and (
negative_ip_adapter_image is not None or negative_ip_adapter_image_embeds is not None
):
ip_adapter_image = np.zeros((width, height, 3), dtype=np.uint8)
if self.joint_attention_kwargs is None:
self._joint_attention_kwargs = {}
image_embeds = None
negative_image_embeds = None
if ip_adapter_image is not None or ip_adapter_image_embeds is not None:
image_embeds = self.prepare_ip_adapter_image_embeds(
ip_adapter_image,
ip_adapter_image_embeds,
device,
batch_size * num_images_per_prompt,
)
if negative_ip_adapter_image is not None or negative_ip_adapter_image_embeds is not None:
negative_image_embeds = self.prepare_ip_adapter_image_embeds(
negative_ip_adapter_image,
negative_ip_adapter_image_embeds,
device,
batch_size * num_images_per_prompt,
)
# 6. Denoising loop # 6. Denoising loop
with self.progress_bar(total=num_inference_steps) as progress_bar: with self.progress_bar(total=num_inference_steps) as progress_bar:
for i, t in enumerate(timesteps): for i, t in enumerate(timesteps):
if self.interrupt: if self.interrupt:
continue continue
if image_embeds is not None:
self._joint_attention_kwargs["ip_adapter_image_embeds"] = image_embeds
# broadcast to batch dimension in a way that's compatible with ONNX/Core ML # broadcast to batch dimension in a way that's compatible with ONNX/Core ML
timestep = t.expand(latents.shape[0]).to(latents.dtype) timestep = t.expand(latents.shape[0]).to(latents.dtype)
noise_pred = self.transformer( noise_pred = self.transformer(
...@@ -966,6 +1119,22 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin): ...@@ -966,6 +1119,22 @@ class FluxInpaintPipeline(DiffusionPipeline, FluxLoraLoaderMixin):
return_dict=False, return_dict=False,
)[0] )[0]
if do_true_cfg:
if negative_image_embeds is not None:
self._joint_attention_kwargs["ip_adapter_image_embeds"] = negative_image_embeds
neg_noise_pred = self.transformer(
hidden_states=latents,
timestep=timestep / 1000,
guidance=guidance,
pooled_projections=negative_pooled_prompt_embeds,
encoder_hidden_states=negative_prompt_embeds,
txt_ids=text_ids,
img_ids=latent_image_ids,
joint_attention_kwargs=self.joint_attention_kwargs,
return_dict=False,
)[0]
noise_pred = neg_noise_pred + true_cfg_scale * (noise_pred - neg_noise_pred)
# compute the previous noisy sample x_t -> x_t-1 # compute the previous noisy sample x_t -> x_t-1
latents_dtype = latents.dtype latents_dtype = latents.dtype
latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0] latents = self.scheduler.step(noise_pred, t, latents, return_dict=False)[0]
......
tests/pipelines/controlnet_flux/test_controlnet_flux.py
View file @ 694f9658
...@@ -39,13 +39,13 @@ from diffusers.utils.testing_utils import ( ...@@ -39,13 +39,13 @@ from diffusers.utils.testing_utils import (
) )
from diffusers.utils.torch_utils import randn_tensor from diffusers.utils.torch_utils import randn_tensor
from ..test_pipelines_common import PipelineTesterMixin from ..test_pipelines_common import FluxIPAdapterTesterMixin, PipelineTesterMixin
enable_full_determinism() enable_full_determinism()
class FluxControlNetPipelineFastTests(unittest.TestCase, PipelineTesterMixin): class FluxControlNetPipelineFastTests(unittest.TestCase, PipelineTesterMixin, FluxIPAdapterTesterMixin):
pipeline_class = FluxControlNetPipeline pipeline_class = FluxControlNetPipeline
params = frozenset(["prompt", "height", "width", "guidance_scale", "prompt_embeds", "pooled_prompt_embeds"]) params = frozenset(["prompt", "height", "width", "guidance_scale", "prompt_embeds", "pooled_prompt_embeds"])
...@@ -128,6 +128,8 @@ class FluxControlNetPipelineFastTests(unittest.TestCase, PipelineTesterMixin): ...@@ -128,6 +128,8 @@ class FluxControlNetPipelineFastTests(unittest.TestCase, PipelineTesterMixin):
"transformer": transformer, "transformer": transformer,
"vae": vae, "vae": vae,
"controlnet": controlnet, "controlnet": controlnet,
"image_encoder": None,
"feature_extractor": None,
} }
def get_dummy_inputs(self, device, seed=0): def get_dummy_inputs(self, device, seed=0):
......
tests/pipelines/flux/test_pipeline_flux_img2img.py
View file @ 694f9658
...@@ -12,13 +12,13 @@ from diffusers.utils.testing_utils import ( ...@@ -12,13 +12,13 @@ from diffusers.utils.testing_utils import (
torch_device, torch_device,
) )
from ..test_pipelines_common import PipelineTesterMixin from ..test_pipelines_common import FluxIPAdapterTesterMixin, PipelineTesterMixin
enable_full_determinism() enable_full_determinism()
class FluxImg2ImgPipelineFastTests(unittest.TestCase, PipelineTesterMixin): class FluxImg2ImgPipelineFastTests(unittest.TestCase, PipelineTesterMixin, FluxIPAdapterTesterMixin):
pipeline_class = FluxImg2ImgPipeline pipeline_class = FluxImg2ImgPipeline
params = frozenset(["prompt", "height", "width", "guidance_scale", "prompt_embeds", "pooled_prompt_embeds"]) params = frozenset(["prompt", "height", "width", "guidance_scale", "prompt_embeds", "pooled_prompt_embeds"])
batch_params = frozenset(["prompt"]) batch_params = frozenset(["prompt"])
...@@ -85,6 +85,8 @@ class FluxImg2ImgPipelineFastTests(unittest.TestCase, PipelineTesterMixin): ...@@ -85,6 +85,8 @@ class FluxImg2ImgPipelineFastTests(unittest.TestCase, PipelineTesterMixin):
"tokenizer_2": tokenizer_2, "tokenizer_2": tokenizer_2,
"transformer": transformer, "transformer": transformer,
"vae": vae, "vae": vae,
"image_encoder": None,
"feature_extractor": None,
} }
def get_dummy_inputs(self, device, seed=0): def get_dummy_inputs(self, device, seed=0):
......
tests/pipelines/flux/test_pipeline_flux_inpaint.py
View file @ 694f9658
...@@ -12,13 +12,13 @@ from diffusers.utils.testing_utils import ( ...@@ -12,13 +12,13 @@ from diffusers.utils.testing_utils import (
torch_device, torch_device,
) )
from ..test_pipelines_common import PipelineTesterMixin from ..test_pipelines_common import FluxIPAdapterTesterMixin, PipelineTesterMixin
enable_full_determinism() enable_full_determinism()
class FluxInpaintPipelineFastTests(unittest.TestCase, PipelineTesterMixin): class FluxInpaintPipelineFastTests(unittest.TestCase, PipelineTesterMixin, FluxIPAdapterTesterMixin):
pipeline_class = FluxInpaintPipeline pipeline_class = FluxInpaintPipeline
params = frozenset(["prompt", "height", "width", "guidance_scale", "prompt_embeds", "pooled_prompt_embeds"]) params = frozenset(["prompt", "height", "width", "guidance_scale", "prompt_embeds", "pooled_prompt_embeds"])
batch_params = frozenset(["prompt"]) batch_params = frozenset(["prompt"])
...@@ -85,6 +85,8 @@ class FluxInpaintPipelineFastTests(unittest.TestCase, PipelineTesterMixin): ...@@ -85,6 +85,8 @@ class FluxInpaintPipelineFastTests(unittest.TestCase, PipelineTesterMixin):
"tokenizer_2": tokenizer_2, "tokenizer_2": tokenizer_2,
"transformer": transformer, "transformer": transformer,
"vae": vae, "vae": vae,
"image_encoder": None,
"feature_extractor": None,
} }
def get_dummy_inputs(self, device, seed=0): def get_dummy_inputs(self, device, seed=0):
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment