mirror of
https://github.com/invoke-ai/InvokeAI
synced 2024-08-30 20:32:17 +00:00
Comment unused IPAdapter generate(...) methods.
This commit is contained in:
Ryan Dick
parent
b2d5b53b5f
commit
d669f0855d
@ -1,8 +1,6 @@
|
||||
# copied from https://github.com/tencent-ailab/IP-Adapter (Apache License 2.0)
|
||||
# and modified as needed
|
||||
|
||||
from typing import List
|
||||
|
||||
import torch
|
||||
|
||||
# FIXME: Getting errors when trying to use PyTorch 2.0 versions of IPAttnProcessor and AttnProcessor
|
||||
@ -120,134 +118,135 @@ class IPAdapter:
|
||||
if isinstance(attn_processor, IPAttnProcessor):
|
||||
attn_processor.scale = scale
|
||||
|
||||
# IPAdapter.generate() method is not used for InvokeAI
|
||||
# left here for reference
|
||||
def generate(
|
||||
self,
|
||||
pil_image,
|
||||
prompt=None,
|
||||
negative_prompt=None,
|
||||
scale=1.0,
|
||||
num_samples=4,
|
||||
seed=-1,
|
||||
guidance_scale=7.5,
|
||||
num_inference_steps=30,
|
||||
**kwargs,
|
||||
):
|
||||
self.set_scale(scale)
|
||||
# IPAdapter.generate() method is not used for InvokeAI. Left here for reference:
|
||||
# def generate(
|
||||
# self,
|
||||
# pil_image,
|
||||
# prompt=None,
|
||||
# negative_prompt=None,
|
||||
# scale=1.0,
|
||||
# num_samples=4,
|
||||
# seed=-1,
|
||||
# guidance_scale=7.5,
|
||||
# num_inference_steps=30,
|
||||
# **kwargs,
|
||||
# ):
|
||||
# self.set_scale(scale)
|
||||
|
||||
if isinstance(pil_image, Image.Image):
|
||||
num_prompts = 1
|
||||
else:
|
||||
num_prompts = len(pil_image)
|
||||
# if isinstance(pil_image, Image.Image):
|
||||
# num_prompts = 1
|
||||
# else:
|
||||
# num_prompts = len(pil_image)
|
||||
|
||||
if prompt is None:
|
||||
prompt = "best quality, high quality"
|
||||
if negative_prompt is None:
|
||||
negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
|
||||
# if prompt is None:
|
||||
# prompt = "best quality, high quality"
|
||||
# if negative_prompt is None:
|
||||
# negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
|
||||
|
||||
if not isinstance(prompt, List):
|
||||
prompt = [prompt] * num_prompts
|
||||
if not isinstance(negative_prompt, List):
|
||||
negative_prompt = [negative_prompt] * num_prompts
|
||||
# if not isinstance(prompt, List):
|
||||
# prompt = [prompt] * num_prompts
|
||||
# if not isinstance(negative_prompt, List):
|
||||
# negative_prompt = [negative_prompt] * num_prompts
|
||||
|
||||
image_prompt_embeds, uncond_image_prompt_embeds = self.get_image_embeds(pil_image)
|
||||
bs_embed, seq_len, _ = image_prompt_embeds.shape
|
||||
image_prompt_embeds = image_prompt_embeds.repeat(1, num_samples, 1)
|
||||
image_prompt_embeds = image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
|
||||
uncond_image_prompt_embeds = uncond_image_prompt_embeds.repeat(1, num_samples, 1)
|
||||
uncond_image_prompt_embeds = uncond_image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
|
||||
# image_prompt_embeds, uncond_image_prompt_embeds = self.get_image_embeds(pil_image)
|
||||
# bs_embed, seq_len, _ = image_prompt_embeds.shape
|
||||
# image_prompt_embeds = image_prompt_embeds.repeat(1, num_samples, 1)
|
||||
# image_prompt_embeds = image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
|
||||
# uncond_image_prompt_embeds = uncond_image_prompt_embeds.repeat(1, num_samples, 1)
|
||||
# uncond_image_prompt_embeds = uncond_image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
|
||||
|
||||
with torch.inference_mode():
|
||||
prompt_embeds = self.pipe._encode_prompt(
|
||||
prompt,
|
||||
device=self.device,
|
||||
num_images_per_prompt=num_samples,
|
||||
do_classifier_free_guidance=True,
|
||||
negative_prompt=negative_prompt,
|
||||
)
|
||||
negative_prompt_embeds_, prompt_embeds_ = prompt_embeds.chunk(2)
|
||||
prompt_embeds = torch.cat([prompt_embeds_, image_prompt_embeds], dim=1)
|
||||
negative_prompt_embeds = torch.cat([negative_prompt_embeds_, uncond_image_prompt_embeds], dim=1)
|
||||
# with torch.inference_mode():
|
||||
# prompt_embeds = self.pipe._encode_prompt(
|
||||
# prompt,
|
||||
# device=self.device,
|
||||
# num_images_per_prompt=num_samples,
|
||||
# do_classifier_free_guidance=True,
|
||||
# negative_prompt=negative_prompt,
|
||||
# )
|
||||
# negative_prompt_embeds_, prompt_embeds_ = prompt_embeds.chunk(2)
|
||||
# prompt_embeds = torch.cat([prompt_embeds_, image_prompt_embeds], dim=1)
|
||||
# negative_prompt_embeds = torch.cat([negative_prompt_embeds_, uncond_image_prompt_embeds], dim=1)
|
||||
|
||||
generator = torch.Generator(self.device).manual_seed(seed) if seed is not None else None
|
||||
images = self.pipe(
|
||||
prompt_embeds=prompt_embeds,
|
||||
negative_prompt_embeds=negative_prompt_embeds,
|
||||
guidance_scale=guidance_scale,
|
||||
num_inference_steps=num_inference_steps,
|
||||
generator=generator,
|
||||
**kwargs,
|
||||
).images
|
||||
# generator = torch.Generator(self.device).manual_seed(seed) if seed is not None else None
|
||||
# images = self.pipe(
|
||||
# prompt_embeds=prompt_embeds,
|
||||
# negative_prompt_embeds=negative_prompt_embeds,
|
||||
# guidance_scale=guidance_scale,
|
||||
# num_inference_steps=num_inference_steps,
|
||||
# generator=generator,
|
||||
# **kwargs,
|
||||
# ).images
|
||||
|
||||
return images
|
||||
# return images
|
||||
|
||||
|
||||
class IPAdapterXL(IPAdapter):
|
||||
"""SDXL"""
|
||||
|
||||
def generate(
|
||||
self,
|
||||
pil_image,
|
||||
prompt=None,
|
||||
negative_prompt=None,
|
||||
scale=1.0,
|
||||
num_samples=4,
|
||||
seed=-1,
|
||||
num_inference_steps=30,
|
||||
**kwargs,
|
||||
):
|
||||
self.set_scale(scale)
|
||||
pass
|
||||
# IPAdapterXL.generate() method is not used for InvokeAI. Left here for reference:
|
||||
# def generate(
|
||||
# self,
|
||||
# pil_image,
|
||||
# prompt=None,
|
||||
# negative_prompt=None,
|
||||
# scale=1.0,
|
||||
# num_samples=4,
|
||||
# seed=-1,
|
||||
# num_inference_steps=30,
|
||||
# **kwargs,
|
||||
# ):
|
||||
# self.set_scale(scale)
|
||||
|
||||
if isinstance(pil_image, Image.Image):
|
||||
num_prompts = 1
|
||||
else:
|
||||
num_prompts = len(pil_image)
|
||||
# if isinstance(pil_image, Image.Image):
|
||||
# num_prompts = 1
|
||||
# else:
|
||||
# num_prompts = len(pil_image)
|
||||
|
||||
if prompt is None:
|
||||
prompt = "best quality, high quality"
|
||||
if negative_prompt is None:
|
||||
negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
|
||||
# if prompt is None:
|
||||
# prompt = "best quality, high quality"
|
||||
# if negative_prompt is None:
|
||||
# negative_prompt = "monochrome, lowres, bad anatomy, worst quality, low quality"
|
||||
|
||||
if not isinstance(prompt, List):
|
||||
prompt = [prompt] * num_prompts
|
||||
if not isinstance(negative_prompt, List):
|
||||
negative_prompt = [negative_prompt] * num_prompts
|
||||
# if not isinstance(prompt, List):
|
||||
# prompt = [prompt] * num_prompts
|
||||
# if not isinstance(negative_prompt, List):
|
||||
# negative_prompt = [negative_prompt] * num_prompts
|
||||
|
||||
image_prompt_embeds, uncond_image_prompt_embeds = self.get_image_embeds(pil_image)
|
||||
bs_embed, seq_len, _ = image_prompt_embeds.shape
|
||||
image_prompt_embeds = image_prompt_embeds.repeat(1, num_samples, 1)
|
||||
image_prompt_embeds = image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
|
||||
uncond_image_prompt_embeds = uncond_image_prompt_embeds.repeat(1, num_samples, 1)
|
||||
uncond_image_prompt_embeds = uncond_image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
|
||||
# image_prompt_embeds, uncond_image_prompt_embeds = self.get_image_embeds(pil_image)
|
||||
# bs_embed, seq_len, _ = image_prompt_embeds.shape
|
||||
# image_prompt_embeds = image_prompt_embeds.repeat(1, num_samples, 1)
|
||||
# image_prompt_embeds = image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
|
||||
# uncond_image_prompt_embeds = uncond_image_prompt_embeds.repeat(1, num_samples, 1)
|
||||
# uncond_image_prompt_embeds = uncond_image_prompt_embeds.view(bs_embed * num_samples, seq_len, -1)
|
||||
|
||||
with torch.inference_mode():
|
||||
(
|
||||
prompt_embeds,
|
||||
negative_prompt_embeds,
|
||||
pooled_prompt_embeds,
|
||||
negative_pooled_prompt_embeds,
|
||||
) = self.pipe.encode_prompt(
|
||||
prompt,
|
||||
num_images_per_prompt=num_samples,
|
||||
do_classifier_free_guidance=True,
|
||||
negative_prompt=negative_prompt,
|
||||
)
|
||||
prompt_embeds = torch.cat([prompt_embeds, image_prompt_embeds], dim=1)
|
||||
negative_prompt_embeds = torch.cat([negative_prompt_embeds, uncond_image_prompt_embeds], dim=1)
|
||||
# with torch.inference_mode():
|
||||
# (
|
||||
# prompt_embeds,
|
||||
# negative_prompt_embeds,
|
||||
# pooled_prompt_embeds,
|
||||
# negative_pooled_prompt_embeds,
|
||||
# ) = self.pipe.encode_prompt(
|
||||
# prompt,
|
||||
# num_images_per_prompt=num_samples,
|
||||
# do_classifier_free_guidance=True,
|
||||
# negative_prompt=negative_prompt,
|
||||
# )
|
||||
# prompt_embeds = torch.cat([prompt_embeds, image_prompt_embeds], dim=1)
|
||||
# negative_prompt_embeds = torch.cat([negative_prompt_embeds, uncond_image_prompt_embeds], dim=1)
|
||||
|
||||
generator = torch.Generator(self.device).manual_seed(seed) if seed is not None else None
|
||||
images = self.pipe(
|
||||
prompt_embeds=prompt_embeds,
|
||||
negative_prompt_embeds=negative_prompt_embeds,
|
||||
pooled_prompt_embeds=pooled_prompt_embeds,
|
||||
negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
|
||||
num_inference_steps=num_inference_steps,
|
||||
generator=generator,
|
||||
**kwargs,
|
||||
).images
|
||||
# generator = torch.Generator(self.device).manual_seed(seed) if seed is not None else None
|
||||
# images = self.pipe(
|
||||
# prompt_embeds=prompt_embeds,
|
||||
# negative_prompt_embeds=negative_prompt_embeds,
|
||||
# pooled_prompt_embeds=pooled_prompt_embeds,
|
||||
# negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
|
||||
# num_inference_steps=num_inference_steps,
|
||||
# generator=generator,
|
||||
# **kwargs,
|
||||
# ).images
|
||||
|
||||
return images
|
||||
# return images
|
||||
|
||||
|
||||
class IPAdapterPlus(IPAdapter):
|
||||
|
||||
Loading…
Reference in New Issue
Block a user