update for 5-3-1

Author: shunk031

notebooks/5-3-1_prompt-to-prompt.ipynb

@@ -0,0 +1,184 @@
+# ---
+# jupyter:
+#   jupytext:
+#     text_representation:
+#       extension: .py
+#       format_name: percent
+#       format_version: '1.3'
+#       jupytext_version: 1.16.7
+#   kernelspec:
+#     display_name: Python 3 (ipykernel)
+#     language: python
+#     name: python3
+# ---
+
+# %% [markdown]
+# # Prompt-to-Prompt の実装
+
+# %% [markdown]
+# [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/py-img-gen/python-image-generation/blob/main/notebooks/5-3-1_prompt-to-prompt.ipynb)
+
+# %% [markdown]
+# 参考: https://github.com/huggingface/diffusers/tree/main/examples/community#prompt2prompt-pipeline
+
+# %% [markdown]
+# ## 準備
+
+# %%
+# !pip install -qq py-img-gen
+
+# %%
+import warnings
+
+import torch
+
+device = torch.device(
+    "cuda" if torch.cuda.is_available() else "cpu"
+)
+dtype = torch.float16
+variant = "fp16"
+seed = 0
+
+warnings.simplefilter("ignore", FutureWarning)
+
+# %%
+from diffusers import DiffusionPipeline
+
+model_id = "stable-diffusion-v1-5/stable-diffusion-v1-5"
+
+prompt2prompt = DiffusionPipeline.from_pretrained(
+    model_id,
+    custom_pipeline="pipeline_prompt2prompt",
+)
+
+
+# %%
+from diffusers_modules.git.pipeline_prompt2prompt import (
+    Prompt2PromptPipeline,
+)
+
+assert isinstance(prompt2prompt, Prompt2PromptPipeline)
+prompt2prompt = prompt2prompt.to(device=device)
+
+# %% [markdown]
+# ## 簡易パイプライン実行関数の定義
+
+# %%
+from typing import Any, Dict, List
+
+from diffusers.utils import make_image_grid
+from PIL.Image import Image as PilImage
+
+
+def prompt_to_prompt(
+    prompt_before: str,
+    prompt_after: str,
+    cross_attention_kwargs: Dict[str, Any],
+    width: int = 512,
+    height: int = 512,
+    num_inference_steps: int = 50,
+) -> List[PilImage]:
+    outputs = prompt2prompt(
+        prompt=[prompt_before, prompt_after],
+        width=width,
+        height=height,
+        num_inference_steps=num_inference_steps,
+        cross_attention_kwargs=cross_attention_kwargs,
+        generator=torch.manual_seed(seed),
+    )
+    return outputs.images
+
+
+# %% [markdown]
+# ## Relace Edit
+
+# %%
+prompt_before = "A painting of a squirrel eating a burger"
+prompt_after = "A painting of a cat eating a burger"
+
+images = prompt_to_prompt(
+    prompt_before=prompt_before,
+    prompt_after=prompt_after,
+    cross_attention_kwargs={
+        "edit_type": "replace",
+        "cross_replace_steps": 0.4,
+        "self_replace_steps": 0.4,
+    },
+)
+make_image_grid(images, rows=1, cols=len(images))
+
+
+# %% [markdown]
+# ## Replace Edit with Loacl Blend
+
+# %%
+prompt_before = "A painting of a squirrel eating a burger"
+prompt_after = "A painting of a cat eating a burger"
+
+images = prompt_to_prompt(
+    prompt_before=prompt_before,
+    prompt_after=prompt_after,
+    cross_attention_kwargs={
+        "edit_type": "replace",
+        "cross_replace_steps": 0.4,
+        "self_replace_steps": 0.4,
+        "local_blend_words": ["squirrel", "cat"],
+    },
+)
+make_image_grid(images, rows=1, cols=len(images))
+
+# %% [markdown]
+# ## Refine Edit
+
+# %%
+prompt_before = "A painting of a squirrel eating a burger"
+prompt_after = "A real photo of a cat eating a burger"
+
+images = prompt_to_prompt(
+    prompt_before=prompt_before,
+    prompt_after=prompt_after,
+    cross_attention_kwargs={
+        "edit_type": "refine",
+        "cross_replace_steps": 0.9,
+        "self_replace_steps": 0.2,
+    },
+)
+make_image_grid(images, rows=1, cols=len(images))
+
+# %% [markdown]
+# ## Refine Edit with Local Blend
+
+# %%
+prompt_before = "A painting of a squirrel eating a burger"
+prompt_after = "A real photo of a cat eating a burger"
+
+images = prompt_to_prompt(
+    prompt_before=prompt_before,
+    prompt_after=prompt_after,
+    cross_attention_kwargs={
+        "edit_type": "refine",
+        "cross_replace_steps": 0.9,
+        "self_replace_steps": 0.2,
+        "local_blend_words": ["squirrel", "cat"],
+    },
+)
+make_image_grid(images, rows=1, cols=len(images))
+
+# %% [markdown]
+# ## Rewiehgt Edit
+
+# %%
+prompt = "a photo of smiling pink bunny doll"
+
+images = prompt_to_prompt(
+    prompt_before=prompt,
+    prompt_after=prompt,
+    cross_attention_kwargs={
+        "edit_type": "reweight",
+        "cross_replace_steps": 0.8,
+        "self_replace_steps": 0.8,
+        "equalizer_words": ["smiling"],
+        "equalizer_strengths": [20],
+    },
+)
+make_image_grid(images, rows=1, cols=len(images))