├── .github
    └── FUNDING.yml
├── LICENSE
├── README.md
├── flax_stable_diffusion.ipynb
├── flax_stable_diffusion_2.ipynb
├── flax_stable_diffusion_3.ipynb
├── pytorch_freeu_stable_diffusion.ipynb
├── pytorch_riffusion_stable_diffusion.ipynb
├── pytorch_stable_diffusion.ipynb
├── pytorch_stable_diffusion_2.ipynb
└── pytorch_stable_diffusion_3.ipynb


/.github/FUNDING.yml:
--------------------------------------------------------------------------------
 1 | # These are supported funding model platforms
 2 | 
 3 | github: # Replace with up to 4 GitHub Sponsors-enabled usernames e.g., [user1, user2]
 4 | patreon: camenduru
 5 | open_collective: # Replace with a single Open Collective username
 6 | ko_fi: camenduru
 7 | tidelift: # Replace with a single Tidelift platform-name/package-name e.g., npm/babel
 8 | community_bridge: # Replace with a single Community Bridge project-name e.g., cloud-foundry
 9 | liberapay: # Replace with a single Liberapay username
10 | issuehunt: # Replace with a single IssueHunt username
11 | otechie: # Replace with a single Otechie username
12 | lfx_crowdfunding: # Replace with a single LFX Crowdfunding project-name e.g., cloud-foundry
13 | custom: # Replace with up to 4 custom sponsorship URLs e.g., ['link1', 'link2']
14 | 


--------------------------------------------------------------------------------
/LICENSE:
--------------------------------------------------------------------------------
 1 | This is free and unencumbered software released into the public domain.
 2 | 
 3 | Anyone is free to copy, modify, publish, use, compile, sell, or
 4 | distribute this software, either in source code form or as a compiled
 5 | binary, for any purpose, commercial or non-commercial, and by any
 6 | means.
 7 | 
 8 | In jurisdictions that recognize copyright laws, the author or authors
 9 | of this software dedicate any and all copyright interest in the
10 | software to the public domain. We make this dedication for the benefit
11 | of the public at large and to the detriment of our heirs and
12 | successors. We intend this dedication to be an overt act of
13 | relinquishment in perpetuity of all present and future rights to this
14 | software under copyright law.
15 | 
16 | THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
17 | EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
18 | MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
19 | IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
20 | OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
21 | ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
22 | OTHER DEALINGS IN THE SOFTWARE.
23 | 
24 | For more information, please refer to <https://unlicense.org>
25 | 


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | 🐣 Please follow me for new updates https://twitter.com/camenduru <br />
 2 | 🔥 Please join our discord server https://discord.gg/k5BwmmvJJU <br />
 3 | 🥳 Please join my patreon community https://patreon.com/camenduru <br />
 4 | 
 5 | ## 🦒 Colab TPU
 6 | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/flax_stable_diffusion.ipynb) flax_stable_diffusion
 7 | 
 8 | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/flax_stable_diffusion_2.ipynb) flax_stable_diffusion_2 (🏗WIP)
 9 | 
10 | ## 🦆 Kaggle TPU
11 | https://www.kaggle.com/code/camenduru/flax-stable-diffusion
12 | 
13 | | 🤗 HuggingFace Path | Model Page |
14 | | --- | --- |
15 | | flax/anything-v3.0 | https://huggingface.co/Linaqruf/anything-v3.0
16 | | flax/samdoesart `samdoesarts style` | https://huggingface.co/lilpotat/sa
17 | | flax/Stable_Diffusion_PaperCut_Model `PaperCut style` | https://huggingface.co/Fictiverse/Stable_Diffusion_PaperCut_Model
18 | | flax/redshift-diffusion `redshift style` | https://huggingface.co/nitrosocke/redshift-diffusion
19 | | flax/sd15 | https://huggingface.co/runwayml/stable-diffusion-v1-5
20 | | flax/sd15-non-ema | https://huggingface.co/runwayml/stable-diffusion-v1-5
21 | | flax/midjourney-v4-diffusion `mdjrny-v4 style` | https://huggingface.co/prompthero/midjourney-v4-diffusion
22 | | flax/StudioGhibli `studio_ghibli_anime_style style` | https://huggingface.co/IShallRiseAgain/StudioGhibli
23 | | flax/Tron-Legacy-diffusion `trnlgcy style` | https://huggingface.co/dallinmackay/Tron-Legacy-diffusion
24 | | flax/Cyberpunk-Anime-Diffusion `dgs illustration style` | https://huggingface.co/DGSpitzer/Cyberpunk-Anime-Diffusion
25 | | flax/waifu-diffusion | https://huggingface.co/hakurei/waifu-diffusion-v1-3
26 | | flax/Arcane-Diffusion `arcane style` | https://huggingface.co/nitrosocke/Arcane-Diffusion
27 | | flax/classic-anim-diffusion `classic disney style` | https://huggingface.co/nitrosocke/classic-anim-diffusion
28 | | flax/mo-di-diffusion `modern disney style` | https://huggingface.co/nitrosocke/mo-di-diffusion
29 | | camenduru/plushies | https://huggingface.co/camenduru/plushies
30 | | flax/stable-diffusion-2 | https://huggingface.co/stabilityai/stable-diffusion-2
31 | | flax/stable-diffusion-2-base | https://huggingface.co/stabilityai/stable-diffusion-2-base
32 | 
33 | ## 🦒 Colab GPU
34 | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/pytorch_stable_diffusion.ipynb) pytorch_stable_diffusion
35 | 
36 | [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/pytorch_stable_diffusion_2.ipynb) pytorch_stable_diffusion_2 (🏗WIP)
37 | 
38 | ## 🦆 Kaggle GPU
39 | https://www.kaggle.com/code/camenduru/pytorch-stable-diffusion
40 | 
41 | | 🤗 HuggingFace Path | Model Page |
42 | | --- | --- |
43 | | Linaqruf/anything-v3.0 | https://huggingface.co/Linaqruf/anything-v3.0
44 | | lilpotat/sa `samdoesarts style` | https://huggingface.co/lilpotat/sa
45 | | Fictiverse/Stable_Diffusion_PaperCut_Model `PaperCut style` | https://huggingface.co/Fictiverse/Stable_Diffusion_PaperCut_Model
46 | | nitrosocke/redshift-diffusion `redshift style` | https://huggingface.co/nitrosocke/redshift-diffusion
47 | | camenduru/sd15 | https://huggingface.co/runwayml/stable-diffusion-v1-5
48 | | prompthero/midjourney-v4-diffusion `mdjrny-v4 style` | https://huggingface.co/prompthero/midjourney-v4-diffusion
49 | | dallinmackay/Tron-Legacy-diffusion `trnlgcy style` | https://huggingface.co/dallinmackay/Tron-Legacy-diffusion
50 | | DGSpitzer/Cyberpunk-Anime-Diffusion `dgs illustration style` | https://huggingface.co/DGSpitzer/Cyberpunk-Anime-Diffusion
51 | | hakurei/waifu-diffusion-v1-3 | https://huggingface.co/hakurei/waifu-diffusion-v1-3
52 | | nitrosocke/Arcane-Diffusion `arcane style` | https://huggingface.co/nitrosocke/Arcane-Diffusion
53 | | nitrosocke/classic-anim-diffusion `classic disney style` | https://huggingface.co/nitrosocke/classic-anim-diffusion
54 | | nitrosocke/mo-di-diffusion `modern disney style` | https://huggingface.co/nitrosocke/mo-di-diffusion
55 | 
56 | ## Models License
57 | https://huggingface.co/spaces/CompVis/stable-diffusion-license
58 | 


--------------------------------------------------------------------------------
/flax_stable_diffusion.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |   "cells": [
  3 |     {
  4 |       "cell_type": "markdown",
  5 |       "metadata": {
  6 |         "id": "view-in-github"
  7 |       },
  8 |       "source": [
  9 |         "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/flax_stable_diffusion.ipynb)"
 10 |       ]
 11 |     },
 12 |     {
 13 |       "cell_type": "code",
 14 |       "execution_count": null,
 15 |       "metadata": {
 16 |         "cellView": "form",
 17 |         "id": "ymcDoPQn41U4"
 18 |       },
 19 |       "outputs": [],
 20 |       "source": [
 21 |         "save_to = \"here\" #@param [\"here\", \"discord\", \"gdrive\"]\n",
 22 |         "if save_to == \"gdrive\":\n",
 23 |         "  from google.colab import drive\n",
 24 |         "  drive.mount('/content/gdrive')\n",
 25 |         "\n",
 26 |         "!pip install -U jax==0.3.25 jaxlib==0.3.25 flax==0.6.2 transformers piexif fold_to_ascii ftfy\n",
 27 |         "\n",
 28 |         "is_tile = False #@param {type: 'boolean'}\n",
 29 |         "if(is_tile):\n",
 30 |         "  !pip install -U git+https://github.com/camenduru/diffusers@padding\n",
 31 |         "else:\n",
 32 |         "  !pip install -U git+https://github.com/camenduru/diffusers@v0.9.1\n",
 33 |         "\n",
 34 |         "import jax.tools.colab_tpu\n",
 35 |         "jax.tools.colab_tpu.setup_tpu('tpu_driver_nightly')\n",
 36 |         "\n",
 37 |         "import jax, random, os, gc, requests, json, piexif\n",
 38 |         "import numpy as np\n",
 39 |         "from flax.jax_utils import replicate\n",
 40 |         "from flax.training.common_utils import shard\n",
 41 |         "from diffusers import FlaxStableDiffusionPipeline\n",
 42 |         "from PIL import Image\n",
 43 |         "from PIL.PngImagePlugin import PngInfo\n",
 44 |         "from fold_to_ascii import fold\n",
 45 |         "metadata = PngInfo()\n",
 46 |         "\n",
 47 |         "def closestNumber(n, m):\n",
 48 |         "    q = int(n / m)\n",
 49 |         "    n1 = m * q\n",
 50 |         "    if (n * m) > 0:\n",
 51 |         "        n2 = m * (q + 1)\n",
 52 |         "    else:\n",
 53 |         "        n2 = m * (q - 1)\n",
 54 |         "    if abs(n - n1) < abs(n - n2):\n",
 55 |         "        return n1\n",
 56 |         "    return n2\n",
 57 |         "\n",
 58 |         "def image_grid(imgs, rows, cols):\n",
 59 |         "    w,h = imgs[0].size\n",
 60 |         "    grid = Image.new('RGB', size=(cols*w, rows*h))\n",
 61 |         "    for i, img in enumerate(imgs): grid.paste(img, box=(i%cols*w, i//cols*h))\n",
 62 |         "    return grid\n",
 63 |         "\n",
 64 |         "folder_max_files = 500 #@param {type: 'integer'}\n",
 65 |         "root_folder = 'ai_images' #@param {type: 'string'}\n",
 66 |         "if save_to == \"gdrive\":\n",
 67 |         "  root_folder = f\"/content/gdrive/MyDrive/{root_folder}\"\n",
 68 |         "\n",
 69 |         "if os.path.exists(f\"{root_folder}\") == False:\n",
 70 |         "    os.mkdir(f\"{root_folder}\")\n",
 71 |         "image_folder = max([int(f) for f in os.listdir(f\"{root_folder}\")], default=0)\n",
 72 |         "if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
 73 |         "    os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
 74 |         "name = max([int(f[: f.index(\".\")]) for f in os.listdir(f\"{root_folder}/{image_folder:04}\")],default=0,)\n",
 75 |         "\n",
 76 |         "model_folder = \"flax/sd15\" #@param [\"flax/anything-v3.0\", \"flax/samdoesart\", \"flax/Stable_Diffusion_PaperCut_Model\", \"flax/redshift-diffusion\", \"flax/sd15\", \"flax/sd15-non-ema\", \"flax/midjourney-v4-diffusion\", \"flax/StudioGhibli\", \"flax/Tron-Legacy-diffusion\", \"flax/Cyberpunk-Anime-Diffusion\", \"flax/waifu-diffusion\", \"flax/Arcane-Diffusion\", \"flax/classic-anim-diffusion\", \"flax/mo-di-diffusion\", \"camenduru/plushies\", \"lilpotat/f2\", \"lilpotat/rbm\"] {allow-input: true}\n",
 77 |         "if model_folder == \"lilpotat/f2\" or model_folder == \"lilpotat/rbm\":\n",
 78 |         "  revision = \"flax\"\n",
 79 |         "else:\n",
 80 |         "  revision = \"main\"\n",
 81 |         "pipe, params = FlaxStableDiffusionPipeline.from_pretrained(model_folder, revision=revision, from_pt=False, dtype=jax.numpy.bfloat16, safety_checker=None)\n",
 82 |         "params = replicate(params)\n",
 83 |         "\n",
 84 |         "def generate(discord_token, discord_channel_id, discord_user, by, num_inference_steps, guidance_scale, sampler, width, height, prompt, negative_prompt, suffix, image_folder, name):\n",
 85 |         "    width = closestNumber(width, 8)\n",
 86 |         "    height = closestNumber(height, 8)\n",
 87 |         "    metadata.add_text(\"Prompt\", f\"{prompt}\")\n",
 88 |         "    metadata.add_text(\"by\", f\"{by}\")\n",
 89 |         "    gc.collect()\n",
 90 |         "    real_seed = random.randint(0, 2147483647)\n",
 91 |         "    prng_seed = jax.random.PRNGKey(real_seed)\n",
 92 |         "    prng_seed = jax.random.split(prng_seed, jax.device_count())\n",
 93 |         "    num_samples = jax.device_count()\n",
 94 |         "    prompt_n = num_samples * [prompt]\n",
 95 |         "    prompt_ids = pipe.prepare_inputs(prompt_n)\n",
 96 |         "    prompt_ids = shard(prompt_ids)\n",
 97 |         "    negative_prompt_n = num_samples * [negative_prompt]\n",
 98 |         "    negative_prompt_ids = pipe.prepare_inputs(negative_prompt_n)\n",
 99 |         "    negative_prompt_ids = shard(negative_prompt_ids)\n",
100 |         "    images = pipe(prompt_ids, params, prng_seed, neg_prompt_ids=negative_prompt_ids, num_inference_steps=num_inference_steps, height=height, width=width, guidance_scale=guidance_scale, jit=True).images\n",
101 |         "    images = pipe.numpy_to_pil(np.asarray(images.reshape((num_samples,) + images.shape[-3:])))\n",
102 |         "    image = image_grid(images, 2, 4)\n",
103 |         "    if(suffix == 'png'):\n",
104 |         "      image.save(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", pnginfo=metadata)\n",
105 |         "    else:\n",
106 |         "      zeroth_ifd = {piexif.ImageIFD.ImageDescription: f\"{fold(prompt)}\", piexif.ImageIFD.Make: f\"{fold(by)}\", piexif.ImageIFD.Model: f\"{model_folder}\"}\n",
107 |         "      exif_dict = {\"0th\": zeroth_ifd}\n",
108 |         "      exif_bytes = piexif.dump(exif_dict)\n",
109 |         "      image.save(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", \"JPEG\", quality=70, exif=exif_bytes)\n",
110 |         "    files = {f\"{image_folder:04}_{name:04}.{suffix}\": open(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", \"rb\").read()}\n",
111 |         "    if save_to == \"discord\":\n",
112 |         "      payload = {\"content\": f\"{prompt}\\nNegative prompt: {negative_prompt}\\nSteps: {num_inference_steps}, Sampler: {sampler}, CFG scale: {guidance_scale}, Seed: {real_seed}, Size: {width}x{height}, Model folder: {model_folder} - {discord_user}\"}\n",
113 |         "      requests.post(f\"https://discord.com/api/v9/channels/{discord_channel_id}/messages\", data=payload, headers={\"authorization\": f\"Bot {discord_token}\"}, files=files)\n",
114 |         "      os.remove(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\")"
115 |       ]
116 |     },
117 |     {
118 |       "cell_type": "code",
119 |       "execution_count": null,
120 |       "metadata": {
121 |         "cellView": "form",
122 |         "id": "DDUev8TOMxJ4"
123 |       },
124 |       "outputs": [],
125 |       "source": [
126 |         "discord_token = \"token\" #@param {type: 'string'}\n",
127 |         "discord_channel_id = 0 #@param {type: 'integer'}\n",
128 |         "prompt = \"duck\" #@param {type: 'string'}\n",
129 |         "negative_prompt = \"\" #@param {type: 'string'}\n",
130 |         "width  = 512 #@param {type: 'integer'}\n",
131 |         "height  = 512 #@param {type: 'integer'}\n",
132 |         "guidance_scale = 7.5 #@param {type: 'number'}\n",
133 |         "num_inference_steps = 50 #@param {type: 'integer'}\n",
134 |         "suffix = \"jpg\" #@param [\"jpg\", \"png\"]\n",
135 |         "by = \"camenduru\" #@param {type: 'string'}\n",
136 |         "template = {\n",
137 |         "    \"discord_token\": discord_token,\n",
138 |         "    \"discord_channel_id\": discord_channel_id,\n",
139 |         "    \"by\": by,\n",
140 |         "    \"num_inference_steps\": num_inference_steps,\n",
141 |         "    \"guidance_scale\": guidance_scale,\n",
142 |         "    \"sampler\": \"PLMS\",\n",
143 |         "    \"width\": width,\n",
144 |         "    \"height\": height,\n",
145 |         "    \"prompt\": prompt,\n",
146 |         "    \"negative_prompt\": negative_prompt,\n",
147 |         "    \"suffix\": suffix\n",
148 |         "  }\n",
149 |         "with open(\"template.json\", \"w\") as outfile:\n",
150 |         "    outfile.write(json.dumps(template))\n",
151 |         "\n",
152 |         "is_loop = False #@param {type:\"boolean\"}\n",
153 |         "\n",
154 |         "if is_loop:\n",
155 |         "  while True:\n",
156 |         "      if name < folder_max_files:\n",
157 |         "          with open(\"template.json\", \"r\") as file:\n",
158 |         "              prompts = file.readlines()\n",
159 |         "          for prompt in prompts:\n",
160 |         "              d = json.loads(prompt)\n",
161 |         "              name += 1\n",
162 |         "              generate(d[\"discord_token\"], d[\"discord_channel_id\"], \"camenduru\", d[\"by\"], d[\"num_inference_steps\"], d[\"guidance_scale\"], d[\"sampler\"], d[\"width\"], d[\"height\"], d[\"prompt\"], d[\"negative_prompt\"], d[\"suffix\"], image_folder, name)\n",
163 |         "      else:\n",
164 |         "          image_folder += 1\n",
165 |         "          if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
166 |         "              os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
167 |         "          name = 0\n",
168 |         "else:\n",
169 |         "  if name < folder_max_files:\n",
170 |         "      with open(\"template.json\", \"r\") as file:\n",
171 |         "          prompts = file.readlines()\n",
172 |         "      for prompt in prompts:\n",
173 |         "          d = json.loads(prompt)\n",
174 |         "          name += 1\n",
175 |         "          generate(d[\"discord_token\"], d[\"discord_channel_id\"], \"camenduru\", d[\"by\"], d[\"num_inference_steps\"], d[\"guidance_scale\"], d[\"sampler\"], d[\"width\"], d[\"height\"], d[\"prompt\"], d[\"negative_prompt\"], d[\"suffix\"], image_folder, name)\n",
176 |         "  else:\n",
177 |         "      image_folder += 1\n",
178 |         "      if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
179 |         "          os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
180 |         "      name = 0"
181 |       ]
182 |     }
183 |   ],
184 |   "metadata": {
185 |     "accelerator": "TPU",
186 |     "colab": {
187 |       "provenance": []
188 |     },
189 |     "gpuClass": "standard",
190 |     "kernelspec": {
191 |       "display_name": "Python 3",
192 |       "name": "python3"
193 |     },
194 |     "language_info": {
195 |       "name": "python"
196 |     }
197 |   },
198 |   "nbformat": 4,
199 |   "nbformat_minor": 0
200 | }
201 | 


--------------------------------------------------------------------------------
/flax_stable_diffusion_2.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |   "cells": [
  3 |     {
  4 |       "cell_type": "markdown",
  5 |       "metadata": {
  6 |         "id": "view-in-github"
  7 |       },
  8 |       "source": [
  9 |         "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/flax_stable_diffusion_2.ipynb)"
 10 |       ]
 11 |     },
 12 |     {
 13 |       "cell_type": "code",
 14 |       "execution_count": null,
 15 |       "metadata": {
 16 |         "cellView": "form",
 17 |         "id": "ymcDoPQn41U4"
 18 |       },
 19 |       "outputs": [],
 20 |       "source": [
 21 |         "save_to = \"here\" #@param [\"here\", \"discord\", \"gdrive\"]\n",
 22 |         "if save_to == \"gdrive\":\n",
 23 |         "  from google.colab import drive\n",
 24 |         "  drive.mount('/content/gdrive')\n",
 25 |         "\n",
 26 |         "!pip install -U jax==0.3.25 jaxlib==0.3.25 flax==0.6.2 transformers piexif fold_to_ascii ftfy\n",
 27 |         "!pip install -U diffusers==0.10.0\n",
 28 |         "\n",
 29 |         "import jax.tools.colab_tpu\n",
 30 |         "jax.tools.colab_tpu.setup_tpu('tpu_driver_nightly')\n",
 31 |         "\n",
 32 |         "import jax, random, os, gc, requests, json, piexif\n",
 33 |         "import numpy as np\n",
 34 |         "from flax.jax_utils import replicate\n",
 35 |         "from flax.training.common_utils import shard\n",
 36 |         "from diffusers import FlaxStableDiffusionPipeline\n",
 37 |         "from PIL import Image\n",
 38 |         "from PIL.PngImagePlugin import PngInfo\n",
 39 |         "from fold_to_ascii import fold\n",
 40 |         "metadata = PngInfo()\n",
 41 |         "\n",
 42 |         "def closestNumber(n, m):\n",
 43 |         "    q = int(n / m)\n",
 44 |         "    n1 = m * q\n",
 45 |         "    if (n * m) > 0:\n",
 46 |         "        n2 = m * (q + 1)\n",
 47 |         "    else:\n",
 48 |         "        n2 = m * (q - 1)\n",
 49 |         "    if abs(n - n1) < abs(n - n2):\n",
 50 |         "        return n1\n",
 51 |         "    return n2\n",
 52 |         "\n",
 53 |         "def image_grid(imgs, rows, cols):\n",
 54 |         "    w,h = imgs[0].size\n",
 55 |         "    grid = Image.new('RGB', size=(cols*w, rows*h))\n",
 56 |         "    for i, img in enumerate(imgs): grid.paste(img, box=(i%cols*w, i//cols*h))\n",
 57 |         "    return grid\n",
 58 |         "\n",
 59 |         "folder_max_files = 500 #@param {type: 'integer'}\n",
 60 |         "root_folder = 'ai_images' #@param {type: 'string'}\n",
 61 |         "if save_to == \"gdrive\":\n",
 62 |         "  root_folder = f\"/content/gdrive/MyDrive/{root_folder}\"\n",
 63 |         "\n",
 64 |         "if os.path.exists(f\"{root_folder}\") == False:\n",
 65 |         "    os.mkdir(f\"{root_folder}\")\n",
 66 |         "image_folder = max([int(f) for f in os.listdir(f\"{root_folder}\")], default=0)\n",
 67 |         "if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
 68 |         "    os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
 69 |         "name = max([int(f[: f.index(\".\")]) for f in os.listdir(f\"{root_folder}/{image_folder:04}\")],default=0,)\n",
 70 |         "\n",
 71 |         "model_folder = \"flax/stable-diffusion-2\" #@param [\"flax/stable-diffusion-2\", \"flax/stable-diffusion-2-base\"] {allow-input: true}\n",
 72 |         "revision = \"main\"\n",
 73 |         "pipe, params = FlaxStableDiffusionPipeline.from_pretrained(model_folder, revision=revision, from_pt=False, dtype=jax.numpy.bfloat16, safety_checker=None)\n",
 74 |         "params = replicate(params)\n",
 75 |         "\n",
 76 |         "def generate(discord_token, discord_channel_id, discord_user, by, num_inference_steps, guidance_scale, sampler, width, height, prompt, negative_prompt, suffix, image_folder, name):\n",
 77 |         "    width = closestNumber(width, 8)\n",
 78 |         "    height = closestNumber(height, 8)\n",
 79 |         "    metadata.add_text(\"Prompt\", f\"{prompt}\")\n",
 80 |         "    metadata.add_text(\"by\", f\"{by}\")\n",
 81 |         "    gc.collect()\n",
 82 |         "    real_seed = random.randint(0, 2147483647)\n",
 83 |         "    prng_seed = jax.random.PRNGKey(real_seed)\n",
 84 |         "    prng_seed = jax.random.split(prng_seed, jax.device_count())\n",
 85 |         "    num_samples = jax.device_count()\n",
 86 |         "    prompt_n = num_samples * [prompt]\n",
 87 |         "    prompt_ids = pipe.prepare_inputs(prompt_n)\n",
 88 |         "    prompt_ids = shard(prompt_ids)\n",
 89 |         "    negative_prompt_n = num_samples * [negative_prompt]\n",
 90 |         "    negative_prompt_ids = pipe.prepare_inputs(negative_prompt_n)\n",
 91 |         "    negative_prompt_ids = shard(negative_prompt_ids)\n",
 92 |         "    images = pipe(prompt_ids, params, prng_seed, neg_prompt_ids=negative_prompt_ids, num_inference_steps=num_inference_steps, height=height, width=width, guidance_scale=guidance_scale, jit=True).images\n",
 93 |         "    images = pipe.numpy_to_pil(np.asarray(images.reshape((num_samples,) + images.shape[-3:])))\n",
 94 |         "    image = image_grid(images, 2, 4)\n",
 95 |         "    if(suffix == 'png'):\n",
 96 |         "      image.save(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", pnginfo=metadata)\n",
 97 |         "    else:\n",
 98 |         "      zeroth_ifd = {piexif.ImageIFD.ImageDescription: f\"{fold(prompt)}\", piexif.ImageIFD.Make: f\"{fold(by)}\", piexif.ImageIFD.Model: f\"{model_folder}\"}\n",
 99 |         "      exif_dict = {\"0th\": zeroth_ifd}\n",
100 |         "      exif_bytes = piexif.dump(exif_dict)\n",
101 |         "      image.save(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", \"JPEG\", quality=70, exif=exif_bytes)\n",
102 |         "    files = {f\"{image_folder:04}_{name:04}.{suffix}\": open(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", \"rb\").read()}\n",
103 |         "    if save_to == \"discord\":\n",
104 |         "      payload = {\"content\": f\"{prompt}\\nNegative prompt: {negative_prompt}\\nSteps: {num_inference_steps}, Sampler: {sampler}, CFG scale: {guidance_scale}, Seed: {real_seed}, Size: {width}x{height}, Model folder: {model_folder} - {discord_user}\"}\n",
105 |         "      requests.post(f\"https://discord.com/api/v9/channels/{discord_channel_id}/messages\", data=payload, headers={\"authorization\": f\"Bot {discord_token}\"}, files=files)\n",
106 |         "      os.remove(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\")"
107 |       ]
108 |     },
109 |     {
110 |       "cell_type": "code",
111 |       "execution_count": null,
112 |       "metadata": {
113 |         "cellView": "form",
114 |         "id": "DDUev8TOMxJ4"
115 |       },
116 |       "outputs": [],
117 |       "source": [
118 |         "discord_token = \"token\" #@param {type: 'string'}\n",
119 |         "discord_channel_id = 0 #@param {type: 'integer'}\n",
120 |         "prompt = \"duck\" #@param {type: 'string'}\n",
121 |         "negative_prompt = \"\" #@param {type: 'string'}\n",
122 |         "width  = 512 #@param {type: 'integer'}\n",
123 |         "height  = 512 #@param {type: 'integer'}\n",
124 |         "guidance_scale = 7.5 #@param {type: 'number'}\n",
125 |         "num_inference_steps = 50 #@param {type: 'integer'}\n",
126 |         "suffix = \"jpg\" #@param [\"jpg\", \"png\"]\n",
127 |         "by = \"camenduru\" #@param {type: 'string'}\n",
128 |         "template = {\n",
129 |         "    \"discord_token\": discord_token,\n",
130 |         "    \"discord_channel_id\": discord_channel_id,\n",
131 |         "    \"by\": by,\n",
132 |         "    \"num_inference_steps\": num_inference_steps,\n",
133 |         "    \"guidance_scale\": guidance_scale,\n",
134 |         "    \"sampler\": \"PLMS\",\n",
135 |         "    \"width\": width,\n",
136 |         "    \"height\": height,\n",
137 |         "    \"prompt\": prompt,\n",
138 |         "    \"negative_prompt\": negative_prompt,\n",
139 |         "    \"suffix\": suffix\n",
140 |         "  }\n",
141 |         "with open(\"template.json\", \"w\") as outfile:\n",
142 |         "    outfile.write(json.dumps(template))\n",
143 |         "\n",
144 |         "is_loop = False #@param {type:\"boolean\"}\n",
145 |         "\n",
146 |         "if is_loop:\n",
147 |         "  while True:\n",
148 |         "      if name < folder_max_files:\n",
149 |         "          with open(\"template.json\", \"r\") as file:\n",
150 |         "              prompts = file.readlines()\n",
151 |         "          for prompt in prompts:\n",
152 |         "              d = json.loads(prompt)\n",
153 |         "              name += 1\n",
154 |         "              generate(d[\"discord_token\"], d[\"discord_channel_id\"], \"camenduru\", d[\"by\"], d[\"num_inference_steps\"], d[\"guidance_scale\"], d[\"sampler\"], d[\"width\"], d[\"height\"], d[\"prompt\"], d[\"negative_prompt\"], d[\"suffix\"], image_folder, name)\n",
155 |         "      else:\n",
156 |         "          image_folder += 1\n",
157 |         "          if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
158 |         "              os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
159 |         "          name = 0\n",
160 |         "else:\n",
161 |         "  if name < folder_max_files:\n",
162 |         "      with open(\"template.json\", \"r\") as file:\n",
163 |         "          prompts = file.readlines()\n",
164 |         "      for prompt in prompts:\n",
165 |         "          d = json.loads(prompt)\n",
166 |         "          name += 1\n",
167 |         "          generate(d[\"discord_token\"], d[\"discord_channel_id\"], \"camenduru\", d[\"by\"], d[\"num_inference_steps\"], d[\"guidance_scale\"], d[\"sampler\"], d[\"width\"], d[\"height\"], d[\"prompt\"], d[\"negative_prompt\"], d[\"suffix\"], image_folder, name)\n",
168 |         "  else:\n",
169 |         "      image_folder += 1\n",
170 |         "      if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
171 |         "          os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
172 |         "      name = 0"
173 |       ]
174 |     }
175 |   ],
176 |   "metadata": {
177 |     "accelerator": "TPU",
178 |     "colab": {
179 |       "provenance": []
180 |     },
181 |     "gpuClass": "standard",
182 |     "kernelspec": {
183 |       "display_name": "Python 3",
184 |       "name": "python3"
185 |     },
186 |     "language_info": {
187 |       "name": "python"
188 |     }
189 |   },
190 |   "nbformat": 4,
191 |   "nbformat_minor": 0
192 | }
193 | 


--------------------------------------------------------------------------------
/flax_stable_diffusion_3.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |   "cells": [
  3 |     {
  4 |       "cell_type": "markdown",
  5 |       "metadata": {
  6 |         "id": "view-in-github"
  7 |       },
  8 |       "source": [
  9 |         "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/flax_stable_diffusion_3.ipynb)"
 10 |       ]
 11 |     },
 12 |     {
 13 |       "cell_type": "code",
 14 |       "execution_count": null,
 15 |       "metadata": {},
 16 |       "outputs": [],
 17 |       "source": [
 18 |         "#@title Install - Step 1\n",
 19 |         "!pip install -q jax==0.3.25 jaxlib==0.3.25 flax==0.6.2 transformers accelerate omegaconf\n",
 20 |         "!pip install -q git+https://github.com/huggingface/diffusers\n",
 21 |         "\n",
 22 |         "!apt -y install -qq aria2\n",
 23 |         "!aria2c --console-log-level=error -c -x 16 -s 16 -k 1M 'https://civitai.com/api/download/models/138176?type=Model&format=SafeTensor&size=pruned&fp=fp16' -d /content/models -o cyberrealistic_v33.safetensors\n",
 24 |         "# !aria2c --console-log-level=error -c -x 16 -s 16 -k 1M 'https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_canny.pth' -d /content/models/controlnet -o control_v11p_sd15_canny.pth\n",
 25 |         "# !aria2c --console-log-level=error -c -x 16 -s 16 -k 1M 'https://huggingface.co/ckpt/ControlNet-v1-1/resolve/main/control_v11p_sd15_canny_fp16.safetensors' -d /content/models/controlnet -o control_v11p_sd15_canny_fp16.safetensors"
 26 |       ]
 27 |     },
 28 |     {
 29 |       "cell_type": "code",
 30 |       "execution_count": null,
 31 |       "metadata": {},
 32 |       "outputs": [],
 33 |       "source": [
 34 |         "#@title Install - Step 2\n",
 35 |         "import jax, random, gc, torch\n",
 36 |         "from flax.jax_utils import replicate\n",
 37 |         "from flax.training.common_utils import shard\n",
 38 |         "from diffusers import StableDiffusionPipeline\n",
 39 |         "\n",
 40 |         "import numpy as np\n",
 41 |         "from PIL import Image\n",
 42 |         "\n",
 43 |         "import jax.tools.colab_tpu\n",
 44 |         "jax.tools.colab_tpu.setup_tpu('tpu_driver_nightly')\n",
 45 |         "jax.device_count()\n",
 46 |         "\n",
 47 |         "def image_grid(imgs, rows, cols):\n",
 48 |         "    w,h = imgs[0].size\n",
 49 |         "    grid = Image.new('RGB', size=(cols*w, rows*h))\n",
 50 |         "    for i, img in enumerate(imgs): grid.paste(img, box=(i%cols*w, i//cols*h))\n",
 51 |         "    return grid"
 52 |       ]
 53 |     },
 54 |     {
 55 |       "cell_type": "code",
 56 |       "execution_count": null,
 57 |       "metadata": {},
 58 |       "outputs": [],
 59 |       "source": [
 60 |         "#@title Vanilla - Step 1\n",
 61 |         "pipeline = StableDiffusionPipeline.from_single_file('/content/models/cyberrealistic_v33.safetensors', torch_dtype=torch.float16, safety_checker=None)\n",
 62 |         "pipeline.save_pretrained('/content/models/torch/cyberrealistic', safe_serialization=False)"
 63 |       ]
 64 |     },
 65 |     {
 66 |       "cell_type": "code",
 67 |       "execution_count": null,
 68 |       "metadata": {},
 69 |       "outputs": [],
 70 |       "source": [
 71 |         "#@title Vanilla - Step 2\n",
 72 |         "from diffusers import FlaxStableDiffusionPipeline\n",
 73 |         "\n",
 74 |         "pipeline, params = FlaxStableDiffusionPipeline.from_pretrained('/content/models/torch/cyberrealistic', from_pt=True, dtype=jax.numpy.bfloat16, safety_checker=None)\n",
 75 |         "params = replicate(params)"
 76 |       ]
 77 |     },
 78 |     {
 79 |       "cell_type": "code",
 80 |       "execution_count": null,
 81 |       "metadata": {},
 82 |       "outputs": [],
 83 |       "source": [
 84 |         "#@title Vanilla - Step 3\n",
 85 |         "gc.collect()\n",
 86 |         "\n",
 87 |         "prompt = \"cute duck\"\n",
 88 |         "negative_prompt = \"blurry\"\n",
 89 |         "\n",
 90 |         "num_samples = jax.device_count()\n",
 91 |         "\n",
 92 |         "prompt_n = num_samples * [prompt]\n",
 93 |         "prompt_ids = pipeline.prepare_inputs(prompt_n)\n",
 94 |         "prompt_ids = shard(prompt_ids)\n",
 95 |         "\n",
 96 |         "negative_prompt_n = num_samples * [negative_prompt]\n",
 97 |         "negative_prompt_ids = pipeline.prepare_inputs(negative_prompt_n)\n",
 98 |         "negative_prompt_ids = shard(negative_prompt_ids)\n",
 99 |         "\n",
100 |         "real_seed = random.randint(0, 2147483647)\n",
101 |         "prng_seed = jax.random.PRNGKey(real_seed)\n",
102 |         "prng_seed = jax.random.split(prng_seed, jax.device_count())\n",
103 |         "\n",
104 |         "images = pipeline(prompt_ids, params, prng_seed, neg_prompt_ids=negative_prompt_ids, num_inference_steps=50, height=512, width=512, guidance_scale=7, jit=True).images\n",
105 |         "images = pipeline.numpy_to_pil(np.asarray(images.reshape((num_samples,) + images.shape[-3:])))\n",
106 |         "\n",
107 |         "image = image_grid(images, 2, 4)\n",
108 |         "image"
109 |       ]
110 |     },
111 |     {
112 |       "cell_type": "code",
113 |       "execution_count": null,
114 |       "metadata": {},
115 |       "outputs": [],
116 |       "source": [
117 |         "#@title ControlNet #Canny - Step 1\n",
118 |         "import jax, torch\n",
119 |         "from diffusers import FlaxStableDiffusionControlNetPipeline, ControlNetModel\n",
120 |         "# controlnet = ControlNetModel.from_single_file('/content/models/controlnet/control_v11p_sd15_canny.pth', dtype=torch.float16) # invalid load key, '<'.\n",
121 |         "controlnet = ControlNetModel.from_single_file('https://huggingface.co/lllyasviel/ControlNet-v1-1/blob/main/control_v11p_sd15_canny.pth', dtype=torch.float16)\n",
122 |         "pipeline, params = FlaxStableDiffusionControlNetPipeline.from_pretrained('/content/models/torch/cyberrealistic', controlnet=controlnet, dtype=jax.numpy.bfloat16, from_pt=True, safety_checker=None)\n",
123 |         "params = replicate(params)"
124 |       ]
125 |     },
126 |     {
127 |       "cell_type": "code",
128 |       "execution_count": null,
129 |       "metadata": {},
130 |       "outputs": [],
131 |       "source": [
132 |         "#@title ControlNet #Canny - Step 2\n",
133 |         "from diffusers.utils import load_image\n",
134 |         "image = load_image(\"https://hf.co/datasets/huggingface/documentation-images/resolve/main/diffusers/input_image_vermeer.png\")\n",
135 |         "image_raw = image\n",
136 |         "\n",
137 |         "import cv2\n",
138 |         "from PIL import Image\n",
139 |         "import numpy as np\n",
140 |         "image = np.array(image)\n",
141 |         "low_threshold = 100\n",
142 |         "high_threshold = 200\n",
143 |         "image = cv2.Canny(image, low_threshold, high_threshold)\n",
144 |         "image = image[:, :, None]\n",
145 |         "image = np.concatenate([image, image, image], axis=2)\n",
146 |         "canny_image = Image.fromarray(image)\n",
147 |         "image = image_grid([image_raw, canny_image], 1, 2)\n",
148 |         "image"
149 |       ]
150 |     },
151 |     {
152 |       "cell_type": "code",
153 |       "execution_count": null,
154 |       "metadata": {},
155 |       "outputs": [],
156 |       "source": [
157 |         "#@title ControlNet #Canny - Step 3\n",
158 |         "gc.collect()\n",
159 |         "\n",
160 |         "prompt = \"cute duck\"\n",
161 |         "negative_prompt = \"blurry, ugly\"\n",
162 |         "\n",
163 |         "num_samples = jax.device_count()\n",
164 |         "\n",
165 |         "prompt_n = num_samples * [prompt]\n",
166 |         "prompt_ids = pipeline.prepare_inputs(prompt_n)\n",
167 |         "prompt_ids = shard(prompt_ids)\n",
168 |         "\n",
169 |         "negative_prompt_n = num_samples * [negative_prompt]\n",
170 |         "negative_prompt_ids = pipeline.prepare_inputs(negative_prompt_n)\n",
171 |         "negative_prompt_ids = shard(negative_prompt_ids)\n",
172 |         "\n",
173 |         "real_seed = random.randint(0, 2147483647)\n",
174 |         "prng_seed = jax.random.PRNGKey(real_seed)\n",
175 |         "prng_seed = jax.random.split(prng_seed, jax.device_count())\n",
176 |         "\n",
177 |         "images = pipeline(prompt_ids, canny_image, params, prng_seed, neg_prompt_ids=negative_prompt_ids, num_inference_steps=50, height=512, width=512, guidance_scale=7, jit=True).images\n",
178 |         "images = pipeline.numpy_to_pil(np.asarray(images.reshape((num_samples,) + images.shape[-3:])))\n",
179 |         "\n",
180 |         "image = image_grid(images, 2, 4)\n",
181 |         "image"
182 |       ]
183 |     }
184 |   ],
185 |   "metadata": {
186 |     "accelerator": "TPU",
187 |     "colab": {
188 |       "provenance": []
189 |     },
190 |     "gpuClass": "standard",
191 |     "kernelspec": {
192 |       "display_name": "Python 3",
193 |       "name": "python3"
194 |     },
195 |     "language_info": {
196 |       "name": "python"
197 |     }
198 |   },
199 |   "nbformat": 4,
200 |   "nbformat_minor": 0
201 | }
202 | 


--------------------------------------------------------------------------------
/pytorch_freeu_stable_diffusion.ipynb:
--------------------------------------------------------------------------------
 1 | {
 2 |   "cells": [
 3 |     {
 4 |       "cell_type": "markdown",
 5 |       "metadata": {
 6 |         "id": "view-in-github"
 7 |       },
 8 |       "source": [
 9 |         "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/pytorch_freeu_stable_diffusion.ipynb)"
10 |       ]
11 |     },
12 |     {
13 |       "cell_type": "code",
14 |       "execution_count": null,
15 |       "metadata": {
16 |         "vscode": {
17 |           "languageId": "plaintext"
18 |         }
19 |       },
20 |       "outputs": [],
21 |       "source": [
22 |         "# Main Repo: https://github.com/lyn-rgb/FreeU_Diffusers\n",
23 |         "\n",
24 |         "%cd /content\n",
25 |         "!pip install -q transformers accelerate\n",
26 |         "!pip install -q diffusers==0.21.2\n",
27 |         "# !pip install -q diffusers==0.19.3\n",
28 |         "!rm -rf /content/FreeU_Diffusers\n",
29 |         "!git clone -b diffusers-v0.21.2 https://github.com/camenduru/FreeU_Diffusers\n",
30 |         "%cd /content/FreeU_Diffusers"
31 |       ]
32 |     },
33 |     {
34 |       "cell_type": "code",
35 |       "execution_count": null,
36 |       "metadata": {
37 |         "vscode": {
38 |           "languageId": "plaintext"
39 |         }
40 |       },
41 |       "outputs": [],
42 |       "source": [
43 |         "from diffusers import StableDiffusionPipeline\n",
44 |         "import torch, sys\n",
45 |         "sys.path.append('/content/FreeU_Diffusers')\n",
46 |         "from free_lunch_utils import register_free_upblock2d, register_free_crossattn_upblock2d\n",
47 |         "\n",
48 |         "model_id = \"runwayml/stable-diffusion-v1-5\"\n",
49 |         "pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16).to(\"cuda\")\n",
50 |         "\n",
51 |         "# -------- freeu block registration\n",
52 |         "register_free_upblock2d(pipe, b1=1.2, b2=1.4, s1=0.9, s2=0.2)\n",
53 |         "register_free_crossattn_upblock2d(pipe, b1=1.2, b2=1.4, s1=0.9, s2=0.2)\n",
54 |         "# -------- freeu block registration"
55 |       ]
56 |     },
57 |     {
58 |       "cell_type": "code",
59 |       "execution_count": null,
60 |       "metadata": {
61 |         "vscode": {
62 |           "languageId": "plaintext"
63 |         }
64 |       },
65 |       "outputs": [],
66 |       "source": [
67 |         "prompt = \"a photo of an astronaut riding a horse on mars\"\n",
68 |         "image = pipe(prompt).images[0]  \n",
69 |         "image.save(\"astronaut_rides_horse.png\")"
70 |       ]
71 |     }
72 |   ],
73 |   "metadata": {
74 |     "accelerator": "TPU",
75 |     "colab": {
76 |       "provenance": []
77 |     },
78 |     "gpuClass": "standard",
79 |     "kernelspec": {
80 |       "display_name": "Python 3",
81 |       "name": "python3"
82 |     },
83 |     "language_info": {
84 |       "name": "python"
85 |     }
86 |   },
87 |   "nbformat": 4,
88 |   "nbformat_minor": 0
89 | }
90 | 


--------------------------------------------------------------------------------
/pytorch_riffusion_stable_diffusion.ipynb:
--------------------------------------------------------------------------------
 1 | {
 2 |   "cells": [
 3 |     {
 4 |       "cell_type": "markdown",
 5 |       "metadata": {
 6 |         "id": "view-in-github"
 7 |       },
 8 |       "source": [
 9 |         "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/pytorch_riffusion_stable_diffusion.ipynb)"
10 |       ]
11 |     },
12 |     {
13 |       "cell_type": "code",
14 |       "execution_count": null,
15 |       "metadata": {
16 |         "id": "ymcDoPQn41U4"
17 |       },
18 |       "outputs": [],
19 |       "source": [
20 |         "!pip install -q torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 torchtext==0.14.1 torchdata==0.5.1 --extra-index-url https://download.pytorch.org/whl/cu116 -U\n",
21 |         "!pip install -q https://github.com/camenduru/stable-diffusion-webui-colab/releases/download/0.0.15/xformers-0.0.15.dev0+189828c.d20221207-cp38-cp38-linux_x86_64.whl\n",
22 |         "!wget -q https://raw.githubusercontent.com/hmartiro/riffusion-inference/main/riffusion/audio.py\n",
23 |         "!pip install -q -U diffusers transformers ftfy pydub\n",
24 |         "\n",
25 |         "import torch, IPython\n",
26 |         "from diffusers import StableDiffusionPipeline\n",
27 |         "from audio import wav_bytes_from_spectrogram_image\n",
28 |         "\n",
29 |         "pipe = StableDiffusionPipeline.from_pretrained(\"riffusion/riffusion-model-v1\", torch_dtype=torch.float16).to(\"cuda\")\n",
30 |         "pipe.enable_xformers_memory_efficient_attention()"
31 |       ]
32 |     },
33 |     {
34 |       "cell_type": "code",
35 |       "execution_count": null,
36 |       "metadata": {
37 |         "id": "BKK30_wrQpww"
38 |       },
39 |       "outputs": [],
40 |       "source": [
41 |         "prompt = \"bongos on a havana street\" #@param {type: 'string'}\n",
42 |         "\n",
43 |         "with torch.autocast(\"cuda\"):\n",
44 |         "  image = pipe(prompt, height=512, width=512).images[0];\n",
45 |         "wav = wav_bytes_from_spectrogram_image(image)\n",
46 |         "with open(\"output.wav\", \"wb\") as f:\n",
47 |         "    f.write(wav[0].getbuffer())\n",
48 |         "IPython.display.Audio(\"/content/output.wav\", rate=44100)"
49 |       ]
50 |     }
51 |   ],
52 |   "metadata": {
53 |     "accelerator": "GPU",
54 |     "colab": {
55 |       "provenance": []
56 |     },
57 |     "gpuClass": "standard",
58 |     "kernelspec": {
59 |       "display_name": "Python 3",
60 |       "name": "python3"
61 |     },
62 |     "language_info": {
63 |       "name": "python"
64 |     }
65 |   },
66 |   "nbformat": 4,
67 |   "nbformat_minor": 0
68 | }
69 | 


--------------------------------------------------------------------------------
/pytorch_stable_diffusion.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |   "cells": [
  3 |     {
  4 |       "cell_type": "markdown",
  5 |       "metadata": {
  6 |         "id": "view-in-github"
  7 |       },
  8 |       "source": [
  9 |         "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/pytorch_stable_diffusion.ipynb)"
 10 |       ]
 11 |     },
 12 |     {
 13 |       "cell_type": "code",
 14 |       "execution_count": null,
 15 |       "metadata": {
 16 |         "cellView": "form",
 17 |         "id": "ymcDoPQn41U4"
 18 |       },
 19 |       "outputs": [],
 20 |       "source": [
 21 |         "save_to = \"here\" #@param [\"here\", \"discord\", \"gdrive\"]\n",
 22 |         "if save_to == \"gdrive\":\n",
 23 |         "  from google.colab import drive\n",
 24 |         "  drive.mount('/content/gdrive')\n",
 25 |         "\n",
 26 |         "!pip install -q torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 torchtext==0.14.1 torchdata==0.5.1 --extra-index-url https://download.pytorch.org/whl/cu116 -U\n",
 27 |         "!pip install -U diffusers transformers piexif fold_to_ascii ftfy\n",
 28 |         "!pip install -q https://github.com/camenduru/stable-diffusion-webui-colab/releases/download/0.0.15/xformers-0.0.15.dev0+189828c.d20221207-cp38-cp38-linux_x86_64.whl\n",
 29 |         "\n",
 30 |         "import torch, os, gc, requests, json, piexif\n",
 31 |         "from diffusers import StableDiffusionPipeline\n",
 32 |         "from PIL import Image\n",
 33 |         "from PIL.PngImagePlugin import PngInfo\n",
 34 |         "from fold_to_ascii import fold\n",
 35 |         "metadata = PngInfo()\n",
 36 |         "\n",
 37 |         "def closestNumber(n, m):\n",
 38 |         "    q = int(n / m)\n",
 39 |         "    n1 = m * q\n",
 40 |         "    if (n * m) > 0:\n",
 41 |         "        n2 = m * (q + 1)\n",
 42 |         "    else:\n",
 43 |         "        n2 = m * (q - 1)\n",
 44 |         "    if abs(n - n1) < abs(n - n2):\n",
 45 |         "        return n1\n",
 46 |         "    return n2\n",
 47 |         "\n",
 48 |         "folder_max_files = 500 #@param {type: 'integer'}\n",
 49 |         "root_folder = \"ai_images\" #@param {type: 'string'}\n",
 50 |         "if save_to == \"gdrive\":\n",
 51 |         "  root_folder = f\"/content/gdrive/MyDrive/{root_folder}\"\n",
 52 |         "\n",
 53 |         "if os.path.exists(f\"{root_folder}\") == False:\n",
 54 |         "    os.mkdir(f\"{root_folder}\")\n",
 55 |         "image_folder = max([int(f) for f in os.listdir(f\"{root_folder}\")], default=0)\n",
 56 |         "if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
 57 |         "    os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
 58 |         "name = max([int(f[: f.index(\".\")]) for f in os.listdir(f\"{root_folder}/{image_folder:04}\")],default=0,)\n",
 59 |         "\n",
 60 |         "model_folder = \"camenduru/sd15\" #@param [\"camenduru/sd15\", \"Linaqruf/anything-v3.0\", \"lilpotat/sa\", \"Fictiverse/Stable_Diffusion_PaperCut_Model\", \"nitrosocke/redshift-diffusion\", \"prompthero/midjourney-v4-diffusion\", \"dallinmackay/Tron-Legacy-diffusion\", \"DGSpitzer/Cyberpunk-Anime-Diffusion\", \"hakurei/waifu-diffusion\", \"nitrosocke/Arcane-Diffusion\", \"nitrosocke/classic-anim-diffusion\", \"nitrosocke/mo-di-diffusion\"] {allow-input: true}\n",
 61 |         "\n",
 62 |         "is_tile = False #@param {type: 'boolean'}\n",
 63 |         "if(is_tile):\n",
 64 |         "  def patch_conv(cls):\n",
 65 |         "    init = cls.__init__\n",
 66 |         "    def __init__(self, *args, **kwargs):\n",
 67 |         "      return init(self, *args, **kwargs, padding_mode='circular')\n",
 68 |         "    cls.__init__ = __init__\n",
 69 |         "  patch_conv(torch.nn.Conv2d)\n",
 70 |         "\n",
 71 |         "pipe = StableDiffusionPipeline.from_pretrained(model_folder, safety_checker=None).to(\"cuda\")\n",
 72 |         "pipe.enable_xformers_memory_efficient_attention()\n",
 73 |         "\n",
 74 |         "def generate(discord_token, discord_channel_id, discord_user, by, num_inference_steps, guidance_scale, sampler, width, height, prompt, negative_prompt, suffix, image_folder, name):\n",
 75 |         "    width = closestNumber(width, 8)\n",
 76 |         "    height = closestNumber(height, 8)\n",
 77 |         "    metadata.add_text(\"Prompt\", f\"{prompt}\")\n",
 78 |         "    metadata.add_text(\"by\", f\"{by}\")\n",
 79 |         "    gc.collect()\n",
 80 |         "    with torch.autocast(\"cuda\"):\n",
 81 |         "      images = pipe(prompt, negative_prompt=negative_prompt, num_inference_steps=num_inference_steps, height=height, width=width, guidance_scale=guidance_scale).images\n",
 82 |         "    image = images[0]\n",
 83 |         "    real_seed = torch.cuda.initial_seed()\n",
 84 |         "    if(suffix == 'png'):\n",
 85 |         "      image.save(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", pnginfo=metadata)\n",
 86 |         "    else:\n",
 87 |         "      zeroth_ifd = {piexif.ImageIFD.ImageDescription: f\"{fold(prompt)}\", piexif.ImageIFD.Make: f\"{fold(by)}\", piexif.ImageIFD.Model: f\"{model_folder}\"}\n",
 88 |         "      exif_dict = {\"0th\": zeroth_ifd}\n",
 89 |         "      exif_bytes = piexif.dump(exif_dict)\n",
 90 |         "      image.save(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", \"JPEG\", quality=70, exif=exif_bytes)\n",
 91 |         "    files = {f\"{image_folder:04}_{name:04}.{suffix}\": open(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", \"rb\").read()}\n",
 92 |         "    if save_to == \"discord\":\n",
 93 |         "      payload = {\"content\": f\"{prompt}\\nNegative prompt: {negative_prompt}\\nSteps: {num_inference_steps}, Sampler: {sampler}, CFG scale: {guidance_scale}, Seed: {real_seed}, Size: {width}x{height}, Model folder: {model_folder} - {discord_user}\"}\n",
 94 |         "      requests.post(f\"https://discord.com/api/v9/channels/{discord_channel_id}/messages\", data=payload, headers={\"authorization\": f\"Bot {discord_token}\"}, files=files)\n",
 95 |         "      os.remove(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\")"
 96 |       ]
 97 |     },
 98 |     {
 99 |       "cell_type": "code",
100 |       "execution_count": null,
101 |       "metadata": {
102 |         "cellView": "form",
103 |         "id": "DDUev8TOMxJ4"
104 |       },
105 |       "outputs": [],
106 |       "source": [
107 |         "discord_token = \"token\" #@param {type: 'string'}\n",
108 |         "discord_channel_id = 0 #@param {type: 'integer'}\n",
109 |         "prompt = \"duck\" #@param {type: 'string'}\n",
110 |         "negative_prompt = \"\" #@param {type: 'string'}\n",
111 |         "width  = 512 #@param {type: 'integer'}\n",
112 |         "height  = 512 #@param {type: 'integer'}\n",
113 |         "guidance_scale = 7.5 #@param {type: 'number'}\n",
114 |         "num_inference_steps = 50 #@param {type: 'integer'}\n",
115 |         "suffix = \"jpg\" #@param [\"jpg\", \"png\"]\n",
116 |         "by = \"camenduru\" #@param {type: 'string'}\n",
117 |         "template = {\n",
118 |         "    \"discord_token\": discord_token,\n",
119 |         "    \"discord_channel_id\": discord_channel_id,\n",
120 |         "    \"by\": by,\n",
121 |         "    \"num_inference_steps\": num_inference_steps,\n",
122 |         "    \"guidance_scale\": guidance_scale,\n",
123 |         "    \"sampler\": \"PLMS\",\n",
124 |         "    \"width\": width,\n",
125 |         "    \"height\": height,\n",
126 |         "    \"prompt\": prompt,\n",
127 |         "    \"negative_prompt\": negative_prompt,\n",
128 |         "    \"suffix\": suffix\n",
129 |         "  }\n",
130 |         "with open(\"template.json\", \"w\") as outfile:\n",
131 |         "    outfile.write(json.dumps(template))\n",
132 |         "\n",
133 |         "is_loop = False #@param {type:\"boolean\"}\n",
134 |         "\n",
135 |         "if is_loop:\n",
136 |         "  while True:\n",
137 |         "      if name < folder_max_files:\n",
138 |         "          with open(\"template.json\", \"r\") as file:\n",
139 |         "              prompts = file.readlines()\n",
140 |         "          for prompt in prompts:\n",
141 |         "              d = json.loads(prompt)\n",
142 |         "              name += 1\n",
143 |         "              generate(d[\"discord_token\"], d[\"discord_channel_id\"], \"camenduru\", d[\"by\"], d[\"num_inference_steps\"], d[\"guidance_scale\"], d[\"sampler\"], d[\"width\"], d[\"height\"], d[\"prompt\"], d[\"negative_prompt\"], d[\"suffix\"], image_folder, name)\n",
144 |         "      else:\n",
145 |         "          image_folder += 1\n",
146 |         "          if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
147 |         "              os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
148 |         "          name = 0\n",
149 |         "else:\n",
150 |         "  if name < folder_max_files:\n",
151 |         "      with open(\"template.json\", \"r\") as file:\n",
152 |         "          prompts = file.readlines()\n",
153 |         "      for prompt in prompts:\n",
154 |         "          d = json.loads(prompt)\n",
155 |         "          name += 1\n",
156 |         "          generate(d[\"discord_token\"], d[\"discord_channel_id\"], \"camenduru\", d[\"by\"], d[\"num_inference_steps\"], d[\"guidance_scale\"], d[\"sampler\"], d[\"width\"], d[\"height\"], d[\"prompt\"], d[\"negative_prompt\"], d[\"suffix\"], image_folder, name)\n",
157 |         "  else:\n",
158 |         "      image_folder += 1\n",
159 |         "      if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
160 |         "          os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
161 |         "      name = 0"
162 |       ]
163 |     }
164 |   ],
165 |   "metadata": {
166 |     "accelerator": "GPU",
167 |     "colab": {
168 |       "provenance": []
169 |     },
170 |     "gpuClass": "standard",
171 |     "kernelspec": {
172 |       "display_name": "Python 3",
173 |       "name": "python3"
174 |     },
175 |     "language_info": {
176 |       "name": "python"
177 |     }
178 |   },
179 |   "nbformat": 4,
180 |   "nbformat_minor": 0
181 | }
182 | 


--------------------------------------------------------------------------------
/pytorch_stable_diffusion_2.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |   "cells": [
  3 |     {
  4 |       "cell_type": "markdown",
  5 |       "metadata": {
  6 |         "id": "view-in-github"
  7 |       },
  8 |       "source": [
  9 |         "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/pytorch_stable_diffusion_2.ipynb)"
 10 |       ]
 11 |     },
 12 |     {
 13 |       "cell_type": "code",
 14 |       "execution_count": null,
 15 |       "metadata": {
 16 |         "cellView": "form",
 17 |         "id": "ymcDoPQn41U4"
 18 |       },
 19 |       "outputs": [],
 20 |       "source": [
 21 |         "save_to = \"here\" #@param [\"here\", \"discord\", \"gdrive\"]\n",
 22 |         "if save_to == \"gdrive\":\n",
 23 |         "  from google.colab import drive\n",
 24 |         "  drive.mount('/content/gdrive')\n",
 25 |         "\n",
 26 |         "!pip install -q torch==1.13.1+cu116 torchvision==0.14.1+cu116 torchaudio==0.13.1 torchtext==0.14.1 torchdata==0.5.1 --extra-index-url https://download.pytorch.org/whl/cu116 -U\n",
 27 |         "!pip install -U git+https://github.com/huggingface/diffusers\n",
 28 |         "!pip install -U transformers piexif fold_to_ascii ftfy\n",
 29 |         "!pip install -q https://github.com/camenduru/stable-diffusion-webui-colab/releases/download/0.0.15/xformers-0.0.15.dev0+189828c.d20221207-cp38-cp38-linux_x86_64.whl\n",
 30 |         "\n",
 31 |         "import torch, os, gc, requests, json, piexif\n",
 32 |         "from diffusers import StableDiffusionPipeline, EulerDiscreteScheduler\n",
 33 |         "from PIL import Image\n",
 34 |         "from PIL.PngImagePlugin import PngInfo\n",
 35 |         "from fold_to_ascii import fold\n",
 36 |         "metadata = PngInfo()\n",
 37 |         "\n",
 38 |         "def closestNumber(n, m):\n",
 39 |         "    q = int(n / m)\n",
 40 |         "    n1 = m * q\n",
 41 |         "    if (n * m) > 0:\n",
 42 |         "        n2 = m * (q + 1)\n",
 43 |         "    else:\n",
 44 |         "        n2 = m * (q - 1)\n",
 45 |         "    if abs(n - n1) < abs(n - n2):\n",
 46 |         "        return n1\n",
 47 |         "    return n2\n",
 48 |         "\n",
 49 |         "folder_max_files = 500 #@param {type: 'integer'}\n",
 50 |         "root_folder = \"ai_images\" #@param {type: 'string'}\n",
 51 |         "if save_to == \"gdrive\":\n",
 52 |         "  root_folder = f\"/content/gdrive/MyDrive/{root_folder}\"\n",
 53 |         "\n",
 54 |         "if os.path.exists(f\"{root_folder}\") == False:\n",
 55 |         "    os.mkdir(f\"{root_folder}\")\n",
 56 |         "image_folder = max([int(f) for f in os.listdir(f\"{root_folder}\")], default=0)\n",
 57 |         "if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
 58 |         "    os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
 59 |         "name = max([int(f[: f.index(\".\")]) for f in os.listdir(f\"{root_folder}/{image_folder:04}\")],default=0,)\n",
 60 |         "\n",
 61 |         "model_folder = \"stabilityai/stable-diffusion-2\" #@param [\"stabilityai/stable-diffusion-2\", \"stabilityai/stable-diffusion-2-base\", \"stabilityai/stable-diffusion-2-inpainting\", \"stabilityai/stable-diffusion-2-depth\"] {allow-input: true}\n",
 62 |         "\n",
 63 |         "is_tile = False #@param {type: 'boolean'}\n",
 64 |         "if(is_tile):\n",
 65 |         "  def patch_conv(cls):\n",
 66 |         "    init = cls.__init__\n",
 67 |         "    def __init__(self, *args, **kwargs):\n",
 68 |         "      return init(self, *args, **kwargs, padding_mode='circular')\n",
 69 |         "    cls.__init__ = __init__\n",
 70 |         "  patch_conv(torch.nn.Conv2d)\n",
 71 |         "\n",
 72 |         "scheduler = EulerDiscreteScheduler.from_pretrained(model_folder, subfolder=\"scheduler\")\n",
 73 |         "pipe = StableDiffusionPipeline.from_pretrained(model_folder, scheduler=scheduler).to(\"cuda\")\n",
 74 |         "pipe.safety_checker = lambda images, clip_input: (images, False)\n",
 75 |         "pipe.enable_xformers_memory_efficient_attention()\n",
 76 |         "\n",
 77 |         "def generate(discord_token, discord_channel_id, discord_user, by, num_inference_steps, guidance_scale, sampler, width, height, prompt, negative_prompt, suffix, image_folder, name):\n",
 78 |         "    width = closestNumber(width, 8)\n",
 79 |         "    height = closestNumber(height, 8)\n",
 80 |         "    metadata.add_text(\"Prompt\", f\"{prompt}\")\n",
 81 |         "    metadata.add_text(\"by\", f\"{by}\")\n",
 82 |         "    gc.collect()\n",
 83 |         "    with torch.autocast(\"cuda\"):\n",
 84 |         "      images = pipe(prompt, negative_prompt=negative_prompt, num_inference_steps=num_inference_steps, height=height, width=width, guidance_scale=guidance_scale).images\n",
 85 |         "    image = images[0]\n",
 86 |         "    real_seed = torch.cuda.initial_seed()\n",
 87 |         "    if(suffix == 'png'):\n",
 88 |         "      image.save(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", pnginfo=metadata)\n",
 89 |         "    else:\n",
 90 |         "      zeroth_ifd = {piexif.ImageIFD.ImageDescription: f\"{fold(prompt)}\", piexif.ImageIFD.Make: f\"{fold(by)}\", piexif.ImageIFD.Model: f\"{model_folder}\"}\n",
 91 |         "      exif_dict = {\"0th\": zeroth_ifd}\n",
 92 |         "      exif_bytes = piexif.dump(exif_dict)\n",
 93 |         "      image.save(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", \"JPEG\", quality=70, exif=exif_bytes)\n",
 94 |         "    files = {f\"{image_folder:04}_{name:04}.{suffix}\": open(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\", \"rb\").read()}\n",
 95 |         "    if save_to == \"discord\":\n",
 96 |         "      payload = {\"content\": f\"{prompt}\\nNegative prompt: {negative_prompt}\\nSteps: {num_inference_steps}, Sampler: {sampler}, CFG scale: {guidance_scale}, Seed: {real_seed}, Size: {width}x{height}, Model folder: {model_folder} - {discord_user}\"}\n",
 97 |         "      requests.post(f\"https://discord.com/api/v9/channels/{discord_channel_id}/messages\", data=payload, headers={\"authorization\": f\"Bot {discord_token}\"}, files=files)\n",
 98 |         "      os.remove(f\"{root_folder}/{image_folder:04}/{name:04}.{suffix}\")"
 99 |       ]
100 |     },
101 |     {
102 |       "cell_type": "code",
103 |       "execution_count": null,
104 |       "metadata": {
105 |         "cellView": "form",
106 |         "id": "DDUev8TOMxJ4"
107 |       },
108 |       "outputs": [],
109 |       "source": [
110 |         "discord_token = \"\" #@param {type: 'string'}\n",
111 |         "discord_channel_id = 0 #@param {type: 'integer'}\n",
112 |         "prompt = \"duck\" #@param {type: 'string'}\n",
113 |         "negative_prompt = \"\" #@param {type: 'string'}\n",
114 |         "width  = 768 #@param {type: 'integer'}\n",
115 |         "height  = 768 #@param {type: 'integer'}\n",
116 |         "guidance_scale = 7.5 #@param {type: 'number'}\n",
117 |         "num_inference_steps = 50 #@param {type: 'integer'}\n",
118 |         "suffix = \"jpg\" #@param [\"jpg\", \"png\"]\n",
119 |         "by = \"camenduru\" #@param {type: 'string'}\n",
120 |         "template = {\n",
121 |         "    \"discord_token\": discord_token,\n",
122 |         "    \"discord_channel_id\": discord_channel_id,\n",
123 |         "    \"by\": by,\n",
124 |         "    \"num_inference_steps\": num_inference_steps,\n",
125 |         "    \"guidance_scale\": guidance_scale,\n",
126 |         "    \"sampler\": \"PLMS\",\n",
127 |         "    \"width\": width,\n",
128 |         "    \"height\": height,\n",
129 |         "    \"prompt\": prompt,\n",
130 |         "    \"negative_prompt\": negative_prompt,\n",
131 |         "    \"suffix\": suffix\n",
132 |         "  }\n",
133 |         "with open(\"template.json\", \"w\") as outfile:\n",
134 |         "    outfile.write(json.dumps(template))\n",
135 |         "\n",
136 |         "is_loop = False #@param {type:\"boolean\"}\n",
137 |         "\n",
138 |         "if is_loop:\n",
139 |         "  while True:\n",
140 |         "      if name < folder_max_files:\n",
141 |         "          with open(\"template.json\", \"r\") as file:\n",
142 |         "              prompts = file.readlines()\n",
143 |         "          for prompt in prompts:\n",
144 |         "              d = json.loads(prompt)\n",
145 |         "              name += 1\n",
146 |         "              generate(d[\"discord_token\"], d[\"discord_channel_id\"], \"camenduru\", d[\"by\"], d[\"num_inference_steps\"], d[\"guidance_scale\"], d[\"sampler\"], d[\"width\"], d[\"height\"], d[\"prompt\"], d[\"negative_prompt\"], d[\"suffix\"], image_folder, name)\n",
147 |         "      else:\n",
148 |         "          image_folder += 1\n",
149 |         "          if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
150 |         "              os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
151 |         "          name = 0\n",
152 |         "else:\n",
153 |         "  if name < folder_max_files:\n",
154 |         "      with open(\"template.json\", \"r\") as file:\n",
155 |         "          prompts = file.readlines()\n",
156 |         "      for prompt in prompts:\n",
157 |         "          d = json.loads(prompt)\n",
158 |         "          name += 1\n",
159 |         "          generate(d[\"discord_token\"], d[\"discord_channel_id\"], \"camenduru\", d[\"by\"], d[\"num_inference_steps\"], d[\"guidance_scale\"], d[\"sampler\"], d[\"width\"], d[\"height\"], d[\"prompt\"], d[\"negative_prompt\"], d[\"suffix\"], image_folder, name)\n",
160 |         "  else:\n",
161 |         "      image_folder += 1\n",
162 |         "      if os.path.exists(f\"{root_folder}/{image_folder:04}\") == False:\n",
163 |         "          os.mkdir(f\"{root_folder}/{image_folder:04}\")\n",
164 |         "      name = 0"
165 |       ]
166 |     }
167 |   ],
168 |   "metadata": {
169 |     "accelerator": "GPU",
170 |     "colab": {
171 |       "provenance": []
172 |     },
173 |     "gpuClass": "standard",
174 |     "kernelspec": {
175 |       "display_name": "Python 3",
176 |       "name": "python3"
177 |     },
178 |     "language_info": {
179 |       "name": "python"
180 |     }
181 |   },
182 |   "nbformat": 4,
183 |   "nbformat_minor": 0
184 | }
185 | 


--------------------------------------------------------------------------------
/pytorch_stable_diffusion_3.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |   "cells": [
  3 |     {
  4 |       "cell_type": "markdown",
  5 |       "metadata": {
  6 |         "id": "view-in-github"
  7 |       },
  8 |       "source": [
  9 |         "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/camenduru/stable-diffusion-diffusers-colab/blob/main/pytorch_stable_diffusion_3.ipynb)"
 10 |       ]
 11 |     },
 12 |     {
 13 |       "cell_type": "code",
 14 |       "execution_count": null,
 15 |       "metadata": {},
 16 |       "outputs": [],
 17 |       "source": [
 18 |         "#@title Install - Step 1\n",
 19 |         "!pip install -q transformers accelerate omegaconf xformers==0.0.20 opencv-contrib-python controlnet_aux\n",
 20 |         "!pip install -q git+https://github.com/huggingface/diffusers\n",
 21 |         "\n",
 22 |         "# !apt -y install -qq aria2\n",
 23 |         "# !aria2c --console-log-level=error -c -x 16 -s 16 -k 1M https://huggingface.co/ckpt/sd_xl_refiner_1.0/resolve/main/sd_xl_refiner_1.0_0.9vae.safetensors -d /content/models -o sd_xl_refiner_1.0.safetensors\n",
 24 |         "# !aria2c --console-log-level=error -c -x 16 -s 16 -k 1M https://huggingface.co/ckpt/juggernaut-xl/resolve/main/juggernautXL_version2.safetensors -d /content/models -o juggernautXL_version2.safetensors\n",
 25 |         "# !aria2c --console-log-level=error -c -x 16 -s 16 -k 1M 'https://huggingface.co/lllyasviel/ControlNet-v1-1/resolve/main/control_v11p_sd15_canny.pth' -d /content/models/controlnet -o control_v11p_sd15_canny.pth\n",
 26 |         "# !aria2c --console-log-level=error -c -x 16 -s 16 -k 1M 'https://huggingface.co/ckpt/ControlNet-v1-1/resolve/main/control_v11p_sd15_canny_fp16.safetensors' -d /content/models/controlnet -o control_v11p_sd15_canny_fp16.safetensors"
 27 |       ]
 28 |     },
 29 |     {
 30 |       "cell_type": "code",
 31 |       "execution_count": null,
 32 |       "metadata": {},
 33 |       "outputs": [],
 34 |       "source": [
 35 |         "#@title Install - Step 2\n",
 36 |         "import random, gc, torch\n",
 37 |         "\n",
 38 |         "def image_grid(imgs, rows, cols):\n",
 39 |         "    w,h = imgs[0].size\n",
 40 |         "    grid = Image.new('RGB', size=(cols*w, rows*h))\n",
 41 |         "    for i, img in enumerate(imgs): grid.paste(img, box=(i%cols*w, i//cols*h))\n",
 42 |         "    return grid"
 43 |       ]
 44 |     },
 45 |     {
 46 |       "cell_type": "code",
 47 |       "execution_count": null,
 48 |       "metadata": {},
 49 |       "outputs": [],
 50 |       "source": [
 51 |         "#@title Vanilla - Step 1\n",
 52 |         "from diffusers import StableDiffusionXLPipeline\n",
 53 |         "# pipeline = StableDiffusionXLPipeline.from_single_file(\"/content/models/juggernautXL_version2.safetensors\", torch_dtype=torch.float16, use_safetensors=True, safety_checker=None ).to(\"cuda\")\n",
 54 |         "pipeline = StableDiffusionXLPipeline.from_pretrained(\"stabilityai/stable-diffusion-xl-base-1.0\", torch_dtype=torch.float16, variant=\"fp16\", use_safetensors=True, safety_checker=None ).to(\"cuda\")\n",
 55 |         "pipeline.enable_xformers_memory_efficient_attention()"
 56 |       ]
 57 |     },
 58 |     {
 59 |       "cell_type": "code",
 60 |       "execution_count": null,
 61 |       "metadata": {},
 62 |       "outputs": [],
 63 |       "source": [
 64 |         "#@title Vanilla - Step 2\n",
 65 |         "# pipeline.load_lora_weights(\"minimaxir/sdxl-wrong-lora\")\n",
 66 |         "image = pipeline(prompt=\"a photo of a dog in a bucket\", negative_prompt=\"blurry\", num_inference_steps=50, height=1024, width=1024, guidance_scale=7).images[0]\n",
 67 |         "image"
 68 |       ]
 69 |     },
 70 |     {
 71 |       "cell_type": "code",
 72 |       "execution_count": null,
 73 |       "metadata": {},
 74 |       "outputs": [],
 75 |       "source": [
 76 |         "#@title ControlNet #Canny - Step 1\n",
 77 |         "import torch\n",
 78 |         "from diffusers import StableDiffusionXLControlNetPipeline, ControlNetModel\n",
 79 |         "# controlnet = ControlNetModel.from_single_file('/content/models/controlnet/control_v11p_sd15_canny.pth', torch_dtype=torch.float16) # invalid load key, '<'.\n",
 80 |         "# controlnet = ControlNetModel.from_single_file('https://huggingface.co/lllyasviel/ControlNet-v1-1/blob/main/control_v11p_sd15_canny.pth', torch_dtype=torch.float16).to(\"cuda\")\n",
 81 |         "controlnet = ControlNetModel.from_pretrained(\"diffusers/controlnet-canny-sdxl-1.0\", variant=\"fp16\", torch_dtype=torch.float16).to(\"cuda\")\n",
 82 |         "pipeline = StableDiffusionXLControlNetPipeline.from_pretrained(\"stabilityai/stable-diffusion-xl-base-1.0\", controlnet=controlnet, torch_dtype=torch.float16, variant=\"fp16\", use_safetensors=True).to(\"cuda\")\n",
 83 |         "# pipeline.enable_model_cpu_offload()\n",
 84 |         "pipeline.enable_xformers_memory_efficient_attention()"
 85 |       ]
 86 |     },
 87 |     {
 88 |       "cell_type": "code",
 89 |       "execution_count": null,
 90 |       "metadata": {},
 91 |       "outputs": [],
 92 |       "source": [
 93 |         "#@title ControlNet #Canny - Step 2\n",
 94 |         "from diffusers import UniPCMultistepScheduler\n",
 95 |         "pipeline.scheduler = UniPCMultistepScheduler.from_config(pipeline.scheduler.config)"
 96 |       ]
 97 |     },
 98 |     {
 99 |       "cell_type": "code",
100 |       "execution_count": null,
101 |       "metadata": {},
102 |       "outputs": [],
103 |       "source": [
104 |         "#@title ControlNet #Canny - Step 3\n",
105 |         "from diffusers.utils import load_image\n",
106 |         "image = load_image(\"https://hf.co/datasets/huggingface/documentation-images/resolve/main/diffusers/input_image_vermeer.png\")\n",
107 |         "image_raw = image\n",
108 |         "\n",
109 |         "import cv2\n",
110 |         "from PIL import Image\n",
111 |         "import numpy as np\n",
112 |         "image = np.array(image)\n",
113 |         "low_threshold = 100\n",
114 |         "high_threshold = 200\n",
115 |         "image = cv2.Canny(image, low_threshold, high_threshold)\n",
116 |         "image = image[:, :, None]\n",
117 |         "image = np.concatenate([image, image, image], axis=2)\n",
118 |         "canny_image = Image.fromarray(image)\n",
119 |         "image = image_grid([image_raw, canny_image], 1, 2)\n",
120 |         "image"
121 |       ]
122 |     },
123 |     {
124 |       "cell_type": "code",
125 |       "execution_count": null,
126 |       "metadata": {},
127 |       "outputs": [],
128 |       "source": [
129 |         "#@title ControlNet #Canny - Step 4\n",
130 |         "controlnet_conditioning_scale = 0.5\n",
131 |         "image = pipeline(prompt=\"a photo of a dog in a bucket\", controlnet_conditioning_scale=controlnet_conditioning_scale, image=canny_image, negative_prompt=\"blurry\", num_inference_steps=50, height=1024, width=1024, guidance_scale=7).images[0]\n",
132 |         "image"
133 |       ]
134 |     },
135 |     {
136 |       "cell_type": "code",
137 |       "execution_count": null,
138 |       "metadata": {},
139 |       "outputs": [],
140 |       "source": [
141 |         "#@title ControlNet #OpenPose - Step 1\n",
142 |         "import torch\n",
143 |         "from diffusers import StableDiffusionXLControlNetPipeline, ControlNetModel\n",
144 |         "controlnet = ControlNetModel.from_pretrained(\"fusing/stable-diffusion-v1-5-controlnet-openpose\", torch_dtype=torch.float16).to(\"cuda\")\n",
145 |         "pipeline = StableDiffusionXLControlNetPipeline.from_pretrained(\"stabilityai/stable-diffusion-xl-base-1.0\", controlnet=controlnet, torch_dtype=torch.float16, variant=\"fp16\", use_safetensors=True).to(\"cuda\")\n",
146 |         "# pipeline.enable_model_cpu_offload()\n",
147 |         "pipeline.enable_xformers_memory_efficient_attention()"
148 |       ]
149 |     },
150 |     {
151 |       "cell_type": "code",
152 |       "execution_count": null,
153 |       "metadata": {},
154 |       "outputs": [],
155 |       "source": [
156 |         "#@title ControlNet #Canny - Step 2\n",
157 |         "from diffusers import UniPCMultistepScheduler\n",
158 |         "pipeline.scheduler = UniPCMultistepScheduler.from_config(pipeline.scheduler.config)"
159 |       ]
160 |     },
161 |     {
162 |       "cell_type": "code",
163 |       "execution_count": null,
164 |       "metadata": {},
165 |       "outputs": [],
166 |       "source": [
167 |         "#@title ControlNet #OpenPose - Step 3\n",
168 |         "from PIL import Image\n",
169 |         "from controlnet_aux import OpenposeDetector\n",
170 |         "from diffusers.utils import load_image\n",
171 |         "image = load_image(\"https://hf.co/datasets/YiYiXu/controlnet-testing/resolve/main/yoga1.jpeg\")\n",
172 |         "image_raw = image\n",
173 |         "model = OpenposeDetector.from_pretrained(\"lllyasviel/ControlNet\")\n",
174 |         "pose = model(image)\n",
175 |         "# image = image_grid([image_raw, pose], 1, 2)\n",
176 |         "# image\n",
177 |         "pose"
178 |       ]
179 |     },
180 |     {
181 |       "cell_type": "code",
182 |       "execution_count": null,
183 |       "metadata": {},
184 |       "outputs": [],
185 |       "source": [
186 |         "#@title ControlNet #OpenPose - Step 4\n",
187 |         "controlnet_conditioning_scale = 0.5\n",
188 |         "image = pipeline(prompt=\"a photo of a dog in a bucket\", controlnet_conditioning_scale=controlnet_conditioning_scale, image=pose, negative_prompt=\"blurry\", num_inference_steps=50, height=1024, width=1024, guidance_scale=7).images[0]\n",
189 |         "image"
190 |       ]
191 |     }
192 |   ],
193 |   "metadata": {
194 |     "accelerator": "TPU",
195 |     "colab": {
196 |       "provenance": []
197 |     },
198 |     "gpuClass": "standard",
199 |     "kernelspec": {
200 |       "display_name": "Python 3",
201 |       "name": "python3"
202 |     },
203 |     "language_info": {
204 |       "name": "python"
205 |     }
206 |   },
207 |   "nbformat": 4,
208 |   "nbformat_minor": 0
209 | }
210 | 


--------------------------------------------------------------------------------