File size: 6,318 Bytes
7aaaf62
 
001a426
 
f30235e
399a445
 
83721e5
7aaaf62
 
 
 
 
 
001a426
f30235e
7aaaf62
 
 
 
f30235e
7aaaf62
 
 
 
001a426
 
7aaaf62
 
 
 
f30235e
7aaaf62
 
f30235e
7aaaf62
 
 
 
 
 
f30235e
7aaaf62
 
f30235e
7aaaf62
96e542f
7aaaf62
 
6caf083
 
 
 
 
 
 
 
7aaaf62
 
f30235e
7aaaf62
 
f30235e
7aaaf62
 
001a426
 
7aaaf62
 
 
 
f30235e
6caf083
7aaaf62
 
 
 
6caf083
 
7aaaf62
6caf083
 
83721e5
7aaaf62
 
 
 
f30235e
6caf083
7aaaf62
f30235e
7aaaf62
83721e5
7aaaf62
 
 
 
6caf083
 
001a426
 
7aaaf62
 
6caf083
 
 
 
 
 
 
 
 
7aaaf62
 
f30235e
6caf083
7aaaf62
f30235e
399a445
83721e5
6caf083
83721e5
 
 
6caf083
 
001a426
 
399a445
 
6caf083
 
 
 
 
 
 
 
399a445
 
f30235e
83721e5
399a445
f30235e
6caf083
 
 
 
 
 
 
 
f30235e
6caf083
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f30235e
6caf083
 
f30235e
6caf083
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7aaaf62
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b451ab22",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import random\n",
    "import numpy as np\n",
    "from PIL import Image\n",
    "from datasets import load_dataset\n",
    "from IPython.display import Audio\n",
    "from diffusers import AutoencoderKL, AudioDiffusionPipeline, Mel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "324cef44",
   "metadata": {},
   "outputs": [],
   "source": [
    "mel = Mel()\n",
    "vae = AutoencoderKL.from_pretrained('../models/autoencoder-kl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "da55ce79",
   "metadata": {},
   "outputs": [],
   "source": [
    "vae.config"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "5fea99ff",
   "metadata": {},
   "outputs": [],
   "source": [
    "ds = load_dataset('teticio/audio-diffusion-256')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3a65ec4d",
   "metadata": {},
   "source": [
    "### Reconstruct audio"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "426c6edd",
   "metadata": {},
   "outputs": [],
   "source": [
    "image = random.choice(ds['train'])['image']\n",
    "display(image)\n",
    "Audio(data=mel.image_to_audio(image), rate=mel.get_sample_rate())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "29c9011d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# encode\n",
    "input_image = np.frombuffer(image.tobytes(), dtype=\"uint8\").reshape(\n",
    "    (image.height, image.width, 1))\n",
    "input_image = ((input_image / 255) * 2 - 1).transpose(2, 0, 1)\n",
    "posterior = vae.encode(torch.tensor([input_image],\n",
    "                                    dtype=torch.float32)).latent_dist\n",
    "latents = posterior.sample()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "323ba46d",
   "metadata": {},
   "outputs": [],
   "source": [
    "# reconstruct\n",
    "output_image = vae.decode(latents)['sample']\n",
    "output_image = torch.clamp(output_image, -1., 1.)\n",
    "output_image = (output_image + 1.0) / 2.0  # -1,1 -> 0,1; c,h,w\n",
    "output_image = (output_image.detach().cpu().numpy() *\n",
    "                255).round().astype(\"uint8\").transpose(0, 2, 3, 1)[0, :, :, 0]\n",
    "output_image = Image.fromarray(output_image)\n",
    "display(output_image)\n",
    "Audio(data=mel.image_to_audio(output_image), rate=mel.get_sample_rate())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "00ff2ffa",
   "metadata": {},
   "source": [
    "### Random sample from latent space\n",
    "(Don't expect interesting results!)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "156a06a2",
   "metadata": {},
   "outputs": [],
   "source": [
    "# sample\n",
    "output_image = vae.decode(torch.randn_like(latents))['sample']\n",
    "output_image = torch.clamp(output_image, -1., 1.)\n",
    "output_image = (output_image + 1.0) / 2.0  # -1,1 -> 0,1; c,h,w\n",
    "output_image = (output_image.detach().cpu().numpy() *\n",
    "                255).round().astype(\"uint8\").transpose(0, 2, 3, 1)[0, :, :, 0]\n",
    "output_image = Image.fromarray(output_image)\n",
    "display(output_image)\n",
    "Audio(data=mel.image_to_audio(output_image), rate=mel.get_sample_rate())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ee3997cf",
   "metadata": {},
   "source": [
    "### Interpolate between two audios in latent space"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "46019770",
   "metadata": {},
   "outputs": [],
   "source": [
    "image2 = random.choice(ds['train'])['image']\n",
    "display(image2)\n",
    "Audio(data=mel.image_to_audio(image2), rate=mel.get_sample_rate())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e6552b19",
   "metadata": {},
   "outputs": [],
   "source": [
    "# encode\n",
    "input_image2 = np.frombuffer(image2.tobytes(), dtype=\"uint8\").reshape(\n",
    "    (image2.height, image2.width, 1))\n",
    "input_image2 = ((input_image2 / 255) * 2 - 1).transpose(2, 0, 1)\n",
    "posterior2 = vae.encode(torch.tensor([input_image2],\n",
    "                                     dtype=torch.float32)).latent_dist\n",
    "latents2 = posterior2.sample()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "060a0b25",
   "metadata": {},
   "outputs": [],
   "source": [
    "# interpolate\n",
    "alpha = 0.5  #@param {type:\"slider\", min:0, max:1, step:0.1}\n",
    "output_image = vae.decode(\n",
    "    AudioDiffusionPipeline.slerp(latents, latents2, alpha))['sample']\n",
    "output_image = torch.clamp(output_image, -1., 1.)\n",
    "output_image = (output_image + 1.0) / 2.0  # -1,1 -> 0,1; c,h,w\n",
    "output_image = (output_image.detach().cpu().numpy() *\n",
    "                255).round().astype(\"uint8\").transpose(0, 2, 3, 1)[0, :, :, 0]\n",
    "output_image = Image.fromarray(output_image)\n",
    "display(output_image)\n",
    "display(Audio(data=mel.image_to_audio(image), rate=mel.get_sample_rate()))\n",
    "display(Audio(data=mel.image_to_audio(image2), rate=mel.get_sample_rate()))\n",
    "display(\n",
    "    Audio(data=mel.image_to_audio(output_image), rate=mel.get_sample_rate()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d6c74105",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "huggingface",
   "language": "python",
   "name": "huggingface"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.6"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}