Spaces:
Running
on
Zero
Running
on
Zero
File size: 6,353 Bytes
22a452a |
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 |
# coding=utf-8
# Copyright 2025 HuggingFace Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import unittest
import PIL.Image
import torch
from diffusers.utils import load_image
from diffusers.utils.constants import (
DECODE_ENDPOINT_FLUX,
DECODE_ENDPOINT_SD_V1,
DECODE_ENDPOINT_SD_XL,
ENCODE_ENDPOINT_FLUX,
ENCODE_ENDPOINT_SD_V1,
ENCODE_ENDPOINT_SD_XL,
)
from diffusers.utils.remote_utils import (
remote_decode,
remote_encode,
)
from diffusers.utils.testing_utils import (
enable_full_determinism,
slow,
)
enable_full_determinism()
IMAGE = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/astronaut.jpg?download=true"
class RemoteAutoencoderKLEncodeMixin:
channels: int = None
endpoint: str = None
decode_endpoint: str = None
dtype: torch.dtype = None
scaling_factor: float = None
shift_factor: float = None
image: PIL.Image.Image = None
def get_dummy_inputs(self):
if self.image is None:
self.image = load_image(IMAGE)
inputs = {
"endpoint": self.endpoint,
"image": self.image,
"scaling_factor": self.scaling_factor,
"shift_factor": self.shift_factor,
}
return inputs
def test_image_input(self):
inputs = self.get_dummy_inputs()
height, width = inputs["image"].height, inputs["image"].width
output = remote_encode(**inputs)
self.assertEqual(list(output.shape), [1, self.channels, height // 8, width // 8])
decoded = remote_decode(
tensor=output,
endpoint=self.decode_endpoint,
scaling_factor=self.scaling_factor,
shift_factor=self.shift_factor,
image_format="png",
)
self.assertEqual(decoded.height, height)
self.assertEqual(decoded.width, width)
# image_slice = torch.from_numpy(np.array(inputs["image"])[0, -3:, -3:].flatten())
# decoded_slice = torch.from_numpy(np.array(decoded)[0, -3:, -3:].flatten())
# TODO: how to test this? encode->decode is lossy. expected slice of encoded latent?
class RemoteAutoencoderKLSDv1Tests(
RemoteAutoencoderKLEncodeMixin,
unittest.TestCase,
):
channels = 4
endpoint = ENCODE_ENDPOINT_SD_V1
decode_endpoint = DECODE_ENDPOINT_SD_V1
dtype = torch.float16
scaling_factor = 0.18215
shift_factor = None
class RemoteAutoencoderKLSDXLTests(
RemoteAutoencoderKLEncodeMixin,
unittest.TestCase,
):
channels = 4
endpoint = ENCODE_ENDPOINT_SD_XL
decode_endpoint = DECODE_ENDPOINT_SD_XL
dtype = torch.float16
scaling_factor = 0.13025
shift_factor = None
class RemoteAutoencoderKLFluxTests(
RemoteAutoencoderKLEncodeMixin,
unittest.TestCase,
):
channels = 16
endpoint = ENCODE_ENDPOINT_FLUX
decode_endpoint = DECODE_ENDPOINT_FLUX
dtype = torch.bfloat16
scaling_factor = 0.3611
shift_factor = 0.1159
class RemoteAutoencoderKLEncodeSlowTestMixin:
channels: int = 4
endpoint: str = None
decode_endpoint: str = None
dtype: torch.dtype = None
scaling_factor: float = None
shift_factor: float = None
image: PIL.Image.Image = None
def get_dummy_inputs(self):
if self.image is None:
self.image = load_image(IMAGE)
inputs = {
"endpoint": self.endpoint,
"image": self.image,
"scaling_factor": self.scaling_factor,
"shift_factor": self.shift_factor,
}
return inputs
def test_multi_res(self):
inputs = self.get_dummy_inputs()
for height in {
320,
512,
640,
704,
896,
1024,
1208,
1384,
1536,
1608,
1864,
2048,
}:
for width in {
320,
512,
640,
704,
896,
1024,
1208,
1384,
1536,
1608,
1864,
2048,
}:
inputs["image"] = inputs["image"].resize(
(
width,
height,
)
)
output = remote_encode(**inputs)
self.assertEqual(list(output.shape), [1, self.channels, height // 8, width // 8])
decoded = remote_decode(
tensor=output,
endpoint=self.decode_endpoint,
scaling_factor=self.scaling_factor,
shift_factor=self.shift_factor,
image_format="png",
)
self.assertEqual(decoded.height, height)
self.assertEqual(decoded.width, width)
decoded.save(f"test_multi_res_{height}_{width}.png")
@slow
class RemoteAutoencoderKLSDv1SlowTests(
RemoteAutoencoderKLEncodeSlowTestMixin,
unittest.TestCase,
):
endpoint = ENCODE_ENDPOINT_SD_V1
decode_endpoint = DECODE_ENDPOINT_SD_V1
dtype = torch.float16
scaling_factor = 0.18215
shift_factor = None
@slow
class RemoteAutoencoderKLSDXLSlowTests(
RemoteAutoencoderKLEncodeSlowTestMixin,
unittest.TestCase,
):
endpoint = ENCODE_ENDPOINT_SD_XL
decode_endpoint = DECODE_ENDPOINT_SD_XL
dtype = torch.float16
scaling_factor = 0.13025
shift_factor = None
@slow
class RemoteAutoencoderKLFluxSlowTests(
RemoteAutoencoderKLEncodeSlowTestMixin,
unittest.TestCase,
):
channels = 16
endpoint = ENCODE_ENDPOINT_FLUX
decode_endpoint = DECODE_ENDPOINT_FLUX
dtype = torch.bfloat16
scaling_factor = 0.3611
shift_factor = 0.1159
|