Spaces:
Running
Running
File size: 5,418 Bytes
6b448ad |
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 |
# coding=utf-8
# Copyright 2023 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 gc
import unittest
import numpy as np
import torch
from diffusers import AutoencoderKL, DDIMScheduler, DiTPipeline, DPMSolverMultistepScheduler, Transformer2DModel
from diffusers.utils import is_xformers_available, load_numpy, slow, torch_device
from diffusers.utils.testing_utils import require_torch_gpu
from ...pipeline_params import (
CLASS_CONDITIONED_IMAGE_GENERATION_BATCH_PARAMS,
CLASS_CONDITIONED_IMAGE_GENERATION_PARAMS,
)
from ...test_pipelines_common import PipelineTesterMixin
torch.backends.cuda.matmul.allow_tf32 = False
class DiTPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
pipeline_class = DiTPipeline
params = CLASS_CONDITIONED_IMAGE_GENERATION_PARAMS
required_optional_params = PipelineTesterMixin.required_optional_params - {
"latents",
"num_images_per_prompt",
"callback",
"callback_steps",
}
batch_params = CLASS_CONDITIONED_IMAGE_GENERATION_BATCH_PARAMS
test_cpu_offload = False
def get_dummy_components(self):
torch.manual_seed(0)
transformer = Transformer2DModel(
sample_size=16,
num_layers=2,
patch_size=4,
attention_head_dim=8,
num_attention_heads=2,
in_channels=4,
out_channels=8,
attention_bias=True,
activation_fn="gelu-approximate",
num_embeds_ada_norm=1000,
norm_type="ada_norm_zero",
norm_elementwise_affine=False,
)
vae = AutoencoderKL()
scheduler = DDIMScheduler()
components = {"transformer": transformer.eval(), "vae": vae.eval(), "scheduler": scheduler}
return components
def get_dummy_inputs(self, device, seed=0):
if str(device).startswith("mps"):
generator = torch.manual_seed(seed)
else:
generator = torch.Generator(device=device).manual_seed(seed)
inputs = {
"class_labels": [1],
"generator": generator,
"num_inference_steps": 2,
"output_type": "numpy",
}
return inputs
def test_inference(self):
device = "cpu"
components = self.get_dummy_components()
pipe = self.pipeline_class(**components)
pipe.to(device)
pipe.set_progress_bar_config(disable=None)
inputs = self.get_dummy_inputs(device)
image = pipe(**inputs).images
image_slice = image[0, -3:, -3:, -1]
self.assertEqual(image.shape, (1, 16, 16, 3))
expected_slice = np.array([0.4380, 0.4141, 0.5159, 0.0000, 0.4282, 0.6680, 0.5485, 0.2545, 0.6719])
max_diff = np.abs(image_slice.flatten() - expected_slice).max()
self.assertLessEqual(max_diff, 1e-3)
def test_inference_batch_single_identical(self):
self._test_inference_batch_single_identical(relax_max_difference=True, expected_max_diff=1e-3)
@unittest.skipIf(
torch_device != "cuda" or not is_xformers_available(),
reason="XFormers attention is only available with CUDA and `xformers` installed",
)
def test_xformers_attention_forwardGenerator_pass(self):
self._test_xformers_attention_forwardGenerator_pass(expected_max_diff=1e-3)
@require_torch_gpu
@slow
class DiTPipelineIntegrationTests(unittest.TestCase):
def tearDown(self):
super().tearDown()
gc.collect()
torch.cuda.empty_cache()
def test_dit_256(self):
generator = torch.manual_seed(0)
pipe = DiTPipeline.from_pretrained("facebook/DiT-XL-2-256")
pipe.to("cuda")
words = ["vase", "umbrella", "white shark", "white wolf"]
ids = pipe.get_label_ids(words)
images = pipe(ids, generator=generator, num_inference_steps=40, output_type="np").images
for word, image in zip(words, images):
expected_image = load_numpy(
f"https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/dit/{word}.npy"
)
assert np.abs((expected_image - image).max()) < 1e-2
def test_dit_512(self):
pipe = DiTPipeline.from_pretrained("facebook/DiT-XL-2-512")
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
pipe.to("cuda")
words = ["vase", "umbrella"]
ids = pipe.get_label_ids(words)
generator = torch.manual_seed(0)
images = pipe(ids, generator=generator, num_inference_steps=25, output_type="np").images
for word, image in zip(words, images):
expected_image = load_numpy(
"https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main"
f"/dit/{word}_512.npy"
)
assert np.abs((expected_image - image).max()) < 1e-1
|