File size: 5,551 Bytes
43b7e92
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# coding=utf-8
# Copyright 2024 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 random
import unittest

import torch

from diffusers import IFInpaintingSuperResolutionPipeline
from diffusers.models.attention_processor import AttnAddedKVProcessor
from diffusers.utils.import_utils import is_xformers_available
from diffusers.utils.testing_utils import floats_tensor, load_numpy, require_torch_gpu, skip_mps, slow, torch_device

from ..pipeline_params import (
    TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS,
    TEXT_GUIDED_IMAGE_INPAINTING_PARAMS,
)
from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference
from . import IFPipelineTesterMixin


@skip_mps
class IFInpaintingSuperResolutionPipelineFastTests(PipelineTesterMixin, IFPipelineTesterMixin, unittest.TestCase):
    pipeline_class = IFInpaintingSuperResolutionPipeline
    params = TEXT_GUIDED_IMAGE_INPAINTING_PARAMS - {"width", "height"}
    batch_params = TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS.union({"original_image"})
    required_optional_params = PipelineTesterMixin.required_optional_params - {"latents"}

    def get_dummy_components(self):
        return self._get_superresolution_dummy_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)

        image = floats_tensor((1, 3, 16, 16), rng=random.Random(seed)).to(device)
        original_image = floats_tensor((1, 3, 32, 32), rng=random.Random(seed)).to(device)
        mask_image = floats_tensor((1, 3, 32, 32), rng=random.Random(seed)).to(device)

        inputs = {
            "prompt": "A painting of a squirrel eating a burger",
            "image": image,
            "original_image": original_image,
            "mask_image": mask_image,
            "generator": generator,
            "num_inference_steps": 2,
            "output_type": "np",
        }

        return inputs

    @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)

    def test_save_load_optional_components(self):
        self._test_save_load_optional_components()

    @unittest.skipIf(torch_device != "cuda", reason="float16 requires CUDA")
    def test_save_load_float16(self):
        # Due to non-determinism in save load of the hf-internal-testing/tiny-random-t5 text encoder
        super().test_save_load_float16(expected_max_diff=1e-1)

    def test_attention_slicing_forward_pass(self):
        self._test_attention_slicing_forward_pass(expected_max_diff=1e-2)

    def test_save_load_local(self):
        self._test_save_load_local()

    def test_inference_batch_single_identical(self):
        self._test_inference_batch_single_identical(
            expected_max_diff=1e-2,
        )


@slow
@require_torch_gpu
class IFInpaintingSuperResolutionPipelineSlowTests(unittest.TestCase):
    def setUp(self):
        # clean up the VRAM before each test
        super().setUp()
        gc.collect()
        torch.cuda.empty_cache()

    def tearDown(self):
        # clean up the VRAM after each test
        super().tearDown()
        gc.collect()
        torch.cuda.empty_cache()

    def test_if_inpainting_superresolution(self):
        pipe = IFInpaintingSuperResolutionPipeline.from_pretrained(
            "DeepFloyd/IF-II-L-v1.0", variant="fp16", torch_dtype=torch.float16
        )
        pipe.unet.set_attn_processor(AttnAddedKVProcessor())
        pipe.enable_model_cpu_offload()

        # Super resolution test
        torch.cuda.empty_cache()
        torch.cuda.reset_max_memory_allocated()
        torch.cuda.reset_peak_memory_stats()

        generator = torch.Generator(device="cpu").manual_seed(0)

        image = floats_tensor((1, 3, 64, 64), rng=random.Random(0)).to(torch_device)
        original_image = floats_tensor((1, 3, 256, 256), rng=random.Random(0)).to(torch_device)
        mask_image = floats_tensor((1, 3, 256, 256), rng=random.Random(1)).to(torch_device)

        output = pipe(
            prompt="anime turtle",
            image=image,
            original_image=original_image,
            mask_image=mask_image,
            generator=generator,
            num_inference_steps=2,
            output_type="np",
        )

        image = output.images[0]

        assert image.shape == (256, 256, 3)

        mem_bytes = torch.cuda.max_memory_allocated()
        assert mem_bytes < 12 * 10**9

        expected_image = load_numpy(
            "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/if/test_if_inpainting_superresolution_stage_II.npy"
        )
        assert_mean_pixel_difference(image, expected_image)

        pipe.remove_all_hooks()