mindone/examples/diffusers/dreambooth/test_dreambooth_sd3.py at master · alien-0119/mindone · GitHub

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
# 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 logging
import os
import sys
import tempfile

from mindone.diffusers import DiffusionPipeline, SD3Transformer2DModel

sys.path.append("..")
from examples.diffusers.test_examples_utils import ExamplesTests, run_command  # noqa: E402

ExamplesTests._launch_args = ["python"]

logging.basicConfig(level=logging.DEBUG)

logger = logging.getLogger()
stream_handler = logging.StreamHandler(sys.stdout)
logger.addHandler(stream_handler)


class DreamBoothSD3(ExamplesTests):
    instance_data_dir = "docs/diffusers/imgs"
    instance_prompt = "photo"
    pretrained_model_name_or_path = "hf-internal-testing/tiny-sd3-pipe"
    script_path = "train_dreambooth_sd3.py"

    def test_dreambooth(self):
        with tempfile.TemporaryDirectory() as tmpdir:
            test_args = f"""
                examples/diffusers/dreambooth/{self.script_path}
                --pretrained_model_name_or_path {self.pretrained_model_name_or_path}
                --instance_data_dir {self.instance_data_dir}
                --instance_prompt {self.instance_prompt}
                --resolution 64
                --train_batch_size 1
                --gradient_accumulation_steps 1
                --max_train_steps 2
                --learning_rate 5.0e-04
                --scale_lr
                --lr_scheduler constant
                --lr_warmup_steps 0
                --output_dir {tmpdir}
                """.split()

            run_command(self._launch_args + test_args)
            # save_pretrained smoke test
            self.assertTrue(os.path.isfile(os.path.join(tmpdir, "transformer", "diffusion_pytorch_model.safetensors")))
            self.assertTrue(os.path.isfile(os.path.join(tmpdir, "scheduler", "scheduler_config.json")))

    def test_dreambooth_checkpointing(self):
        with tempfile.TemporaryDirectory() as tmpdir:
            # Run training script with checkpointing
            # max_train_steps == 4, checkpointing_steps == 2
            # Should create checkpoints at steps 2, 4

            initial_run_args = f"""
                examples/diffusers/dreambooth/{self.script_path}
                --pretrained_model_name_or_path {self.pretrained_model_name_or_path}
                --instance_data_dir {self.instance_data_dir}
                --instance_prompt {self.instance_prompt}
                --resolution 64
                --train_batch_size 1
                --gradient_accumulation_steps 1
                --max_train_steps 4
                --learning_rate 5.0e-04
                --scale_lr
                --lr_scheduler constant
                --lr_warmup_steps 0
                --output_dir {tmpdir}
                --checkpointing_steps=2
                --seed=0
                """.split()

            run_command(self._launch_args + initial_run_args)

            # check can run the original fully trained output pipeline
            pipe = DiffusionPipeline.from_pretrained(tmpdir)
            pipe(self.instance_prompt, num_inference_steps=1)

            # check checkpoint directories exist
            self.assertTrue(os.path.isdir(os.path.join(tmpdir, "checkpoint-2")))
            self.assertTrue(os.path.isdir(os.path.join(tmpdir, "checkpoint-4")))

            # check can run an intermediate checkpoint
            transformer = SD3Transformer2DModel.from_pretrained(tmpdir, subfolder="checkpoint-2/transformer")
            pipe = DiffusionPipeline.from_pretrained(self.pretrained_model_name_or_path, transformer=transformer)
            pipe(self.instance_prompt, num_inference_steps=1)

    def test_dreambooth_checkpointing_checkpoints_total_limit(self):
        with tempfile.TemporaryDirectory() as tmpdir:
            test_args = f"""
            examples/diffusers/dreambooth/{self.script_path}
            --pretrained_model_name_or_path={self.pretrained_model_name_or_path}
            --instance_data_dir={self.instance_data_dir}
            --output_dir={tmpdir}
            --instance_prompt={self.instance_prompt}
            --resolution=64
            --train_batch_size=1
            --gradient_accumulation_steps=1
            --max_train_steps=6
            --checkpoints_total_limit=2
            --checkpointing_steps=2
            """.split()

            run_command(self._launch_args + test_args)

            self.assertEqual(
                {x for x in os.listdir(tmpdir) if "checkpoint" in x},
                {"checkpoint-4", "checkpoint-6"},
            )