File size: 6,717 Bytes
dde5d93
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# 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 logging
import os
import sys
import tempfile

import safetensors


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


logging.basicConfig(level=logging.DEBUG)

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


class DreamBoothLoRAFlux(ExamplesTestsAccelerate):
    instance_data_dir = "docs/source/en/imgs"
    instance_prompt = "photo"
    pretrained_model_name_or_path = "hf-internal-testing/tiny-flux-pipe"
    script_path = "examples/dreambooth/train_dreambooth_lora_flux.py"

    def test_dreambooth_lora_flux(self):
        with tempfile.TemporaryDirectory() as tmpdir:
            test_args = f"""
                {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, "pytorch_lora_weights.safetensors")))

            # make sure the state_dict has the correct naming in the parameters.
            lora_state_dict = safetensors.torch.load_file(os.path.join(tmpdir, "pytorch_lora_weights.safetensors"))
            is_lora = all("lora" in k for k in lora_state_dict.keys())
            self.assertTrue(is_lora)

            # when not training the text encoder, all the parameters in the state dict should start
            # with `"transformer"` in their names.
            starts_with_transformer = all(key.startswith("transformer") for key in lora_state_dict.keys())
            self.assertTrue(starts_with_transformer)

    def test_dreambooth_lora_text_encoder_flux(self):
        with tempfile.TemporaryDirectory() as tmpdir:
            test_args = f"""
                {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
                --train_text_encoder
                --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, "pytorch_lora_weights.safetensors")))

            # make sure the state_dict has the correct naming in the parameters.
            lora_state_dict = safetensors.torch.load_file(os.path.join(tmpdir, "pytorch_lora_weights.safetensors"))
            is_lora = all("lora" in k for k in lora_state_dict.keys())
            self.assertTrue(is_lora)

            starts_with_expected_prefix = all(
                (key.startswith("transformer") or key.startswith("text_encoder")) for key in lora_state_dict.keys()
            )
            self.assertTrue(starts_with_expected_prefix)

    def test_dreambooth_lora_flux_checkpointing_checkpoints_total_limit(self):
        with tempfile.TemporaryDirectory() as tmpdir:
            test_args = f"""
            {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"},
            )

    def test_dreambooth_lora_flux_checkpointing_checkpoints_total_limit_removes_multiple_checkpoints(self):
        with tempfile.TemporaryDirectory() as tmpdir:
            test_args = f"""
            {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=4
            --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-2", "checkpoint-4"})

            resume_run_args = f"""
            {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=8
            --checkpointing_steps=2
            --resume_from_checkpoint=checkpoint-4
            --checkpoints_total_limit=2
            """.split()

            run_command(self._launch_args + resume_run_args)

            self.assertEqual({x for x in os.listdir(tmpdir) if "checkpoint" in x}, {"checkpoint-6", "checkpoint-8"})