[24.0.6] Train toml config seed type error

[hinablue]

Error when start the training. And I check the tmpfilelora.toml found the seed is float not the int.

Change this line will be good to go.

"seed": int(seed) if seed != 0 else None,  // Force the seed to int(seed)

Full Error trackback,

services-kohya-1  |   torch.utils._pytree._register_pytree_node(
services-kohya-1  | 2024-04-23 03:57:31 INFO     Loading settings from            train_util.py:3744
services-kohya-1  |                              ./outputs/tmpfilelora.toml...
services-kohya-1  |                     INFO     ./outputs/tmpfilelora            train_util.py:3763
services-kohya-1  | Traceback (most recent call last):
services-kohya-1  |   File "_mt19937.pyx", line 180, in numpy.random._mt19937.MT19937._legacy_seeding
services-kohya-1  | TypeError: 'float' object cannot be interpreted as an integer
services-kohya-1  |
services-kohya-1  | During handling of the above exception, another exception occurred:
services-kohya-1  |
services-kohya-1  | Traceback (most recent call last):
services-kohya-1  |   File "/app/sd-scripts/sdxl_train_network.py", line 185, in <module>
services-kohya-1  |     trainer.train(args)
services-kohya-1  |   File "/app/sd-scripts/train_network.py", line 151, in train
services-kohya-1  |     set_seed(args.seed)
services-kohya-1  |   File "/app/venv/lib/python3.10/site-packages/accelerate/utils/random.py", line 44, in set_seed
services-kohya-1  |     np.random.seed(seed)
services-kohya-1  |   File "numpy/random/mtrand.pyx", line 4806, in numpy.random.mtrand.seed
services-kohya-1  |   File "numpy/random/mtrand.pyx", line 250, in numpy.random.mtrand.RandomState.seed
services-kohya-1  |   File "_mt19937.pyx", line 168, in numpy.random._mt19937.MT19937._legacy_seeding
services-kohya-1  |   File "_mt19937.pyx", line 188, in numpy.random._mt19937.MT19937._legacy_seeding
services-kohya-1  | TypeError: Cannot cast scalar from dtype('float64') to dtype('int64') according to the rule 'safe'
services-kohya-1  | Traceback (most recent call last):
services-kohya-1  |   File "/app/venv/bin/accelerate", line 8, in <module>
services-kohya-1  |     sys.exit(main())
services-kohya-1  |   File "/app/venv/lib/python3.10/site-packages/accelerate/commands/accelerate_cli.py", line 47, in main
services-kohya-1  |     args.func(args)
services-kohya-1  |   File "/app/venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 1017, in launch_command
services-kohya-1  |     simple_launcher(args)
services-kohya-1  |   File "/app/venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 637, in simple_launcher
services-kohya-1  |     raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
services-kohya-1  | subprocess.CalledProcessError: Command '['/app/venv/bin/python3', '/app/sd-scripts/sdxl_train_network.py', '--config_file', './outputs/tmpfilelora.toml', '--log_prefix=xl-lora']' returned non-zero exit status 1.
services-kohya-1  | 03:57:33-429097 INFO     Training has ended.

[hinablue]

Also those need to be int,

"caption_dropout_every_n_epochs": int(caption_dropout_every_n_epochs),

"lr_scheduler_num_cycles": (
    int(lr_scheduler_num_cycles) if lr_scheduler_num_cycles != "" else int(epoch)
),

"max_train_epochs": int(max_train_epochs) if max_train_epochs != 0 else None,

"max_train_steps": int(max_train_steps) if max_train_steps != 0 else None,

"persistent_data_loader_workers": int(persistent_data_loader_workers),


config_toml_data["max_data_loader_n_workers"] = int(max_data_loader_n_workers)

Error trackback,

services-kohya-1  | Using decoupled weight decay
services-kohya-1  | enable fp8 training.
services-kohya-1  | running training / 学習開始
services-kohya-1  |   num train images * repeats / 学習画像の数×繰り返し回数: 1650
services-kohya-1  |   num reg images / 正則化画像の数: 1400
services-kohya-1  |   num batches per epoch / 1epochのバッチ数: 839
services-kohya-1  |   num epochs / epoch数: 15
services-kohya-1  |   batch size per device / バッチサイズ: 4
services-kohya-1  |   gradient accumulation steps / 勾配を合計するステップ数 = 1
services-kohya-1  |   total optimization steps / 学習ステップ数: 12000.0
services-kohya-1  | Traceback (most recent call last):
services-kohya-1  |   File "/app/sd-scripts/sdxl_train_network.py", line 185, in <module>
services-kohya-1  |     trainer.train(args)
services-kohya-1  |   File "/app/sd-scripts/train_network.py", line 739, in train
services-kohya-1  |     progress_bar = tqdm(range(args.max_train_steps), smoothing=0, disable=not accelerator.is_local_main_process, desc="steps")
services-kohya-1  | TypeError: 'float' object cannot be interpreted as an integer
services-kohya-1  | Traceback (most recent call last):
services-kohya-1  |   File "/app/venv/bin/accelerate", line 8, in <module>
services-kohya-1  |     sys.exit(main())
services-kohya-1  |   File "/app/venv/lib/python3.10/site-packages/accelerate/commands/accelerate_cli.py", line 47, in main
services-kohya-1  |     args.func(args)
services-kohya-1  |   File "/app/venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 1017, in launch_command
services-kohya-1  |     simple_launcher(args)
services-kohya-1  |   File "/app/venv/lib/python3.10/site-packages/accelerate/commands/launch.py", line 637, in simple_launcher
services-kohya-1  |     raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
services-kohya-1  | subprocess.CalledProcessError: Command '['/app/venv/bin/python3', '/app/sd-scripts/sdxl_train_network.py', '--config_file', './outputs/tmpfilelora.toml', '--log_prefix=xl-locon']' returned non-zero exit status 1.
services-kohya-1  | 04:14:24-424949 INFO     Training has ended.

[bmaltais]

Fixed in latest dev commit.