Dev分支不允许64以下的bucket_reso_steps

[PercyDan54]

在dev/sdxl_train_network.py的39行有验证bucket_reso_steps能否被32整除

lora-scripts/scripts/dev/sdxl_train_network.py

Lines 26 to 40 in 16749ca

	def assert_extra_args(self, args, train_dataset_group):
	super().assert_extra_args(args, train_dataset_group)
	sdxl_train_util.verify_sdxl_training_args(args)
	if args.cache_text_encoder_outputs:
	assert (
	train_dataset_group.is_text_encoder_output_cacheable()
	), "when caching Text Encoder output, either caption_dropout_rate, shuffle_caption, token_warmup_step or caption_tag_dropout_rate cannot be used / Text Encoderの出力をキャッシュするときはcaption_dropout_rate, shuffle_caption, token_warmup_step, caption_tag_dropout_rateは使えません"
	assert (
	args.network_train_unet_only or not args.cache_text_encoder_outputs
	), "network for Text Encoder cannot be trained with caching Text Encoder outputs / Text Encoderの出力をキャッシュしながらText Encoderのネットワークを学習することはできません"
	train_dataset_group.verify_bucket_reso_steps(32)

但是可以看到在27行的super().assert_extra_args(args, train_dataset_group)

这个方法在stable里是空的

lora-scripts/scripts/stable/train_network.py

Lines 97 to 98 in 16749ca

	def assert_extra_args(self, args, train_dataset_group):
	pass

但是在dev里验证了bucket_reso_steps能否被64整除

lora-scripts/scripts/dev/train_network.py

Lines 124 to 125 in 16749ca

	def assert_extra_args(self, args, train_dataset_group):
	train_dataset_group.verify_bucket_reso_steps(64)

导致dev不能使用64以下的bucket_reso_steps。
我不清楚到底预期行为是什么

[PercyDan54]

kohya-ss/sd-scripts#1915