How to use custom learning scheduler?

parser.add_argument(
"--lr_scheduler_args",
type=str,
default=None,
nargs="*",
help='additional arguments for scheduler (like "T_max=100") / スケジューラの追加引数（例： "T_max100"）',
)

parser.add_argument(
"--lr_scheduler",
type=str,
default="constant",
help="scheduler to use for learning rate / 学習率のスケジューラ: linear, cosine, cosine_with_restarts, polynomial, constant (default), constant_with_warmup, adafactor",
)
parser.add_argument(
"--lr_warmup_steps",
type=int_or_float,
default=0,
help="Int number of steps for the warmup in the lr scheduler (default is 0) or float with ratio of train steps"
" / 学習率のスケジューラをウォームアップするステップ数（デフォルト0）、または学習ステップの比率（1未満のfloat値の場合）",
)
parser.add_argument(
"--lr_decay_steps",
type=int_or_float,
default=0,
help="Int number of steps for the decay in the lr scheduler (default is 0) or float (<1) with ratio of train steps"
" / 学習率のスケジューラを減衰させるステップ数（デフォルト0）、または学習ステップの比率（1未満のfloat値の場合）",
)
parser.add_argument(
"--lr_scheduler_num_cycles",
type=int,
default=1,
help="Number of restarts for cosine scheduler with restarts / cosine with restartsスケジューラでのリスタート回数",
)
parser.add_argument(
"--lr_scheduler_power",
type=float,
default=1,
help="Polynomial power for polynomial scheduler / polynomialスケジューラでのpolynomial power",
)
parser.add_argument(
"--fused_backward_pass",
action="store_true",
help="Combines backward pass and optimizer step to reduce VRAM usage. Only available in SDXL, SD3 and FLUX"
" / バックワードパスとオプティマイザステップを組み合わせてVRAMの使用量を削減します。SDXL、SD3、FLUXでのみ利用可能",
)
parser.add_argument(
"--lr_scheduler_timescale",
type=int,
default=None,
help="Inverse sqrt timescale for inverse sqrt scheduler,defaults to `num_warmup_steps`"
+ " / 逆平方根スケジューラのタイムスケール、デフォルトは`num_warmup_steps`",
)
parser.add_argument(
"--lr_scheduler_min_lr_ratio",
type=float,
default=None,
help="The minimum learning rate as a ratio of the initial learning rate for cosine with min lr scheduler and warmup decay scheduler"
+ " / 初期学習率の比率としての最小学習率を指定する、cosine with min lr と warmup decay スケジューラ で有効",
)

I want to assign a different learning rate to specific, predefined steps within the total number of training steps.

Start with the argument definitions for --lr_scheduler, --lr_scheduler_args, --lr_warmup_steps, and --lr_decay_steps shown in the issue. Trace how these options are used during training and determine the scope of supporting different learning rates at predefined steps. Done should be a clear supported behavior or documented limitation with an appropriate verification path.

Written by the indexing model from the issue text.