Validation does not work with Anima LoRa

[Hirmuolio]

Trying to have validation loss while training Anima LoRa does not work.

Enabling validation loss with validation_split = 0.5 and validate_every_n_steps = 4 results in the following error:

traceback (most recent call last):
  File "/ai/sd-scripts/anima_train_network.py", line 448, in <module>
    trainer.train(args)
  File "/ai/sd-scripts/train_network.py", line 1558, in train
    loss = self.process_batch(
           ^^^^^^^^^^^^^^^^^^^
  File "/ai/sd-scripts/anima_train_network.py", line 353, in process_batch
    text_encoder_outputs_list = anima_text_encoding_strategy.drop_cached_text_encoder_outputs(
                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: AnimaTextEncodingStrategy.drop_cached_text_encoder_outputs() missing 1 required positional argument: 't5_attn_mask'
steps:   0%|                                                                                     | 4/5000 [00:12<4:18:41,  3.11s/it, avr_loss=0.0595]
validation steps:   0%|▏                                                           | 1/300 [00:01<08:13,  1.65s/it, val_avg_loss=0.102, timestep=200]
Traceback (most recent call last):
  File "/ai/kohya_ss/.venv/bin/accelerate", line 10, in <module>
    sys.exit(main())
             ^^^^^^
  File "/ai/kohya_ss/.venv/lib/python3.11/site-packages/accelerate/commands/accelerate_cli.py", line 50, in main
    args.func(args)
  File "/ai/kohya_ss/.venv/lib/python3.11/site-packages/accelerate/commands/launch.py", line 1199, in launch_command
    simple_launcher(args)
  File "/ai/kohya_ss/.venv/lib/python3.11/site-packages/accelerate/commands/launch.py", line 785, in simple_launcher
    raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
subprocess.CalledProcessError: Command '['/ai/kohya_ss/.venv/bin/python', 'anima_train_network.py', '--config_file=/ai/training/anima_test.toml']' returned non-zero exit status 1.

[kohya-ss]

Thank you for this! I've fixed this in #2291. It will be merged to main sooner.