sd-scripts for lumina training in colab doesnt work

Im trying update old notebook colabs for Lumina training.

I will tried updated XL Lora Trainer by Hollowstrawberry

But, when I start training it stopped, on settings loading.
https://colab.research.google.com/drive/1711hIGxFmJ_leFe7lZzCsvN5Yxkc6Ovg?usp=sharing
Warnings under closing:

ipex flag is deprecated, will be removed in Accelerate v1.10. From 2.7.0, PyTorch has all needed optimizations for Intel CPU and XPU.
2025-07-28 17:07:05.835713: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2025-07-28 17:07:05.852711: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:477] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
E0000 00:00:1753722425.873980 2027 cuda_dnn.cc:8310] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1753722425.880420 2027 cuda_blas.cc:1418] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
2025-07-28 17:07:05.901506: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
/content/drive/MyDrive/sd-scripts-sd3/library/lumina_models.py:51: UserWarning: Cannot import apex RMSNorm, switch to vanilla implementation
warnings.warn("Cannot import apex RMSNorm, switch to vanilla implementation")

Also, I find Modified lora-scripts by Akegarasu, and it setting loading stopped too
https://drive.google.com/file/d/1nyVN8avR7Ko65-T73VWkR8W4ijCjsxnh/view?usp=sharing
https://drive.google.com/file/d/1m_r2upJz1s5z1SZjHuXx-wmJxRW2j4vj/view?usp=sharing

There I only getting warning:
WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
E0000 00:00:1753705231.371468 8011 cuda_dnn.cc:8310] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
E0000 00:00:1753705231.378031 8011 cuda_blas.cc:1418] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered

Where it can be problem?

Start by reproducing the failure in the linked Colab notebooks and inspect the settings-loading path around library/lumina_models.py:51. Separate the warnings from the condition that stops training, then verify that settings load successfully and training begins in a current Lumina notebook.

Written by the indexing model from the issue text.