AttributeError: 'InternalData' object has no attribute 'aspect_ratio'

[zmy-08]

I'm very sorry to bother you, I'm trying to train PixArt_lcm and I directly modified image_size=512 and the root location of my dataset in both InternalData.py and PixArt_xl2_img1024_lcm.py files. Because I see in the readme.md that it says to just modify the files in the diffusion/dataset/ directory & the configs/ directory. But I got this error and I don't know how to change it。

2024-04-30 09:36:34,827 - PixArt - INFO - World_size: 2, seed: 43
2024-04-30 09:36:34,828 - PixArt - INFO - Initializing: DDP for training
2024-04-30 09:36:44,444 - PixArt - WARNING - lewei scale: (2.0,), base size: 32
2024-04-30 09:36:53,677 - PixArt - INFO - PixArtMS Model Parameters: 611,349,152
Traceback (most recent call last):
  File "/abc/l2/zmy/PixArt-alpha/train_scripts/train_pixart_lcm.py", line 418, in <module>
    batch_size=config.train_batch_size, aspect_ratios=dataset.aspect_ratio, drop_last=True,
  File "/abc/l2/zmy/PixArt-alpha/diffusion/data/datasets/InternalData.py", line 144, in __getattr__
    raise AttributeError(f"'{type(self).__name__}' object has no attribute '{name}'")
AttributeError: 'InternalData' object has no attribute 'aspect_ratio'
2024-04-30 09:36:56,792 - PixArt - INFO - Constructing dataset InternalData...
2024-04-30 09:36:56,793 - PixArt - INFO - T5 max token length: 120
2024-04-30 09:36:56,793 - PixArt - INFO - Dataset InternalData constructed. time: 0.00 s, length (use/ori): 5/5
Traceback (most recent call last):
  File "/abc/l2/zmy/PixArt-alpha/train_scripts/train_pixart_lcm.py", line 418, in <module>
    batch_size=config.train_batch_size, aspect_ratios=dataset.aspect_ratio, drop_last=True,
  File "/abc/l2/zmy/PixArt-alpha/diffusion/data/datasets/InternalData.py", line 144, in __getattr__
    raise AttributeError(f"'{type(self).__name__}' object has no attribute '{name}'")
AttributeError: 'InternalData' object has no attribute 'aspect_ratio'
[2024-04-30 09:36:58,498] torch.distributed.elastic.multiprocessing.api: [WARNING] Sending process 13386 closing signal SIGTERM
[2024-04-30 09:36:58,502] torch.distributed.elastic.multiprocessing.api: [ERROR] failed (exitcode: 1) local_rank: 1 (pid: 13387) of binary: /icislab/volume4/zmy/Myhome/2023028046zmy/envs/pixart/bin/python

[lawrence-cj]

Use InternalData_ms.py instead.