examples/image/diffusion 怎么设置data的file_path?

[bobooscar]

1、访问：https://github.com/hpcaitech/ColossalAI/tree/main/examples/images/diffusion，并安装依赖包
2、下载 laion 1B dataset：

1.   huggingface-cli download --resume-download --repo-type dataset laion/laion1B-nolang-safety
   

2.   cp -r /root/.cache/huggingface/hub/datasets--laion--laion1B-nolang-safety/ ./data/
   

3、修改configs/train_colossalai.yaml
file_path: "data/datasets--laion--laion1B-nolang-safety/"
4、运行train

1. python main.py --logdir /tmp/ --train --base configs/train_colossalai.yaml --ckpt 512-base-ema.ckpt
   

提示如下：

Using base config ['configs/train_colossalai.yaml']
Global seed set to 23
Using ckpt_path = 512-base-ema.ckpt
LatentDiffusion: Running in v-prediction mode
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is None and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is 1024 and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is None and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is 1024 and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is None and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is 1024 and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is None and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is 1024 and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is None and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 1280, context_dim is 1024 and using 20 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is None and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is 1024 and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is None and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is 1024 and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is None and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 640, context_dim is 1024 and using 10 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is None and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is 1024 and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is None and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is 1024 and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is None and using 5 heads.
Setting up MemoryEfficientCrossAttention. Query dim is 320, context_dim is 1024 and using 5 heads.
building MemoryEfficientAttnBlock with 512 in_channels...
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
building MemoryEfficientAttnBlock with 512 in_channels...
/root/anaconda3/envs/ldm/lib/python3.9/site-packages/lightning/pytorch/trainer/connectors/accelerator_connector.py:578: LightningDeprecationWarning: The Trainer argument auto_select_gpus has been deprecated in v1.9.0 and will be removed in v2.0.0. Please use the function lightning.pytorch.accelerators.find_usable_cuda_devices instead.
rank_zero_deprecation(
GPU available: True (cuda), used: True
TPU available: False, using: 0 TPU cores
IPU available: False, using: 0 IPUs
HPU available: False, using: 0 HPUs
Txt2ImgIterableBaseDataset dataset contains -1 examples.
Txt2ImgIterableBaseDataset dataset contains -1 examples.

╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮
│ /mnt/ColossalAI/examples/images/diffusion/main.py:813 in │
│ │
│ 810 │ │ │
│ 811 │ │ # Print some information about the datasets in the data module │
│ 812 │ │ for k in data.datasets: │
│ ❱ 813 │ │ │ rank_zero_info(f"{k}, {data.datasets[k].class.name}, {len(data.datas │
│ 814 │ │ │
│ 815 │ │ # Configure learning rate based on the batch size, base learning rate and number │
│ 816 │ │ # If scale_lr is true, calculate the learning rate based on additional factors │
╰──────────────────────────────────────────────────────────────────────────────────────────────────╯
ValueError: len() should return >= 0

5、尝试修改第3步当中的file_path，加上refs/blobs，提示not a directory，加上snapshots，提示和第4步结果相同。

请问是什么原因？应该怎么配置？