VisionPartitioner such as CIFAR10Partitioner and MNISTPartitioner gets stuck when partitioning non-iid data with dir_alpha si small(for eg 0.1 or 0.3)

[williamium3000]

Describe the bug
CIFAR10Partitioner gets stuck when partitioning non-iid data with dir_alpha si small(for eg 0.1 or 0.3)

Environment
Environment you use when bug appears:

1. Python version:3.8.5

2. PyTorch Version:1.9.0+cu111

3. FedLab version:1.1.4

4. code you run
   CIFAR10Partitioner(targets=targets, num_clients=100,
   balance=True, partition="dirichlet",
   unbalance_sgm=0,
   num_shards=None,
   dir_alpha=0.1)

5. the detailed error
   It gets stuck when there is still remaining data. For example, a typical error log looks like:
   ....
   Remaining Data: 1340
   Remaining Data: 1339
   Remaining Data: 1338
   Remaining Data: 1337
   Remaining Data: 1336
   Remaining Data: 1336
   Remaining Data: 1335
   Remaining Data: 1334
   Remaining Data: 1333
   Remaining Data: 1332
   Remaining Data: 1331
   Remaining Data: 1330
   Remaining Data: 1329
   Remaining Data: 1329
   Remaining Data: 1328
   Remaining Data: 1328
   Remaining Data: 1327
   Remaining Data: 1327
   Remaining Data: 1326
   Remaining Data: 1325
   Remaining Data: 1324
   Remaining Data: 1324
   Remaining Data: 1324
   Remaining Data: 1323
   Remaining Data: 1323
   Remaining Data: 1323
   Remaining Data: 1322
   Remaining Data: 1322
   Remaining Data: 1321
   Remaining Data: 1320
   Remaining Data: 1320
   Remaining Data: 1319
   Remaining Data: 1318
   Remaining Data: 1318
   Remaining Data: 1318
   Remaining Data: 1317
   Remaining Data: 1316
   Remaining Data: 1316
   Remaining Data: 1315
   Remaining Data: 1314
   Remaining Data: 1314
   Remaining Data: 1314
   Remaining Data: 1314
   Remaining Data: 1314
   Remaining Data: 1313
   Remaining Data: 1313

And it gets stuck here.

[GiuseppeGalilei]

I'm encountering the same issue. Unfortunately the answer has been already given in issue #230 .
If it can help you I was able to generate the partition fairly quickly with balance=True, alpha=0.3 and seed was 119 or 120.
I would be grateful in case you find some seed that works for other values of alpha, if you post them, thanks :)

[Zengdun-cs]

Thanks for your attention.
As @GiuseppeGalilei has mentioned, the answer has been given in issue #230. You may encounter this issue with certain seeds. The solution is to change your random seed.

[williamium3000]

Thanks for the commet and reply. I have succeeded in partioning with balance=True, alpha=0.1 and 0.3 using seed of 260.