`random.samples` not working in subconfigs.

[n-gao]

Expected Behavior

The example_config.yaml should generate 192 configs.
for small_datasets: 5(learning_rate) * 3(dropout) * 2(dataset) * 2(hidden_size) * 3(max_epoch)=180
for large_dataset: 3(dropout) * 2(dataset) * 2(hidden_size)=12

Actual Behavior

Only 72 configs are generated:
for small_datasets: 5(learning_rate) * 3(dropout) * 2(dataset) * 2(hidden_size) * 1(max_epoch)=60
for large_dataset: 3(dropout) * 2(dataset) * 2(hidden_size)=12

Steps to Reproduce the Problem

seml seml_test add example_config.yaml

Specifications

Details

• Version: master branch
• Python version: 3.9
• Platform: linux