The order of parameters in sweep config will influence the result - Githubissues

wandb / wandb

The AI developer platform. Use Weights & Biases to train and fine-tune models, and manage models from experimentation to production.

https://wandb.ai

MIT License

8.95k stars 658 forks source link

The order of parameters in sweep config will influence the result #2451

Closed possible1402 closed 3 years ago

possible1402 commented 3 years ago

Description I used the sweep example of pytorch version https://colab.research.google.com/github/wandb/examples/blob/master/colabs/pytorch/Organizing_Hyperparameter_Sweeps_in_PyTorch_with_W%26B.ipynb# and found that if I change the order of parameters, the corresponding will be different. For example, in this example, if the 'dropout' in parameters_dict is changing from [0.3, 0.4, 0.5] to [0.4,0.3,0.5], the results don't match.

Wandb features wandb.sweep()

How to reproduce https://colab.research.google.com/github/wandb/examples/blob/master/colabs/pytorch/Organizing_Hyperparameter_Sweeps_in_PyTorch_with_W%26B.ipynb#, I used grid search.

Environment

Environment: Google Colab

vanpelt commented 3 years ago

Closing as a duplicate of https://github.com/wandb/examples/issues/82