Error while hyperparameter search

When I try wandb.sweep, it gives following error: wandb.errors.CommError: 400 Bad Request: The browser (or proxy) sent a request that this server could not understand.

Following this, is my sweep config
{‘method’: ‘random’,
‘metric’: {‘goal’: ‘minimize’, ‘name’: ‘loss’},
‘parameters’: {‘batch_size’: {‘distribution’: ‘q_log_uniform_values’,
‘max’: 256,
‘min’: 32,
‘q’: 8},
‘epochs’: {‘value’: 10},
‘fc_layer_size’: {‘values’: [16, 32, 64]},
‘learning_rate’: {‘distribution’: ‘uniform’,
‘max’: 0.1,
‘min’: 0},
‘optimizer’: {‘values’: [‘adam’, ‘sgd’]},
‘training_snr’: {‘values’: [0.3981071705534972,
0.44668359215096315,
0.5011872336272722,
0.5623413251903491,
0.6309573444801932,
0.7079457843841379,
0.7943282347242815,
0.8912509381337456,
1.0,
1.1220184543019633,
1.2589254117941673,
1.4125375446227544,
1.5848931924611136,
1.7782794100389228,
1.9952623149688795,
2.2387211385683394,
2.51188643150958,
2.8183829312644537,
3.1622776601683795,
3.548133892335755,
3.9810717055349722,
4.466835921509632,
5.011872336272722,
5.623413251903491,
6.309573444801933,
7.079457843841379,
7.943282347242816,
8.912509381337454,
10.0]}}}

Hi @raikar_sumanth ,
Could you share the debug.log and debug-internal.log files associated with one of the runs which displays this error? It would be very helpful in order to gain more visibility into this error and understand why you are seeing CommErrors here.

Thanks

There was no run, it gets stuck at wandb.sweep(sweep_config). if I disable the training_snr parameter from the sweep config , it works perfectly.
the training_snr parameter is given at the forward pass in pytorch like this
for epoch in range(config.epochs):
avg_loss = train_epoch(network,loader,optimizer,training_snr)
wandb.log({“loss”:avg_loss, “epoch”:epoch,“training_snr”:config.training_snr})

Can you suggest any changes?

Hi @raikar_sumanth ,

Your sweep dictionary configuration is setup correctly and I was able to use it for a an experiment and didn’t run into the same error. To better assist you can you please provide the following:

  • wandb version
  • Full traceback of error
  • Description/Summary of the experiment you are running and which integration ,if any, you are using
  • Example colab of your code to attempt to reproduce your specific error

Thank-you.

Hi @raikar_sumanth ,

Since we have not heard back from you we are going to close this request. If you would like to re-open the conversation, please let us know!

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.