gpuIndices

[creater-zq]

Describe the issue:

Environment:

• NNI version:2.2
• Training service (local|remote|pai|aml|etc):local
• Client OS:Windows 10
• Python version:3.8.5
• PyTorch/TensorFlow version:1.7.1
• Is conda/virtualenv/venv used?:conda version 10.2
• Is running in Docker?:False

Configuration:

• Experiment config (remember to remove secrets!):config.yml
  authorName: default
  experimentName: test_nni
  trialConcurrency: 1
  maxExecDuration: 1d
  maxTrialNum: 30
  trainingServicePlatform: local
  searchSpacePath: search_space.json
  useAnnotation: false
  tuner:
  builtinTunerName: TPE
  classArgs:
  optimize_mode: minimize
  trial:
  command: python train_nni.py
  codeDir: .
  gpuNum: 1
  localConfig:
  useActiveGpu: true
  gpuIndices: 3

• Search space:
  { "start_lr": {"_type": "choice", "_value": [0.0001, 0.001, 0.01, 0.1]},
  "batch_size": {"_type": "choice", "_value": [8, 12, 16, 32, 64, 128]}
  }

error:
ERROR: Config V1 validation failed: SchemaError('gpuIndex format error!')

[scarlett2018]

error:
ERROR: Config V1 validation failed: SchemaError('gpuIndex format error!')

#3540
Adding error msg pr owner in to suggest.

[creater-zq]

error:
ERROR: Config V1 validation failed: SchemaError('gpuIndex format error!')

#3540
Adding error msg pr owner in to suggest.

Thank you, but I don't quite understand your suggestion. Could you tell me more about it

[liuzhe-lz]

This bug will be fixed in next release.
For now you can walkaround by replacing 3 (int) with '3' (str).

[creater-zq]

This bug will be fixed in next release.
For now you can walkaround by replacing 3 (int) with '3' (str).

nice， thanks!

[liuzhe-lz]

Expected to be fixed in v2.3 release next week.

[kvartet]

hello @creater-zq, I'm closing this issue as it had been fixed, free feel to reopen it if you still think it an active issue.

[Delong-Zhu]

This bug will be fixed in next release. For now you can walkaround by replacing 3 (int) with '3' (str).

thanks!