hparams need to allow None values

[elistevens]

🐛 Bug

I can't set hparams.gpus to None:

I0318 11:12:45.466972 15554 lightning_utils.py:182] <class '__main__.ResnetLightningExample'> hparams: Namespace(amp_level='O2', backend='', batch_size=16, debug_print_env=False, debug_skip_loaded_hparams_check=False, do_test=False, early_stop_metric='val_loss', early_stop_mode='min', early_stop_patience=10, enable_batch_size_scaling=True, enable_early_stop=False, gpus=None, learning_rate=0.01, max_epochs=1, min_epochs=1, model_load_checkpoint_path='', model_save_path='', nodes=1, use_amp=False)
Traceback (most recent call last):
  File "/home/estevens/.cache/bazel/_bazel_estevens/7fa7f74bbe03dba6c4e36403df17d704/execroot/zoox/bazel-out/k8-py3-fastbuildcuda/bin/experimental/estevens/pytorch/lightning_resnet50.runfiles/zoox/experimental/estevens/pytorch/lightning_resnet50.py", line 128, in <module>
    ResnetLightningExample.init_from_cli(sys.argv[1:]).main()
  File "/home/estevens/.cache/bazel/_bazel_estevens/7fa7f74bbe03dba6c4e36403df17d704/execroot/zoox/bazel-out/k8-py3-fastbuildcuda/bin/experimental/estevens/pytorch/lightning_resnet50.runfiles/zoox/tflight/lightning_utils/lightning_utils.py", line 255, in main
    trainer.fit(self)
  File "/home/estevens/.cache/bazel/_bazel_estevens/7fa7f74bbe03dba6c4e36403df17d704/execroot/zoox/bazel-out/k8-py3-fastbuildcuda/bin/experimental/estevens/pytorch/lightning_resnet50.runfiles/pypi__pytorch_lightning_python3_deps/pytorch_lightning/trainer/trainer.py", line 630, in fit
    self.run_pretrain_routine(model)
  File "/home/estevens/.cache/bazel/_bazel_estevens/7fa7f74bbe03dba6c4e36403df17d704/execroot/zoox/bazel-out/k8-py3-fastbuildcuda/bin/experimental/estevens/pytorch/lightning_resnet50.runfiles/pypi__pytorch_lightning_python3_deps/pytorch_lightning/trainer/trainer.py", line 748, in run_pretrain_routine
    self.logger.log_hyperparams(ref_model.hparams)
  File "/home/estevens/.cache/bazel/_bazel_estevens/7fa7f74bbe03dba6c4e36403df17d704/execroot/zoox/bazel-out/k8-py3-fastbuildcuda/bin/experimental/estevens/pytorch/lightning_resnet50.runfiles/pypi__pytorch_lightning_python3_deps/pytorch_lightning/loggers/base.py", line 18, in wrapped_fn
    fn(self, *args, **kwargs)
  File "/home/estevens/.cache/bazel/_bazel_estevens/7fa7f74bbe03dba6c4e36403df17d704/execroot/zoox/bazel-out/k8-py3-fastbuildcuda/bin/experimental/estevens/pytorch/lightning_resnet50.runfiles/pypi__pytorch_lightning_python3_deps/pytorch_lightning/loggers/tensorboard.py", line 113, in log_hyperparams
    exp, ssi, sei = hparams(params, {})
  File "/home/estevens/.cache/bazel/_bazel_estevens/7fa7f74bbe03dba6c4e36403df17d704/execroot/zoox/bazel-out/k8-py3-fastbuildcuda/bin/experimental/estevens/pytorch/lightning_resnet50.runfiles/pypi__torch_python3_deps/torch/utils/tensorboard/summary.py", line 156, in hparams
    raise ValueError('value should be one of int, float, str, bool, or torch.Tensor')
ValueError: value should be one of int, float, str, bool, or torch.Tensor

To Reproduce

To reproduce, have parser.add_argument('--gpus', default=None, type=str) and then don't give the --gpus CLI argument.

Expected behavior

Screen out None values before sending them on.

• PyTorch Version (e.g., 1.0): 1.4.0
• OS (e.g., Linux): Ubuntun 14.04
• How you installed PyTorch (conda, pip, source): pip
• Build command you used (if compiling from source):
• Python version: 3.6.8
• CUDA/cuDNN version: 10.0
• GPU models and configuration: 2080 Ti
• Any other relevant information:

[Borda]

this is set by the tensorboard, not lightning...
Why would you need to set gpus to NOne when NOne is the default value?
We did the conversion to primitives in #1130

[awaelchli]

I tested this and it is fixed on master. Hparams with unsupported types (such as None in your case) are converted to string and now appear in the hparams tab in Tensorboard.