使用DeepChem 2.0.0遇TypeError：需字符串/缓冲区而非NoneType

This error pops up because one of the variables you’re trying to stitch into the file path is None—Python 2.7 doesn’t let you add None to a string, which triggers the "coercing to Unicode" type mismatch. Let’s walk through how to fix this:

Root Cause #

Looking at the error line:

save_dir = os.path.join(data_dir, "bace_c/" + featurizer + "/" + str(split))

The issue is almost certainly that featurizer is None (since split has a default value of 'random', and a missing data_dir would throw a different error). Even though load_bace_classification defines a default featurizer='ECFP', your call to run_benchmark (in test_run_benchmark_models.py line 30) is probably passing None for this parameter.

Step-by-Step Fixes #

1. Correct the run_benchmark Call #

First, check your code in test_run_benchmark_models.py line 30. Make sure you’re passing a valid string for featurizer, not None. For example:

# Example of a valid call
run_benchmark(
    # Your other parameters here...
    featurizer='ECFP',  # Use a valid featurizer like 'ECFP' or 'GraphConv'
    seed=123
)

2. Make the load_bace_classification Function More Robust #

To prevent this issue from recurring, add parameter validation and use safer path concatenation:

def load_bace_classification(featurizer='ECFP', split='random', reload=True):
  """Load bace datasets."""
  # Add checks to catch invalid parameters early
  if not isinstance(featurizer, basestring):
      raise ValueError(f"featurizer must be a string (e.g., 'ECFP'), got {type(featurizer)} instead")
  if not isinstance(split, basestring):
      raise ValueError(f"split must be a string (e.g., 'random'), got {type(split)} instead")
  
  # Featurize bace dataset
  print("About to featurize bace dataset.")
  data_dir = deepchem.utils.get_data_dir()
  if reload:
    # Use os.path.join with separate arguments instead of string concatenation
    save_dir = os.path.join(data_dir, "bace_c", featurizer, str(split))
    dataset_file = os.path.join(data_dir, "bace.csv")
    if not os.path.exists(dataset_file):
      deepchem.utils.download_url('http://deepchem.io.s3-website-us-west-1.amazonaws.com/datasets/bace.csv' )

Using os.path.join with multiple arguments avoids string addition pitfalls, and the validation checks will give you a clearer error message if invalid parameters are passed.

3. Verify get_data_dir() Returns a Valid Path (Optional) #

If the above fixes don’t work, double-check that deepchem.utils.get_data_dir() returns a valid string path. Add a print statement to confirm:

data_dir = deepchem.utils.get_data_dir()
print(f"Data directory: {data_dir}, type: {type(data_dir)}")

If it returns None, you may need to set the DEEPCHEM_DATA_DIR environment variable or check your DeepChem configuration.

内容的提问来源于stack exchange，提问作者Ha Young Kim