passing a numpy array to predict_fn when making inference for xgboost model

Question

I have a model that's trained locally and deployed to SageMaker to make inferences / invoke endpoint. When I try to make predictions, I get the following exception. 
 
    raise ValueError('Input numpy.ndarray must be 2 dimensional')
    ValueError: Input numpy.ndarray must be 2 dimensional

My `model` is a xgboost model with some pre-processing (variable encoding) and hyper-parameter tuning. Here's what `model` object looks like:

XGBRegressor(colsample_bytree=xxx, gamma=xxx,
                 learning_rate=xxx, max_depth=x, n_estimators=xxx,
                 subsample=xxx)

My test data is a string of float values which is turned into an array as the data must be passed as numpy array. 
 
    testdata = [........., 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 2000, 200, 85, 412412, 123, 41, 552, 50000, 512, 0.1, 10.0, 2.0, 0.05]

I have tried to reshape the numpy array from 1d to 2d, however, that doesn't work as the number of features between test data and trained model do not match.

My question is how do I pass a numpy array same as the length of # of features in trained model? I am able to make predictions by passing test data as a list locally.

More info on inference script here: https://github.com/aws-samples/amazon-sagemaker-local-mode/blob/main/xgboost_script_mode_local_training_and_serving/code/inference.py

Traceback (most recent call last):
    File "/miniconda3/lib/python3.6/site-packages/sagemaker_containers/_functions.py", line 93, in wrapper
    return fn(*args, **kwargs)
    File "/opt/ml/code/inference.py", line 75, in predict_fn
    prediction = model.predict(input_data)
    File "/miniconda3/lib/python3.6/site-packages/xgboost/sklearn.py", line 448, in predict
    test_dmatrix = DMatrix(data, missing=self.missing, nthread=self.n_jobs)
    File "/miniconda3/lib/python3.6/site-packages/xgboost/core.py", line 404, in __init__
    self._init_from_npy2d(data, missing, nthread)
    File "/miniconda3/lib/python3.6/site-packages/xgboost/core.py", line 474, in _init_from_npy2d
    raise ValueError('Input numpy.ndarray must be 2 dimensional')
    ValueError: Input numpy.ndarray must be 2 dimensional

Answer

Try converting your list to a numpy 2d array like so:

`a = np.array([1, 2, 3])`

and replace `[1, 2, 3]` with your list.

Answer

XGBoost, similar to scikit-learn, expects X as 2D data (n_samples, n_features). In order to predict one sample, you need to reshape your list or feature vector to a 2D array.
   
    import numpy as np

lst = [1, 2, 3]
    lst_reshaped = np.array(lst).reshape((1,-1))
    clf.predict(lst_reshaped)

passing a numpy array to predict_fn when making inference for xgboost model

Relevant content