Windows上的TensorFlow tensorflow_transform.beam问题

Question

在本地Win10计算机上运行以下代码并面临错误。

进口下列包装：

import tensorflow as tf
import tensorflow_transform as tft
import tensorflow_transform.beam as tft_beam

from tensorflow_transform.tf_metadata import dataset_metadata
from tensorflow_transform.tf_metadata import schema_utils

import pprint
import tempfile

python: 3.7.11；windows: 10；tensorflow-transform: 1.5.0

# Ignore the warnings
tf.get_logger().setLevel('ERROR')

# a temporary directory is needed when analyzing the data
with tft_beam.Context(temp_dir=tempfile.mkdtemp()):
    
    # define the pipeline using Apache Beam syntax
    transformed_dataset, transform_fn = (
        
        # analyze and transform the dataset using the preprocessing function
        (raw_data, raw_data_metadata) | tft_beam.AnalyzeAndTransformDataset(
            preprocessing_fn)
    )

# unpack the transformed dataset
transformed_data, transformed_metadata = transformed_dataset

# print the results
print('\nRaw data:\n{}\n'.format(pprint.pformat(raw_data)))
print('Transformed data:\n{}'.format(pprint.pformat(transformed_data)))

错误快照：

usage: ipykernel_launcher.py [-h] [--dataflow_endpoint DATAFLOW_ENDPOINT]
                             [--project PROJECT] [--job_name JOB_NAME]
                             [--staging_location STAGING_LOCATION]
                             [--temp_location TEMP_LOCATION] [--region REGION]
                             [--service_account_email SERVICE_ACCOUNT_EMAIL]
                             [--no_auth]
                             [--template_location TEMPLATE_LOCATION]
                             [--label LABELS] [--update]
                             [--transform_name_mapping TRANSFORM_NAME_MAPPING]
                             [--enable_streaming_engine]
                             [--dataflow_kms_key DATAFLOW_KMS_KEY]
                             [--create_from_snapshot CREATE_FROM_SNAPSHOT]
                             [--flexrs_goal {COST_OPTIMIZED,SPEED_OPTIMIZED}]
                             [--dataflow_service_option DATAFLOW_SERVICE_OPTIONS]
                             [--enable_hot_key_logging]
ipykernel_launcher.py: error: argument --flexrs_goal: invalid choice: 'C:\\Users\\abc\\AppData\\Local\\Temp\\tmp-17163boS9n25ijAn.json' (choose from 'COST_OPTIMIZED', 'SPEED_OPTIMIZED')
ERROR:root:Internal Python error in the inspect module.
Below is the traceback from this internal error.

有人能帮忙吗？

Answer 1

此问题可能是VS代码特有的。我在macOS上遇到了同样的问题。由于某些原因，ipykernel似乎不能很好地与tft_beam一起工作。

这里有几个对我有用的解决办法。

1. 从终端运行一个木星笔记本，并使用远程选项

将VS代码连接到它。

1. 使用it转换将笔记本转换为python文件并运行它。