从TensorFlow导入MNIST数据集遇SSL证书错误（Anaconda环境）求助

Hey there, let's work through this SSL certificate issue and get your MNIST dataset loaded properly. First, let's address a couple of key points that might be causing your trouble:

1. You might have installed the wrong package (not certified but certifi) #

You mentioned uninstalling/reinstalling certified, but the correct Python package for managing SSL certificates is certifi. That's likely a typo that's been holding you back. Run these commands to fix it:

pip uninstall certified
pip install --upgrade certifi

2. The tensorflow.examples.tutorials.mnist module is deprecated #

This module is no longer maintained in newer TensorFlow versions (especially TF 2.x), which can lead to unexpected errors including SSL issues. The recommended approach is to use the built-in Keras dataset loader instead—it's more reliable and avoids these legacy problems:

import tensorflow as tf

# Load MNIST directly via Keras
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()

# If you need the same structure as input_data (like one-hot labels), you can convert it:
y_train = tf.keras.utils.to_categorical(y_train, 10)
y_test = tf.keras.utils.to_categorical(y_test, 10)

3. If you must use the legacy input_data module (TF 1.x only) #

If you're stuck using TensorFlow 1.x and the old module, here are two ways to bypass the SSL error:

Option A: Disable SSL verification (temporary fix) #

This skips certificate checks for the download. Note: This isn't secure for production, but works for testing:

import ssl
# Disable SSL verification
ssl._create_default_https_context = ssl._create_unverified_context

from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("tmp/data/", one_hot=True)

Option B: Manually download the dataset #

Download the MNIST files directly and place them in your tmp/data/ folder. The required files are:

• train-images-idx3-ubyte.gz
• train-labels-idx1-ubyte.gz
• t10k-images-idx3-ubyte.gz
• t10k-labels-idx1-ubyte.gz

When you run read_data_sets, it will detect the local files and skip the download, avoiding the SSL error entirely.

内容的提问来源于stack exchange，提问作者Prashanth Pradeep