Hey there, I totally get how frustrating this can be—TensorFlow setup has tripped up way more people than you think, and half the battle is figuring out where the mismatch is happening. Let’s walk through what you’ve tried so far, then tackle the common pitfalls that might be derailing you.

What You’ve Attempted So Far: #

• Downloaded Anaconda with Python 3.6
• Launched Command Prompt as Administrator
• Ran conda create -n tensorflow pip python=3.6 to create a dedicated environment
• Activated the tensorflow environment
• Ran pip install --upgrade tensorflow
• Tried importing TensorFlow with >>> import tensorflow

Common Fixes to Get You Up and Running: #

1. Double-Check Environment Activation
   Make sure your Command Prompt actually shows (tensorflow) at the start of the line after activation. If it doesn’t, you might be using an older Anaconda version where the activation command is activate tensorflow instead of conda activate tensorflow. No activation means you’re installing TensorFlow to your base environment (or worse, your system Python) instead of the dedicated one.

2. Pin TensorFlow to a Version Compatible with Python 3.6
   Here’s a big one: modern TensorFlow versions (2.4+) don’t support Python 3.6 anymore. When you run pip install --upgrade tensorflow, pip will try to grab the latest version, which won’t work for your Python version. Instead, install a compatible release:

   pip install tensorflow==2.3.4
   # Or if you prefer TensorFlow 1.x:
   pip install tensorflow==1.15.5
   

3. Verify You’re Using the Right Pip
   After activating the tensorflow environment, run pip --version. The output should show a path pointing to your Anaconda envs/tensorflow folder. If it shows your system Python’s pip path, you’re not actually using the environment’s package manager—this usually happens if your system Python is higher in the PATH than Anaconda.

4. Fix Missing DLL Errors (If You Get Them)
   If importing TensorFlow throws a DLL-related error, you’re likely missing the Microsoft Visual C++ Redistributable packages. Install the 2015-2019 version (it includes all the required runtimes for older TensorFlow builds).

Test the Installation #

Once you’ve worked through the above steps, activate your environment again, launch Python, and run:

import tensorflow as tf
print(tf.__version__)

If this prints out the version number you installed, you’re good to go! If you hit a specific error message (like import errors, missing packages, etc.), drop it here and we can troubleshoot further.

内容的提问来源于stack exchange，提问作者mathsnerd