ModuleNotFoundError: No module named 'tensorflow' (Python) — how to fix
Traceback (most recent call last):
File "train.py", line 1, in <module>
import tensorflow as tf
ModuleNotFoundError: No module named 'tensorflow'
Quick fix: Check your Python version (python --version) — TensorFlow only ships wheels for 64-bit CPython 3.9-3.12 (3.13 needs TF 2.20+). Then install with the same interpreter: python -m pip install tensorflow (Linux GPU: python -m pip install "tensorflow[and-cuda]"). Verify: python -c "import tensorflow as tf; print(tf.__version__)".
What it means
Python could not find a tensorflow package on sys.path. Either it was never installed successfully, or it was installed into a different interpreter (system Python vs. venv vs. conda vs. Jupyter kernel).
TensorFlow is stricter than most packages: it only publishes wheels for specific Python versions and 64-bit platforms. If no wheel matches, pip install tensorflow fails with ERROR: No matching distribution found for tensorflow, and the later import fails with this error.
Fix 1: use a supported Python version
python --version
python -c "import struct; print(struct.calcsize('P') * 8, 'bit')"
As a rule of thumb, TensorFlow supports the four or so Python releases before the newest one, and support for a brand new Python arrives months later. TF 2.16-2.19 support Python 3.9-3.12; Python 3.13 support came with TF 2.20. 32-bit Python is never supported. Check the current table at tensorflow.org/install/pip.
If your Python is too new, create a venv with a supported version:
# Ubuntu, e.g. with deadsnakes or pyenv
pyenv install 3.12
pyenv local 3.12
python -m venv .venv && source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install tensorflow
Upgrading pip matters: old pip versions don't recognise newer wheel tags and report "no matching distribution".
Fix 2: install into the interpreter that runs your script
which python
python -m pip --version # which Python this pip belongs to
python -m pip install tensorflow
python -c "import tensorflow as tf; print(tf.__version__)"
Always prefer python -m pip over a bare pip. On Debian/Ubuntu 23.04+ installing into the system Python is blocked (externally-managed-environment) — use a venv as above.
Fix 3: GPU and platform-specific packages
| Platform | Install |
|---|---|
| Linux / WSL2 with NVIDIA GPU | python -m pip install "tensorflow[and-cuda]" |
| Linux / Windows CPU | python -m pip install tensorflow |
| macOS Apple Silicon | python -m pip install tensorflow (TF 2.13+); add tensorflow-metal for GPU |
| Windows native GPU | not supported after TF 2.10 — use WSL2 |
The old names tensorflow-gpu and (for TF 2.13+) tensorflow-macos are deprecated; installing them on current versions fails or installs an outdated release.
Fix 4: Jupyter / VS Code use a different kernel
Install from inside the notebook so it targets the kernel's Python:
%pip install tensorflow
Then restart the kernel. In VS Code, select the venv interpreter (Ctrl+Shift+P → "Python: Select Interpreter").
Fix 5: conda environments
conda create -n tf python=3.11
conda activate tf
python -m pip install tensorflow
Google recommends installing TensorFlow with pip even inside conda; mixing conda install tensorflow and pip often yields broken installs.
Fix 6: a local file shadows the package
A file named tensorflow.py (or a folder tensorflow/) in your project shadows the real package and causes confusing import errors. Rename it and delete __pycache__.
Keras 3 note
If the error mentions tensorflow when you only imported keras, Keras 3 needs a backend: install tensorflow (or set KERAS_BACKEND=jax/torch and install that).
Last updated 2026-11-16.