diff --git a/Dockerfile.train.tmpl b/Dockerfile.train.tmpl
index 3e534b6e..6d33cb25 100644
--- a/Dockerfile.train.tmpl
+++ b/Dockerfile.train.tmpl
@@ -43,6 +43,9 @@ RUN pip3 install --upgrade native_client/ctcdecode/dist/*.whl
 # Prepare deps
 RUN pip3 install --upgrade pip==20.2.2 wheel==0.34.2 setuptools==49.6.0
 
+# FIXME: Hack because no 1.15.4 Docker image
+RUN pip3 install --upgrade tensorflow-gpu==1.15.4
+
 # Install DeepSpeech
 #  - No need for the decoder since we did it earlier
 #  - There is already correct TensorFlow GPU installed on the base image,
diff --git a/doc/TRAINING.rst b/doc/TRAINING.rst
index 08215e8a..ba38b64b 100644
--- a/doc/TRAINING.rst
+++ b/doc/TRAINING.rst
@@ -67,7 +67,7 @@ If you have a capable (NVIDIA, at least 8GB of VRAM) GPU, it is highly recommend
 .. code-block:: bash
 
    pip3 uninstall tensorflow
-   pip3 install 'tensorflow-gpu==1.15.2'
+   pip3 install 'tensorflow-gpu==1.15.4'
 
 Please ensure you have the required :ref:`CUDA dependency <cuda-deps>`.
 
diff --git a/setup.py b/setup.py
index 91a9af5b..16c17eb7 100644
--- a/setup.py
+++ b/setup.py
@@ -74,7 +74,7 @@ def main():
     ]
 
     tensorflow_pypi_dep = [
-        'tensorflow == 1.15.2'
+        'tensorflow == 1.15.4'
     ]
 
     # Due to pip craziness environment variables are the only consistent way to