Sometimes, we want to prevent tensorflow from allocating all of a GPU’s memory with Python.
In this article, we’ll look at how to prevent tensorflow from allocating all of a GPU’s memory with Python.
How to prevent tensorflow from allocating all of a GPU’s memory with Python?
To prevent tensorflow from allocating all of a GPU’s memory with Python, we can create a GPUOptions object.
For instance, we write
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.333)
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))
to create a GPUOptions object with the per_process_gpu_memory_fraction argument set to 0.333 to make tensorflow use 33.3% of the GPU’s memory.
Then we start a session with gpu_options with
sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))
Conclusion
To prevent tensorflow from allocating all of a GPU’s memory with Python, we can create a GPUOptions object.
