# tf.config.experimental.reset_memory_stats

<br />

|---|
| [![](https://www.tensorflow.org/images/GitHub-Mark-32px.png) View source on GitHub](https://github.com/tensorflow/tensorflow/blob/v2.13.1/tensorflow/python/framework/config.py#L603-L641) |

Resets the tracked memory stats for the chosen device.

#### View aliases


**Compat aliases for migration**

See
[Migration guide](https://www.tensorflow.org/guide/migrate) for
more details.

[`tf.compat.v1.config.experimental.reset_memory_stats`](https://www.tensorflow.org/api_docs/python/tf/config/experimental/reset_memory_stats)

<br />

    tf.config.experimental.reset_memory_stats(
        device
    )

This function sets the tracked peak memory for a device to the device's
current memory usage. This allows you to measure the peak memory usage for a
specific part of your program. For example:

    if tf.config.list_physical_devices('GPU'):
      # Sets the peak memory to the current memory.
      tf.config.experimental.reset_memory_stats('GPU:0')
      # Creates the first peak memory usage.
      x1 = tf.ones(1000 * 1000, dtype=tf.float64)
      del x1 # Frees the memory referenced by `x1`.
      peak1 = tf.config.experimental.get_memory_info('GPU:0')['peak']
      # Sets the peak memory to the current memory again.
      tf.config.experimental.reset_memory_stats('GPU:0')
      # Creates the second peak memory usage.
      x2 = tf.ones(1000 * 1000, dtype=tf.float32)
      del x2
      peak2 = tf.config.experimental.get_memory_info('GPU:0')['peak']
      assert peak2 < peak1  # tf.float32 consumes less memory than tf.float64.

Currently only supports GPU and TPU. If called on a CPU device, an exception
will be raised.

<br />

<br />

<br />

| ## Args ||
|---|---|
| `device` | Device string to reset the memory stats, e.g. `"GPU:0"`, `"TPU:0"`. See https://www.tensorflow.org/api_docs/python/tf/device for specifying device strings. |

<br />

<br />

<br />

<br />

| ## Raises ||
|---|---|
| `ValueError` | No device found with the device name, like '"nonexistent"'. |
| `ValueError` | Invalid device name, like '"GPU"', '"CPU:GPU"', '"CPU:"'. |
| `ValueError` | Multiple devices matched with the device name. |
| `ValueError` | Memory statistics not tracked or clearing memory statistics not supported, like '"CPU:0"'. |

<br />