tf / DeviceSpec

See Stable See Nightly

TensorFlow 1 version

View source on GitHub

Represents a (possibly partial) specification for a TensorFlow device.

tf.DeviceSpec(
    job=None, replica=None, task=None, device_type=None, device_index=None
)

DeviceSpecs are used throughout TensorFlow to describe where state is stored and computations occur. Using DeviceSpec allows you to parse device spec strings to verify their validity, merge them or compose them programmatically.

Example:

# Place the operations on device "GPU:0" in the "ps" job.
device_spec = DeviceSpec(job="ps", device_type="GPU", device_index=0)
with tf.device(device_spec.to_string()):
  # Both my_var and squared_var will be placed on /job:ps/device:GPU:0.
  my_var = tf.Variable(..., name="my_variable")
  squared_var = tf.square(my_var)

With eager execution disabled (by default in TensorFlow 1.x and by calling disable_eager_execution() in TensorFlow 2.x), the following syntax can be used:

tf.compat.v1.disable_eager_execution()

# Same as previous
device_spec = DeviceSpec(job="ps", device_type="GPU", device_index=0)
# No need of .to_string() method.
with tf.device(device_spec):
  my_var = tf.Variable(..., name="my_variable")
  squared_var = tf.square(my_var)
 ```

If a `DeviceSpec` is partially specified, it will be merged with other
`DeviceSpec`s according to the scope in which it is defined. `DeviceSpec`
components defined in inner scopes take precedence over those defined in
outer scopes.

```python
gpu0_spec = DeviceSpec(job="ps", device_type="GPU", device_index=0)
with tf.device(DeviceSpec(job="train").to_string()):
  with tf.device(gpu0_spec.to_string()):
    # Nodes created here will be assigned to /job:ps/device:GPU:0.
  with tf.device(DeviceSpec(device_type="GPU", device_index=1).to_string()):
    # Nodes created here will be assigned to /job:train/device:GPU:1.

A DeviceSpec consists of 5 components -- each of which is optionally specified:

• Job: The job name.
• Replica: The replica index.
• Task: The task index.
• Device type: The device type string (e.g. "CPU" or "GPU").
• Device index: The device index.

Methods

from_string

View source

@classmethod
from_string(
    spec
)

Construct a DeviceSpec from a string.

make_merged_spec

View source

make_merged_spec(
    dev
)

Returns a new DeviceSpec which incorporates dev.

When combining specs, dev will take precedence over the current spec. So for instance:

first_spec = tf.DeviceSpec(job=0, device_type="CPU")
second_spec = tf.DeviceSpec(device_type="GPU")
combined_spec = first_spec.make_merged_spec(second_spec)

is equivalent to:

combined_spec = tf.DeviceSpec(job=0, device_type="GPU")

parse_from_string

View source

parse_from_string(
    spec
)

Parse a DeviceSpec name into its components.

2.x behavior change: In TensorFlow 1.x, this function mutates its own state and returns itself. In 2.x, DeviceSpecs are immutable, and this function will return a DeviceSpec which contains the spec.

Recommended:

```
# my_spec and my_updated_spec are unrelated.
my_spec = tf.DeviceSpec.from_string("/CPU:0")
my_updated_spec = tf.DeviceSpec.from_string("/GPU:0")
with tf.device(my_updated_spec):
  ...
```

Will work in 1.x and 2.x (though deprecated in 2.x):

```
my_spec = tf.DeviceSpec.from_string("/CPU:0")
my_updated_spec = my_spec.parse_from_string("/GPU:0")
with tf.device(my_updated_spec):
  ...
```

Will NOT work in 2.x:

```
my_spec = tf.DeviceSpec.from_string("/CPU:0")
my_spec.parse_from_string("/GPU:0")  # <== Will not update my_spec
with tf.device(my_spec):
  ...
```

In general, DeviceSpec.from_string should completely replace DeviceSpec.parse_from_string, and DeviceSpec.replace should completely replace setting attributes directly.

replace

View source

replace(
    **kwargs
)

Convenience method for making a new DeviceSpec by overriding fields.

For instance:

my_spec = DeviceSpec=(job="my_job", device="CPU")
my_updated_spec = my_spec.replace(device="GPU")
my_other_spec = my_spec.replace(device=None)

to_string

View source

to_string()

Return a string representation of this DeviceSpec.

__eq__

View source

__eq__(
    other
)

Checks if the other DeviceSpec is same as the current instance, eg have

same value for all the internal fields.