How do you clean up a tf.data.Iterator?
I am attempting to train a Keras model using date from a tensorflow.data input pipleline. Instead of training the model in one single run, I intend to do early stopping when validation performance deteriorates and continue the training with a larger batch size. My code for doing this is shown below:
batch_sizes = [256, 512, 1024]
for batch_size in batch_sizes:
input_fn = input_fn_helper(batch_size, ...)
training_set = input_fn().make_one_shot_iterator()
input_fn_test = input_fn_test_helper(batch_size, ...)
testing_set = input_fn_test().make_one_shot_iterator()
model.fit(training_set,
steps_per_epoch=(n_train / batch_size),
epochs=max_epochs,
validation_data=testing_set,
validation_steps=(n_test / batch_size),
callbacks=callbacks)
As you can see I construct a new input pipeline for each increase in batch size (input_fn()returns a tf.data.Dataset). The behaviour I get from this is what I would expect, so it does what it is supposed to do. The problem I do have is that my scripts memory footprint increases every time the loop runs, i.e. it seems the preceding instance of training_set and testing_set are not freed up by being overwritten in the next step. This begs a few questions:
- Am I doing something terribly wrong here?
- Is there a canonical way of ensuring that a
tf.data.Iteratoris properly disposed of when no longer needed?
python tensorflow memory keras
asked Nov 10 '18 at 12:20
Harald Husum
346210
add a comment |
I am attempting to train a Keras model using date from a tensorflow.data input pipleline. Instead of training the model in one single run, I intend to do early stopping when validation performance deteriorates and continue the training with a larger batch size. My code for doing this is shown below:
batch_sizes = [256, 512, 1024]
for batch_size in batch_sizes:
input_fn = input_fn_helper(batch_size, ...)
training_set = input_fn().make_one_shot_iterator()
input_fn_test = input_fn_test_helper(batch_size, ...)
testing_set = input_fn_test().make_one_shot_iterator()
model.fit(training_set,
steps_per_epoch=(n_train / batch_size),
epochs=max_epochs,
validation_data=testing_set,
validation_steps=(n_test / batch_size),
callbacks=callbacks)
As you can see I construct a new input pipeline for each increase in batch size (input_fn()returns a tf.data.Dataset). The behaviour I get from this is what I would expect, so it does what it is supposed to do. The problem I do have is that my scripts memory footprint increases every time the loop runs, i.e. it seems the preceding instance of training_set and testing_set are not freed up by being overwritten in the next step. This begs a few questions:
- Am I doing something terribly wrong here?
- Is there a canonical way of ensuring that a
tf.data.Iteratoris properly disposed of when no longer needed?
python tensorflow memory keras
asked Nov 10 '18 at 12:20
Harald Husum
346210
have you tried using reinitializable iterator to define one iterator structure and reinitialize it with whatever new dataset object you want to use? Check out the importing data guide for more detailed explanation!
– kvish
Nov 12 '18 at 15:31
add a comment |
I am attempting to train a Keras model using date from a tensorflow.data input pipleline. Instead of training the model in one single run, I intend to do early stopping when validation performance deteriorates and continue the training with a larger batch size. My code for doing this is shown below:
batch_sizes = [256, 512, 1024]
for batch_size in batch_sizes:
input_fn = input_fn_helper(batch_size, ...)
training_set = input_fn().make_one_shot_iterator()
input_fn_test = input_fn_test_helper(batch_size, ...)
testing_set = input_fn_test().make_one_shot_iterator()
model.fit(training_set,
steps_per_epoch=(n_train / batch_size),
epochs=max_epochs,
validation_data=testing_set,
validation_steps=(n_test / batch_size),
callbacks=callbacks)
As you can see I construct a new input pipeline for each increase in batch size (input_fn()returns a tf.data.Dataset). The behaviour I get from this is what I would expect, so it does what it is supposed to do. The problem I do have is that my scripts memory footprint increases every time the loop runs, i.e. it seems the preceding instance of training_set and testing_set are not freed up by being overwritten in the next step. This begs a few questions:
- Am I doing something terribly wrong here?
- Is there a canonical way of ensuring that a
tf.data.Iteratoris properly disposed of when no longer needed?
python tensorflow memory keras
asked Nov 10 '18 at 12:20
Harald Husum
346210
I am attempting to train a Keras model using date from a tensorflow.data input pipleline. Instead of training the model in one single run, I intend to do early stopping when validation performance deteriorates and continue the training with a larger batch size. My code for doing this is shown below:
batch_sizes = [256, 512, 1024]
for batch_size in batch_sizes:
input_fn = input_fn_helper(batch_size, ...)
training_set = input_fn().make_one_shot_iterator()
input_fn_test = input_fn_test_helper(batch_size, ...)
testing_set = input_fn_test().make_one_shot_iterator()
model.fit(training_set,
steps_per_epoch=(n_train / batch_size),
epochs=max_epochs,
validation_data=testing_set,
validation_steps=(n_test / batch_size),
callbacks=callbacks)
As you can see I construct a new input pipeline for each increase in batch size (input_fn()returns a tf.data.Dataset). The behaviour I get from this is what I would expect, so it does what it is supposed to do. The problem I do have is that my scripts memory footprint increases every time the loop runs, i.e. it seems the preceding instance of training_set and testing_set are not freed up by being overwritten in the next step. This begs a few questions:
- Am I doing something terribly wrong here?
- Is there a canonical way of ensuring that a
tf.data.Iteratoris properly disposed of when no longer needed?
python tensorflow memory keras
python tensorflow memory keras
asked Nov 10 '18 at 12:20
Harald Husum
346210
asked Nov 10 '18 at 12:20
Harald Husum
346210
asked Nov 10 '18 at 12:20
Harald Husum
346210
asked Nov 10 '18 at 12:20
Harald Husum
346210
asked Nov 10 '18 at 12:20
Harald Husum
346210
346210
have you tried using reinitializable iterator to define one iterator structure and reinitialize it with whatever new dataset object you want to use? Check out the importing data guide for more detailed explanation!
– kvish
Nov 12 '18 at 15:31
add a comment |
have you tried using reinitializable iterator to define one iterator structure and reinitialize it with whatever new dataset object you want to use? Check out the importing data guide for more detailed explanation!
– kvish
Nov 12 '18 at 15:31
have you tried using reinitializable iterator to define one iterator structure and reinitialize it with whatever new dataset object you want to use? Check out the importing data guide for more detailed explanation!
– kvish
Nov 12 '18 at 15:31
have you tried using reinitializable iterator to define one iterator structure and reinitialize it with whatever new dataset object you want to use? Check out the importing data guide for more detailed explanation!
– kvish
Nov 12 '18 at 15:31
add a comment |
0
active
oldest
votes
Your Answer
StackExchange.ifUsing("editor", function ()
StackExchange.using("externalEditor", function ()
StackExchange.using("snippets", function ()
StackExchange.snippets.init();
);
);
, "code-snippets");
StackExchange.ready(function()
var channelOptions =
tags: "".split(" "),
id: "1"
;
initTagRenderer("".split(" "), "".split(" "), channelOptions);
StackExchange.using("externalEditor", function()
// Have to fire editor after snippets, if snippets enabled
if (StackExchange.settings.snippets.snippetsEnabled)
StackExchange.using("snippets", function()
createEditor();
);
else
createEditor();
);
function createEditor()
StackExchange.prepareEditor(
heartbeatType: 'answer',
autoActivateHeartbeat: false,
convertImagesToLinks: true,
noModals: true,
showLowRepImageUploadWarning: true,
reputationToPostImages: 10,
bindNavPrevention: true,
postfix: "",
imageUploader:
brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
allowUrls: true
,
onDemand: true,
discardSelector: ".discard-answer"
,immediatelyShowMarkdownHelp:true
);
);
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f53238889%2fhow-do-you-clean-up-a-tf-data-iterator%23new-answer', 'question_page');
);
Post as a guest
Required, but never shown
0
active
oldest
votes
0
active
oldest
votes
active
oldest
votes
active
oldest
votes
Thanks for contributing an answer to Stack Overflow!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
To learn more, see our tips on writing great answers.
Some of your past answers have not been well-received, and you're in danger of being blocked from answering.
Please pay close attention to the following guidance:
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
To learn more, see our tips on writing great answers.
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function ()
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f53238889%2fhow-do-you-clean-up-a-tf-data-iterator%23new-answer', 'question_page');
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function ()
StackExchange.helpers.onClickDraftSave('#login-link');
);
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
have you tried using reinitializable iterator to define one iterator structure and reinitialize it with whatever new dataset object you want to use? Check out the importing data guide for more detailed explanation!
– kvish
Nov 12 '18 at 15:31