If including tracebacks, please include the full Other info / logs Include any logs or source code that would be helpful toĭiagnose the problem. Print("Total allocated size: %.1f KiB" % (total / 1024))ĭataset = tf._generator(generator, output_types=tf.int32, output_shapes=) Total = sum(stat.size for stat in top_stats) Line = linecache.getline(frame.filename, frame.lineno).strip() % (index, filename, frame.lineno, stat.size / 1024)) # replace "/path/to/module/file.py" with "module/file.py"įilename = os.sep.join((os.sep)) Top_stats = snapshot.statistics(key_type)įor index, stat in enumerate(top_stats, 1): Os.environ = "-1"ĭef display_top(snapshot, key_type='lineno', limit=3): If possible, please share a link to Colab/Jupyter/any notebook. Provide a reproducible test case that is the bare minimum necessary to generate Memory should be released when no reference exists for the dataset. Tf._generator leaks memory after each call even if followed by gc.collect(). TF 2.0: python -c "import tensorflow as tf print(tf.version.GIT_VERSION, tf.version.VERSION)" TF 1.0: python -c "import tensorflow as tf print(tf.GIT_VERSION, tf.VERSION)" 2. You can also obtain the TensorFlow version with: 1. You can collect some of this information using our environment capture GCC/Compiler version (if compiling from.Bazel version (if compiling from source):.TensorFlow version (use command below): v2.1.0-rc2-17-ge5bf8de 2.1.0.Have I written custom code (as opposed to using a stockĮxample script provided in TensorFlow): No.We only address code/doc bugs, performance issues, feature requests andīuild/installation issues on GitHub.