(tool 1 ) : graph_metrics 

bazel-bin/third_party/tensorflow/python/tools/graph_metrics \

--graph tensorflow_inception_graph.pb \

--statistics=weight_parameters,flops

reference: https://github.com/tensorflow/tensorflow/blob/fe454464681b036ff7fed3e42c6bb541fa52dd7c/tensorflow/python/tools/graph_metrics.py 

(tool 2 ) : Timeline 

 an example :

    run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)

    run_metadata = tf.RunMetadata()

    summary, _ = sess.run([merged, train_step],

                          feed_dict=feed_dict(True),

                          options=run_options,

                          run_metadata=run_metadata)

    train_writer.add_run_metadata(run_metadata, 'step%03d' % i)

    train_writer.add_summary(summary, i)

    print('Adding run metadata for', i)

    tl = timeline.Timeline(run_metadata.step_stats)

    print(tl.generate_chrome_trace_format(show_memory=True))

    trace_file = tf.gfile.Open(name='timeline', mode='w')

    trace_file.write(tl.generate_chrome_trace_format(show_memory=True))

reference: https://stackoverflow.com/questions/36123740/is-there-a-way-of-determining-how-much-gpu-memory-is-in-use-by-tensorflow

(tool 3 ) : Profiler Tool

#To get memory information, you need --graph_path and --run_meta_path

tfprof> graph -max_depth 10000000 -step 0 -account_type_regexes .* -output timeline:outfile=<filename>

generating trace file.

reference: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/profiler/g3doc/profile_memory.md

https://github.com/tensorflow/tensorflow/tree/master/tensorflow/core/profiler#visualize-time-and-memory

(tool 4 ) : Benchmark Tool 

bazel-bin/tensorflow/tools/benchmark/benchmark_model \

  --graph=tensorflow_inception_graph.pb \

  --input_layer="input:0" \

  --input_layer_shape="1,224,224,3" \

  --input_layer_type="float" \

  --output_layer="output:0"

reference: https://github.com/tensorflow/tensorflow/tree/master/tensorflow/tools/benchmark