I have recently done my course on Tensorflow and while doing my assignments, I have encountered a problem when using tensorflow profiler to analysis my code.
I followed the instructions on this website:
https://github.com/tensorflow/tensorflow/tree/master/tensorflow/core/profiler. Also, google gperftools ( https://github.com/gperftools/gperftools ) was built and installed on my computer.
I follow the steps below to find the most expensive operations of my code .
Step 1. Generate pprof file by running the following code : "code -select accelerator_micros -max_depth 100000 -output pprof:outfile=/tmp/profiles/deep.prof".
Step 2. Visualize the generated file by running "pprof -pdf --nodecount=100 --sample_index=1 /usr/bin/python /tmp/profiles/deep.prof >/tmp/profiles/deep.pdf".
I expect to get a call graph with tensorflow operations and running time. However I only got a table as shown below. I wonder whether I misunderstood some of the instructions. Would someone give me some advice? pprof output pdf file
Thanks & Regards