I tried using script of image label classification from examples folder where I tried to compare two scripts one with ngraph bridge configuration and other without it, I found that when I calculate time elapsed using the code mentioned below,
with tf.compat.v1.Session(graph=graph, config=config_ngraph_enabled) as sess:
Warmup
sess.run(output_operation.outputs[0], {
input_operation.outputs[0]: t})
# Run
import time
start = time.time()
results = sess.run(output_operation.outputs[0], {
input_operation.outputs[0]: t
})
elapsed = time.time() - start
print('Time elapsed: %f seconds' % elapsed)
results = np.squeeze(results)
I observe that the script without this above ngraph config implementation has faster time elapsed than the script with above ngraph config implementation.
Can anyone please guide me where I am going wrong?
Ideally the script with ngraph should have lesser computation time than the script without ngraph!!
I tried using script of image label classification from examples folder where I tried to compare two scripts one with ngraph bridge configuration and other without it, I found that when I calculate time elapsed using the code mentioned below,
config = tf.compat.v1.ConfigProto() config_ngraph_enabled = ngraph_bridge.update_config(config)
with tf.compat.v1.Session(graph=graph, config=config_ngraph_enabled) as sess:
Warmup
results = np.squeeze(results)
I observe that the script without this above ngraph config implementation has faster time elapsed than the script with above ngraph config implementation.
Can anyone please guide me where I am going wrong?
Ideally the script with ngraph should have lesser computation time than the script without ngraph!!