TF's documentation on their quantization script appears to only include --mode=eightbit (July 24, 2017).
What are each of the modes used for? i.e. as indicated in quantize_graph.py (round, quantize, weights, weights_rounded):
flags.DEFINE_string("mode", "round",
"""What transformation to apply (round, quantize,"""
""" eightbit, weights, or weights_rounded).""")
As a complementary question, I tried to trace through the script and noticed that eightbit doesn't actually call any of the functions that compute min/max values (ex. quantize_weight_eightbit). So what is mode=eightbit doing exactly?
