11-26-2019 04:52 AM
Is there any documentation about DECENT or how can I check all weights after quanzation? I want to know the exact value of INT8 weights.
Which method dose DENCT use, mixed FP32/INT8 or just INT8?
And last one, does DENCENT support all operators?
11-27-2019 02:51 AM
11-27-2019 04:11 AM - edited 11-27-2019 04:12 AM
I have read UG1327, but it did not have details about DNNDK or DECNET.
I want to know what DECNET did to my pb model and the INT8 weights after quantization