Part Number:TDA2EXEVM
Hi
in TIDeepLearningLibrary_UserGuide.pdf version 0305, there are some TIDL Limitation:
Convolution Layer
– We have tested the kernel size up to 7x7 (Shall work for higher values also, but not validated)
– Dilation is tested with 1,2,4.
– We support only stride 1 and 2. Any value higher than 2 is not supported.
– Dense convolution flow is supported for only 1x1 and 3x3 kernels with stride = 1 and dilation =1
– Maximum number of input and output channel supported in 1024
My question:
(1) currently we use the first convolution layer of the standard Alex net, can kernel 11 stride4, TIDL latest version be supported?
(2) currently support kernels with Dense convolution flows of 5x5 or 7x7?
(3) my input of the image is the normalized floating point . Is it supported?
(4) the network designed by us has gone through initial l1reg and sparse. After using tidl_model_import.out.exe, it only displays the statistics of Macs at each layer, but there is no sparse statistical result. Why?