search

nvdla / sw

NVDLA SW
Other
489 stars 193 forks source link

Running AlexNet on nv_small #185

Open gitosu67 opened 4 years ago

gitosu67 commented 4 years ago

Hi all,

I am running the preloadable AlexNet file given in the repo, using a nv_small architecture. I am giving an input image of size 227x227x3. I execute the loadable as follows:

./nvdla_runtime --loadable regression/flatbufs/kmd/NN/NN_L0_1_small_fbuf --image images/image0.jpg --rawdump

I am getting an output file as output.dimg with 142560 class labels. Also, it gives the same output classification for different images of size 227x227x3. I am not sure if I am doing something wrong here or how to fix this problem. I am attaching the output file and the image.

Any help will be highly appreciated, Thanks! image_57 output.dimg.txt

albertlee987 commented 4 years ago

Hi Okaymaddy， do you run the nvdla_runtime use Virtual Platform (vp)?

Scriabing commented 4 years ago

Hi， Okaymaddy ， i am using zcu102 to deploy nv_small . Do you use https://github.com/SCLUO/ITRI-OpenDLA/blob/master/FPGA_RTL_nvsmall64/？or use nvdla/hw/nv_small to package ip? Could you have a blog to description all your steps? Could you contact with me with z1403154226@163.com

nagendra7890 commented 3 years ago

Hi, @Scriabing I am also using ZCU102 to deploy nv_small. I generated the bitstream file using Xilinx vivado. how to control the NVDLA through a software environment. it's quite confusing can you please help me out to approach the process.

nagendra7890 commented 3 years ago

hello, @Okaymaddy @Scriabing I am not able to see the execution time and test result and there was no error. can you give any suggestions? https://github.com/nvdla/sw/issues/177#issuecomment-917419962