16:52 tomeu: pH5: btw, I found this list of vsimulator configs: https://github.com/STMicroelectronics/gcnano-binaries/blob/5d02efd5cb4cfa85307633891f3cf87550a8bc1d/gcnano-driver-stm32mp/hal/kernel/inc/gc_vsim_configs.h#L56
17:00 tomeu: there is something quite weird though: when setting VIPPICO_V3_PID0X99 the simulator generates a non-huffman BO, but the driver on the device generates huffman
17:02 tomeu: but the huffman-like BO that the real driver generates is different from that generated by the vsimulator for VIP8000NANOSI_PLUS_PID0X9F
17:02 tomeu: the header is the same, but the bitstream is different
17:16 tomeu: hmm, I think they just have the bias before the weights, for each output channel
17:34 tomeu: yep, seems to be that