liang1232018
liang1232018
I have read the upstream source code git from https://github.com/riscv/riscv-gnu-toolchain . ALL the patch in baremetal.patch has already been added in the source code. so , I think it's not...
Thank you for your feedback. It seems that the input datatype is wrong when starting to execute the NPU garphs. Have you ever changed the model datatype? It's hard to...
Thank you for the detailed log. It seems that the prefilling procedure has completed correctly, with the `lm_head` operator executing and the first token `Fre` being generated. However, the issue...
> @oreomaker Thanks for your help. I'm using S24U (Mem 12GB). This is the tail log. I'm wondering why it consumes a lot of memory, even if running a small...
> It is Segmentation fault too on OPPO Find X7 Ultral (Snapdrag8Gen3， DDR = 16GB) ! > > load time: 1474.77 ms token time: nan ms inference speed: nan tokens/s...
你好， 感谢关注！ 该部分设计主要是考虑地址可能存在不对齐的可能性。该指令将地址传进去是依据地址不对齐的偏移来拼接sline1c和sline1p两相邻的vector。
Hi, Thank you for your attention on the mllm and for providing this issue. In the earlier version of QNN 2.20, it was possible for LinearINT8 parameter names to have...