nullname

And for the `graph_compute cost`, thought we can have a breakdown with the profiler output that capture from logcat

> what other inference frameworks support Qualcomm NPU deployment? Can check out [MLLM](https://github.com/UbiquitousLearning/mllm) - they have their own model format and NPU support. > Does Qualcomm officially use QNN-SDK and...

Hey @finneyyan , it’s been a while! We’ve made several improvements to the hexagon-npu backend and are seeing significant performance gains. When you have a moment, could you test your...

Hi > 当前支持视觉大模型在npu上的调度吗 这个backend在hexagon的npu上实现了部分op，所以只要是支持的op，会跑在npu上 > 似乎没有在文档或者其他地方看见有关性能测试的数据 现阶段有针对op的测试，可以参考下我发在这个discussion里面的comment https://github.com/ggml-org/llama.cpp/discussions/8273#discussioncomment-13274821