nihui

background ref https://github.com/futz12/ncnn_llm/issues/17

feature plan

4

- purge onnx2ncnn (https://github.com/Tencent/ncnn/pull/6324) - arch: layer support any packing (https://github.com/Tencent/ncnn/pull/6392) - arch: optional packing in vulkan layer (https://github.com/Tencent/ncnn/pull/6389) - feat: pnnx input npy files - feat: vulkan pipeline cache...

https://docs.pytorch.org/docs/stable/export.html ```python import torch import torchvision.models as models net = models.resnet18() net.eval() torch.manual_seed(0) x = torch.rand(1, 3, 224, 224) a = net(x) # export pt2 program = torch.export.export(net, (x,)) torch.export.save(program,...

allow loading custom vulkan driver https://github.com/bylaws/libadrenotools

batch inference in partial model graph

3

During model inference, some operators/subgraphs require multiple inferences using batches. Here, a GRU is used to represent the batched portion Often, there are 'across-batch-axis' reshape operations before and after the...

【多人竞赛】support cpu count > 64

10

难度：困难 ncnn的cpu.h/cpu.cpp中有关 cpu 信息，topology 信息，大小核判断，绑定核心等逻辑实现未考虑超过64个cpu的情形，尤其是 cpuset class。修改相关实现以便支持 >64 cpu count 环境中能正常工作，应当形成测试，如使用qemu虚拟更多cpu 在 ncnn github 上提交 pull request，通过全部单元测试，合并pr 本issue为2025犀牛鸟开源人才培养活动专属issue TBA