Bing Xu

For anyone who is interested, here is how to do it for basic enablement: 1. check dynamic codegen using a MIN strategy to create a default kernel instance 2. maybe...

FYI: v0.1.1 is released: https://github.com/facebookincubator/AITemplate/pull/74 New attention is more friendly to dynamic shape, and new runtime supports external memory allocators.

Could you directly modify this unittest to see if there anything abnormal? https://github.com/facebookincubator/AITemplate/blob/main/tests/unittest/ops/test_conv.py

This visualization tool is very helpful for investigating: https://facebookincubator.github.io/AITemplate/tutorial/how_to_visualize.html

Yes, check attributes especially op_type. I suspect frontend conv2d doesn’t map to correct ops. On Tue, Nov 1, 2022 at 17:31 Ehsan Azar ***@***.***> wrote: > The visualization is pretty...

Group (especially depth) conv is not fully supported in this release with CUDA backend, mainly due to depthwise conv is SIMT workload rather than TensorCore workload. In v0.2 will bring...

Also you can modify the CMake building system On Mon, Jun 5, 2017 at 09:42 Mu Li wrote: > I suggest adding a compilation flag USE_HIP. Then in makefile, add...

For LLAMA or other generative AI model needs, you may check out HippoML: https://blog.hippoml.com/large-language-model-inference-from-datacenter-to-edge-ed2f94da4a81 @drxmy @dhawalkp

> > For LLAMA or other generative AI model needs, you may check out HippoML: https://blog.hippoml.com/large-language-model-inference-from-datacenter-to-edge-ed2f94da4a81 > > @drxmy @dhawalkp > > Thank you! I just joined the waitlist. Is...

Done: https://github.com/facebookincubator/AITemplate/pull/74