Any TVM Automatic Quantization Examples?


#1

Hi,

I have been try to test a quantization example described in the following post:

https://tvm.ai/2019/04/29/opt-cuda-quantized.html

However, that machine that I am using for now has a GPU with a CUDA Compute Capability of 3.0, which is not enough for the dp4a instruction.

Is there any other example(s) about automatic quantization that does not use the dp4a instruction on a GPU. Or maybe examples targeting x86. My goal for now is getting started with TVM and its quantization pass.

I would appreciate if someone could provide me with some of this quantization code examples.

Thanks!


#2

it seems x86 quantization is not supported yet.


#3

x86 quantization is supported, though the accuracy and performance of models will vary

see merged PR https://github.com/dmlc/tvm/pull/2116
and open PR https://github.com/dmlc/tvm/pull/3294


#4

Is SSE supported by TVM besides AVX?