mirror of https://github.com/llvm/torch-mlir
60bf6c25af
We can plumb the linear matmul into pytorch using its quantized types with side channel information. To handle the final int8 operation we dequantize and requantize. |
||
|---|---|---|
| .. | ||
| CMakeLists.txt | ||
| DefaultDomainAtoF.cpp | ||
| DefaultDomainGtoP.cpp | ||
| DefaultDomainQtoZ.cpp | ||
| PassDetail.h | ||
| Passes.cpp | ||
| Patterns.cpp | ||
| TorchOnnxToTorch.cpp | ||
| Utils.cpp | ||