tvm
tvm copied to clipboard
apache
Open deep learning compiler stack for cpu, gpu and specialized accelerators
This PR introduce a new LLVM ORC JIT executor for the runtime. The new ORCJIT may obsolete MCJIT due to better upstream maintenance. --- #### Changes * Old MCJIT is...
This commit adds fp16 test cases to the conv2d NHWC TOPI schedules for `arm_cpu`. Following the example of #8529, the numpy reference conv2d output is computed in fp32 instead of...
When reduction axis is small, it's not necessary to use rfactor. This PR updates the gemv rule to use rfactor only when the reduction axis is large enough.
This PR aims to support backend dispatching for `multinomial_from_uniform`, which includes: - Relax Op `multinomial_from_uniform` - TIR gpu kernel for `multinomial_from_uniform` - dispatching pass - TVMScript parser support for pure-python...
Support Reverse sequence quantization operation as part of #15148
Below script can be used to reproduce the issue. You may need to run it multiple times to reproduce, because sample_perfect_tile may sometime to hide the issue with some decision....
This commit adds a scalable `arm_cpu` conv2d NHWC schedule for fp32 which generates SME instructions by using the tensor intrinsics introduced in #16921. Alongside the SME schedule, the logic of...
In https://github.com/apache/tvm/issues/9187 we implemented quantised version of operators in TFLite frontend. Recently, I just noticed a few more operators (with varying priorities) that can be taken as beginner friendly tasks,...
Support ARG_MIN quantization operation as part of #15148
When analyzing scalable expressions, the analyzer will iterate over a series of known vscale values in the range 1-16. However, we can tighten this range to only values that are...
