| commit | f62445cdd96a415d332585aa9702eaf1df3cf972 | [log] [tgz] |
|---|---|---|
| author | Wuwei Lin <wuwei@apache.org> | Sun Jul 28 13:57:09 2024 -0700 |
| committer | GitHub <noreply@github.com> | Sun Jul 28 13:57:09 2024 -0700 |
| tree | dd78adc057a7120538995354c0c3702391c0a903 | |
| parent | 4330c110550242571da017a1b15ae0b765723ae8 [diff] |
[Relax] Disable fusion for fetching from the packed params in FuseOps (#17198) * [Relax] Disable fusion for fetching from the packed params in FuseOps The order of bindings in the fusion result is determined by the first binding in each partition group. When the packed param tuple is used, the function usually begins with a numbers of `TupleGetItem` to unpack the param tuple. Previously `TupleGetItem` is treated as `kInjective`, this causes any operation that relies purely on these params to be moved to the beginning of the function and increases the memory usage of the intermediate results. * lint
Documentation | Contributors | Community | Release Notes
Apache TVM is a compiler stack for deep learning systems. It is designed to close the gap between the productivity-focused deep learning frameworks, and the performance- and efficiency-focused hardware backends. TVM works with deep learning frameworks to provide end to end compilation to different backends.
TVM is licensed under the Apache-2.0 license.
Check out the TVM Documentation site for installation instructions, tutorials, examples, and more. The Getting Started with TVM tutorial is a great place to start.
TVM adopts apache committer model, we aim to create an open source project that is maintained and owned by the community. Check out the Contributor Guide.
We learned a lot from the following projects when building TVM.