[Feature] [Language] Tensor Sharding #21
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
add mesh_tensor functions: annotate_mesh_tensor_info, mesh_tensor_copy
For many NMP, memory is often distributed. To support distributed memory, we add a set of helper functions for working with mesh-distributed tensors. With this, we can let users to control the layout of tiles over distributed memory (cores) and add annotations for the compiling passes about the layout. Then, compiler can optimize the computing performance by utilizing the distributed properties of memory and the layout tiles.
* Update README.md * Update README.md * Configure docker * Update docker * Fix pre-commit --------- Co-authored-by: Jiaqi Guo <guojiaqi@sunmmio.com>
This was
linked to
issues
Jan 6, 2026
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
MeshTensorandMeshShardingPolicyin tilelang language v2 interface. MeshShardingPolicy denotes how to shard a multi-dimensional tensor on a mesh device. With the policy, MeshTensor computes the sharded shape for a tensor based on the given mesh device configuration. Also, to keep track of the original tensor shape, MeshTensor curates some meta data and attaches them to the func_attr when building prim_func. Note that the buffer created by MeshTensor is based on sharded shape, so we can directly get the buffer shape to get the sharded shape.testing/python/languageto perform unit-test.