Releases: tenstorrent/tt-mlir
Releases · tenstorrent/tt-mlir
Nightly 0.2.0.dev20250725
Install via pip
pip install ttmlir==0.2.0.dev20250725 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-mlir/actions/runs/16510054737
- Status: ✅
🎯 Other Changes
- Golden Op executor for TTIR (#4041) by @ndrakulicTT
- Tell claude to source env for test commands too (#4183) by @nsmithtt
- Fixing bugs with lowering TTIR to LINALG and adding tests for them. (#4169) by @djstefanovicTT
- tt-kernel custom print format issue 3204 (#4175) by @mbagherbeikTT
- Migrated benchmark flag to ttrt run and added additional version skips. (#4152) by @tapspatel
- Use pytest instead of large CMake command to test Pykernel (#4161) by @vprajapati-tt
- D2M: Enable compute loop tiling, move subblocking logic lower in the pipeline. (#4070) by @jdesousa-TT
- D2M: Add affine loop coalescing optimization to ttmetal pipeline (#4191) by @jdesousa-TT
- [PyKernel] Enable TensorAccessor (#4116) by @vtangTT
- Swap the order of parsing the mesh shape (#4203) by @ctodTT
- Add ci_compile_dylib.py support for standalone exes (#4184) by @nsmithtt
- Add insertRuntimeTensorAndValidate API to preserve tensor version info (#4200) by @jnie-TT
Contributors
nsmithtt, tapspatel, and 8 other contributors
Assets 3
Nightly 0.2.0.dev20250724
Install via pip
pip install ttmlir==0.2.0.dev20250724 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-mlir/actions/runs/16486445310
- Status: ✅
🎯 Other Changes
- MLIR matching (#4010) by @ndrakulicTT
- Remove arg attrs from FuncOp when going to EmitC (#4144) by @svuckovicTT
- Use md book anchors to hoist deps from dockerfile #3755 (#4115) by @nsmithtt
- Updated PyKernel Docs w/ Multicore References (#4158) by @vprajapati-tt
- [Pykernel] Matmul SingleCore Demo (#4140) by @vprajapati-tt
- [Optimizer] Search over conv2d op configs (#3451) by @rpavlovicTT
- Add Commute Patterns for Concat and Slice (#4135) by @LPanosTT
- #4038: Refactored automatic shardy pass into stablehlo pipeline pass. (#4046) by @tapspatel
- Skip creation ops for trace capture, update conv verify logic, don't allocate empty slots for constants/params (#4165) by @jnie-TT
- [PyKernel] Matmul multi-core demo (#4172) by @vtangTT
Contributors
nsmithtt, rpavlovicTT, and 7 other contributors
Assets 3
Nightly 0.2.0.dev20250723
Install via pip
pip install ttmlir==0.2.0.dev20250723 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-mlir/actions/runs/16457455671
- Status: ✅
🎯 Other Changes
- Disable MLA option by default (#4136) by @rpavlovicTT
- Remove sharding workaround on in mode (#4122) by @azecevicTT
- Uplift model-explorer to a333e83c65fe8dc86f340c606c49ee262420b4ec 2025-07-22 (#4153) by @vmilosevic
- [TTIR, TTNN] Refactor ImplicitBroadcastFold pass (#3820) by @jserbedzijaTT
- [TTNNWorkaround] ConcatOp decomposition (#4142) by @mmanzoorTT
- expand reduce_window -> pooling support (#4064) by @anuragsingh-tt
- Run workarounds pass for full op when optimizer enabled (#4149) by @jnie-TT
- configure fabric for new CCL Ops (#3975) by @hkwonTT
- Add vtangTT to pykernel codeowners (#4163) by @vtangTT
Contributors
rpavlovicTT, vtangTT, and 7 other contributors
Assets 3
Nightly 0.2.0.dev20250722
Install via pip
pip install ttmlir==0.2.0.dev20250722 --pre --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-mlir/actions/runs/16429431180
- Status: ❌
🎯 Other Changes
- Uplift third_party/tt-metal to 81412069f02b01a483cc0fb29fec5c9bf28ac1ab 2025-07-18 (#4119) by @vmilosevic
- [OpModel] support to_memory_config op (#3722) by @gfengTT
- [explorer] Feature: add separate run data (#3664) by @madcampos
- Refactor trace end to end (#4085) by @jnie-TT
- Add ttir-builder support for ttnn-standalone cpp target (#4134) by @nsmithtt
- Invert flag (#4130) by @ctodTT
- #1869 Remove ARCH specific includes and use HAL APIs instead (#4132) by @brataTT
Contributors
madcampos, nsmithtt, and 5 other contributors
Assets 3
Nightly 0.2.0.dev20250721
Install via pip
pip install ttmlir==0.2.0.dev20250721 --pre --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-mlir/actions/runs/16376268052
- Status: ✅
- no changes
Nightly 0.2.0.dev20250720
Install via pip
pip install ttmlir==0.2.0.dev20250720 --pre --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-mlir/actions/runs/16376268052
- Status: ✅
- no changes
Nightly 0.2.0.dev20250719
Install via pip
pip install ttmlir==0.2.0.dev20250719 --pre --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-mlir/actions/runs/16376268052
- Status: ✅
🎯 Other Changes
- Implemented ttir relu to linalg (#4091) by @djstefanovicTT
- Fix L1 calc in ShardSolver & add progress tracker to MLA (#4088) by @rpavlovicTT
- Uplift model-explorer to 99da5ea9334119ad6bcaaf181b5c64c9504ccf82 2025-07-18 (#4087) by @vmilosevic
- Multicore Runtime Arguments in PyKernel + Multicore Example (#4048) by @vprajapati-tt
- Support arm build (#4117) by @nsmithtt
- extend default allocator L1 algorithm to DRAM memspace (#4118) by @vroubtsovTT
Contributors
nsmithtt, rpavlovicTT, and 4 other contributors
Assets 3
Nightly 0.2.0.dev20250718
Install via pip
pip install ttmlir==0.2.0.dev20250718 --pre --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-mlir/actions/runs/16359613714
- Status: ✅
🎯 Other Changes
- Add test for conv2d sharding (#3984) by @mvasiljevicTT
- Runtime API for getting and replacing tensors between operations (#3356) by @ndrakulicTT
- Add caching layer to constraint api (#4061) by @sgholamiTT
- [TTIR] Add flag to enable Conv2dWithMultiply pattern in fusing pass (#4090) by @jserbedzijaTT
- Fuse average pooling when divisor is a pool of a constant (#4060) by @LPanosTT
- Add CLAUDE.md (#4102) by @nsmithtt
- Support cpu fallback max reduce op (#4026) by @Mountagha
- D2M: Revert "Support blocking over dst (#3886)" (#4097) by @jdesousa-TT
- rm patchelf dependency (#4101) by @nsmithtt
- Uplift third_party/tt-metal to 7cd5738b35e8dc4862ef13a21468cb98264dbef6 2025-07-16 (#4086) by @vmilosevic
- Add runtime context to track runtime global state (#4110) by @jnie-TT
Contributors
Mountagha, nsmithtt, and 8 other contributors
Assets 3
Nightly 0.2.0.dev20250717
Install via pip
pip install ttmlir==0.2.0.dev20250717 --pre --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-mlir/actions/runs/16333857500
- Status: ✅
📦 Build
🎯 Other Changes
- Adding a unit test for legal layout analysis (#4044) by @bmalesevicTT
- Simplifying/canonicalizing affine map creation (#4054) by @sdjordjevicTT
- Run builder tests on s (blackholes) in CI (#3913) by @ctodTT
- D2M: ApplyHostMemrefCallingConvention pass (#4071) by @phizalev-TT
- Support ComputeConfigDescriptor::dst_full_sync_en #4095 (#4096) by @nsmithtt
Contributors
nsmithtt, sdjordjevicTT, and 4 other contributors
Assets 3
Nightly 0.2.0.dev20250716
Install via pip
pip install ttmlir==0.2.0.dev20250716 --pre --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Test status: ✅
Workflow run: https://github.com/tenstorrent/tt-mlir/actions/runs/16305254181
🎯 Other Changes
Contributors
madcampos, vmilosevic, and 2 other contributors