Releases: tenstorrent/tt-xla
Nightly 0.6.0.dev20251126
Install via pip
pip install pjrt-plugin-tt==0.6.0.dev20251126 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Docker container
docker pull ghcr.io/tenstorrent/tt-xla-slim:0.6.0.dev20251126
More detailed instructions can be found in the Getting Started docker section
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-xla/actions/runs/19678299330
- Status: ✅
- Docker:
🎯 Other Changes
- Correct handling and detection of Out-of-Memory errors (#2302) by @meenakshiramanathan1
- Uplift third_party/tt-mlir to 83ad36534f9a92af70b5d58ac1803a5217e6b063 2025-11-25 (#2312) by @ajakovljevicTT
- Add Xfail Preset to Weekly Workflow (#2307) by @mmilosevicTT
- CopyFromBuffer Device2Device BufferInstance transfer workaround (#2277) by @jameszianxuTT
Nightly 0.6.0.dev20251125
Install via pip
pip install pjrt-plugin-tt==0.6.0.dev20251125 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Docker container
docker pull ghcr.io/tenstorrent/tt-xla-slim:0.6.0.dev20251125
More detailed instructions can be found in the Getting Started docker section
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-xla/actions/runs/19653316614
- Status: ✅
- Docker:
🎯 Other Changes
- Uplift third_party/tt-mlir to cbbed3dec7d0968924fa0548ce754a8af8c05bcb 2025-11-24 (#2286) by @ajakovljevicTT
- Adds Test Config for Attention DenseUnet Model (#2295) by @saiarthiraguram
- [CI] Build docker stacked on top of tt-mlir docker (#2274) by @nsumrakTT
- Fix Training Tests for Nightly (#2297) by @mmilosevicTT
- Add virtual to releaseResources in LoadedExecutableInstance (#2298) by @sgligorijevicTT
- Moving a torch test to large models (#2300) by @ajakovljevicTT
- Update Error Messages for Failing Training Tests (#2301) by @mmilosevicTT
- Run extended tests on PR with mlir version change (#2299) by @vmilosevic
- Bringup LLVM FileCheck infra to assert that fusion patterns still work (#2230) by @jameszianxuTT
- Add test configs for yolov9 pytorch model (#2289) by @kamalrajkannan78
- [vLLM plugin] Add test for Qwen3-Embedding-0.6B model (#2284) by @mmanzoorTT
Nightly 0.6.0.dev20251124
Install via pip
pip install pjrt-plugin-tt==0.6.0.dev20251124 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Docker container
docker pull ghcr.io/tenstorrent/tt-xla-slim:0.6.0.dev20251124
More detailed instructions can be found in the Getting Started docker section
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-xla/actions/runs/19618666774
- Status: ✅
- Docker:
🎯 Other Changes
Nightly 0.6.0.dev20251123
Install via pip
pip install pjrt-plugin-tt==0.6.0.dev20251123 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Docker container
docker pull ghcr.io/tenstorrent/tt-xla-slim:0.6.0.dev20251123
More detailed instructions can be found in the Getting Started docker section
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-xla/actions/runs/19600129100
- Status: ❌
- Docker:
🎯 Other Changes
- Uplift third_party/tt_forge_models to e3fc28d1c34e51969e82b8780438f326541e2785 2025-11-21 (#2268) by @ajakovljevicTT
- Pass the action output to inspect changes job by @vmilosevic
- Add Experimental passing models to main (#2269) by @ctr-pmuruganTT
- Fix extended test runs (#2271) by @ajakovljevicTT
- Cleanup compile config and add performance user docs (#1936) by @odjuricicTT
- [bugfix] Closing device on tt::runtime::submit failure (#2108) by @acolicTT
- Torch graph test fixes and improvements (#2264) by @gengelageTT
- Fix 3 tt-xla main breaks on Nov21 affecting onPR / onPush (#2278) by @kmabeeTT
- Split training tests to nightly and weekly (#2273) by @ndrakulicTT
Nightly 0.6.0.dev20251121
Install via pip
pip install pjrt-plugin-tt==0.6.0.dev20251121 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Docker container
docker pull ghcr.io/tenstorrent/tt-xla-slim:0.6.0.dev20251121
More detailed instructions can be found in the Getting Started docker section
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-xla/actions/runs/19544338401
- Status: ✅
- Docker:
🎯 Other Changes
- Uplift third_party/tt-mlir to d8f8aa475d180d012f9cc17f9501a30a52a840be 2025-11-20 (#2244) by @ajakovljevicTT
- Add test config for unet_480x640 variant and change the failure reason for phi3.5-vision (#2245) by @kamalrajkannan78
- Record otel data from workflows (#2248) by @vmilosevic
- Add newline by @vmilosevic
- Add mapping for the new n300-high-memory label (#2246) by @vvukomanTT
- Run model tests on uplift PR (#2222) by @vmilosevic
- Xfail crashed tests in nightly (#2255) by @ctr-pmuruganTT
Nightly 0.6.0.dev20251120
Install via pip
pip install pjrt-plugin-tt==0.6.0.dev20251120 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Docker container
docker pull ghcr.io/tenstorrent/tt-xla-slim:0.6.0.dev20251120
More detailed instructions can be found in the Getting Started docker section
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-xla/actions/runs/19519178724
- Status: ✅
- Docker:
🎯 Other Changes
- Uplift third_party/tt_forge_models to 3feef86b842abc8712f514b8aa24d8ceb48769ae 2025-11-19 (#2232) by @ajakovljevicTT
- Uplift third_party/tt-mlir to 81c665e199ee9eda3ff336217ad50110766364fe 2025-11-19 (#2231) by @ajakovljevicTT
- Fixing example reference in emitpy codegen tutorial (#2235) by @sdjordjevicTT
- [vLLM plugin] Add multi device support for pooling runner (#2195) by @mmanzoorTT
- Uplift torch_xla to a5be1f8 (Remove aggressive assert in clear_computation_cache) (#2226) by @jameszianxuTT
- Change file caching env variables behaviour in CI (#2236) by @vvukomanTT
- Model and . Alter vLLM plugin to utilize paged attention. (#2215) by @LPanosTT
- Update nightly failing models on 19_11 (#2239) by @ctr-pmuruganTT
- Uplift third_party/tt_forge_models to 61571741c57a1719859768600258236c9f5d9d43 2025-11-19 (#2240) by @ajakovljevicTT
- Update JAX Training Tests Error Messages (#2233) by @mmilosevicTT
- Add runtime multihost infra, add multihost llama test (#1918) by @jnie-TT
- Re-Enable PCC Check for qwen_3/causal_lm models after padding fix in tt-forge-models (#2157) by @sonalibaskaran2499
Rc 0.6.0rc1
Install via pip
pip install pjrt-plugin-tt==0.6.0rc1 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Docker container
docker pull ghcr.io/tenstorrent/tt-xla-slim:0.6.0rc1
More detailed instructions can be found in the Getting Started docker section
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-xla/actions/runs/19449086178
- Status: ❌
- Docker:
🛠 Fixes
🎯 Other Changes
- Add test config for resnet50 for target input size (1280x800) (#1583) by @kamalrajkannan78
- Remove assert_pcc=False from yolov10 variants as it now have pcc>0.99 after disabling endtoend mode (#1694) by @kamalrajkannan78
- Add entrypoint script to ird docker image (#1666) by @vmilosevic
- Update JAX models for whisper and wav2vec2 (#1333) by @ctr-pmuruganTT
- Update config from Experimental models for OOM issues (#1723) by @ctr-pmuruganTT
- Add unit test for topk op to keep track of pcc drop (#1702) by @kamalrajkannan78
- Injecting FX metadata via TorchDispatchMode (#1699) by @vkovinicTT
- Tensor dumping (#1612) by @sgligorijevicTT
- Update test durations based on latest nightly 2025-10-16 (#1710) by @ajakovljevicTT
- Fix Build tt-xla job in CI (#1733) by @vvukomanTT
- Add a Tensor Parallel JAX Gemma3-27b Model to TT-xla Model Demos (#675) by @lanchongyizu
- Uplift tt_forge_model 2025.10.20 (#1731) by @ndrakulicTT
- Tighten PCC checking in handful of generality models and report on PCC details in INCORRECT_RESULT tests for superset (#1727) by @kmabeeTT
- Fix pre-commit (#1736) by @LPanosTT
- Convert test_config*.py to test_config*.yaml to support scripted/automated updates (#1729) by @kmabeeTT
- Add model auto discovery (test_models.py) initial documentation as model_auto_discovery_tests.md (#1724) by @kmabeeTT
- Few test_config fixes for duplicate red models (#1739) by @kmabeeTT
- Add support to redirect large tests to civ2 shared-runners for test_models.py (#1576) by @kmabeeTT
- Build docker image on push to main (#1732) by @vmilosevic
- adding sdxl unet test to test infra (#1679) by @ppadjinTT
- Fix single test dispatch (#1740) by @vvukomanTT
- modify CI + uplift mlir (#1669) by @sgligorijevicTT
- Fix unmarshallable object error in nightly tests under --forked for tests that set required_pcc, exposed by yaml changes (#1744) by @kmabeeTT
- Run nightly torch multidevice tests without forked (#1748) by @jameszianxuTT
- Uplift third_party/tt-mlir to ebe1f8dd008f5417b5a726bf712804c0367b6977 2025-10-22 (#1631) by @ajakovljevicTT
- Fix IR dumping (#1746) by @sgligorijevicTT
- Adjust issues with model training JAX tests (#1720) by @mmilosevicTT
- Add Pointpillars config and update the bringup status (#1592) by @ashokkumarkannan1
- Forward pipeline options to alchemist (#1752) by @sgligorijevicTT
- Add BERT MHA create heads test (#1747) by @ddilbazTT
- Revert test_config_single_device_inference python file and update pointpillars status (#1753) by @ashokkumarkannan1
- xfail falcon/pytorch-tiiuae/falcon-7b-instruct-tensor_parallel-full-inference DRAM oom test from nightly (#1756) by @ajakovljevicTT
- Add xfail reason for topk failed cases (#1764) by @ctr-pmuruganTT
- Revise CODEOWNERS for improved ownership clarity (#1734) by @mrakitaTT
- [nightly bug] Updated test config for 9 nightly tests (xfail, lower pcc) due to torch-xla changes (#1763) by @vkovinicTT
- Edit vLLM plugin code to support TT static cache (#1377) by @LPanosTT
- Fix debug mode in flatbuffer_loaded_executable_instance (#1758) by @ctr-pmuruganTT
- Ingest TTNN via alchemist (#1762) by @sgligorijevicTT
- Uplift third_party/tt-mlir to cd2498cc1d915659242942984b27e599451d3058 2025-10-23 (#1769) by @ajakovljevicTT
- uplift of the torch-xla (#1777) by @vkovinicTT
- update torch tests for runtime binary op issue (#1786) by @ctr-pmuruganTT
- Xfailing nightly model tests that fail on matmul shape errors (#1782) by @ajakovljevicTT
- Add CLAUDE.md file for Claude Code development guidance (#1467) by @odjuricicTT
- Rename output files for dumping inputs (#1775) by @sgligorijevicTT
- Update placeholders models status (#1780) by @ashokkumarkannan1
- PJRT implementation refactor (#1394) by @sdjukicTT
- Dropping PCC for nightly CI (#1790) by @ajakovljevicTT
- Multichip graph tests (#1745) by @AleksKnezevic
- Use default attention mask for single input per batch inference. (#1785) by @mmanzoorTT
- Make tt-explorer available from tt-xla (#1600) by @brataTT
- Add element-wise scatter tests (#1749) by @ddilbazTT
- Fix docs build in main (#1803) by @brataTT
- Uplift third_party/tt_forge_models to 8868ee46118cce901cd425c4889e2fb5c811aa12 2025-10-25 (#1806) by @ajakovljevicTT
- Uplift third_party/tt_forge_models to 6fb6d2ce17b2a0e7036d458d1702c141317748b2 2025-10-25 (#1807) b...
Nightly 0.6.0.dev20251119
Install via pip
pip install pjrt-plugin-tt==0.6.0.dev20251119 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Docker container
docker pull ghcr.io/tenstorrent/tt-xla-slim:0.6.0.dev20251119
More detailed instructions can be found in the Getting Started docker section
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-xla/actions/runs/19468580240
- Status: ✅
- Docker:
🎯 Other Changes
- Uplift third_party/tt_forge_models to 342a516af9dbd2f1747fe9e71a7011cc677ed019 2025-11-18 (#2220) by @ajakovljevicTT
- Uplift third_party/tt-mlir to d0ea5f1233206bd15608392aef4ba69c243b95fc 2025-11-18 (#2219) by @ajakovljevicTT
- Fix yolox nightly issue (#1974) by @chandrasekaranpradeep
- Update JAX Error Messages in Training Tests (#2106) by @mmilosevicTT
- Uncomment YOLOv6 and YOLOv7 test configs as their ultralytics dependency is now removed (#2221) by @kamalrajkannan78
Nightly 0.6.0.dev20251118
Install via pip
pip install pjrt-plugin-tt==0.6.0.dev20251118 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Docker container
docker pull ghcr.io/tenstorrent/tt-xla-slim:0.6.0.dev20251118
More detailed instructions can be found in the Getting Started docker section
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-xla/actions/runs/19436286267
- Status: ❌
- Docker:
🎯 Other Changes
- Uplift third_party/tt-mlir to 938ea4040e1518279ca9fffd830472a66d19c32a 2025-11-17 (#2199) by @ajakovljevicTT
- [CI] Unfurl links for slack notifications (#2200) by @nsumrakTT
- Only produce release vllm plugin (#2203) by @vmilosevic
- Lazily call toHost for BufferInstances (#1657) by @jameszianxuTT
- Add inference test config for efficientdet model (#2207) by @kamalrajkannan78
- Fixing nightly 17-11-2025 (#2205) by @ajakovljevicTT
- Updating training tests in (#2211) by @ndrakulicTT
- Refactoring jax traning tests to go through automatic test discovery infra (#2180) by @ajakovljevicTT
Nightly 0.6.0.dev20251117
Install via pip
pip install pjrt-plugin-tt==0.6.0.dev20251117 --extra-index-url https://pypi.eng.aws.tenstorrent.com/
Docker container
docker pull ghcr.io/tenstorrent/tt-xla-slim:0.6.0.dev20251117
More detailed instructions can be found in the Getting Started docker section
Tests:
- Workflow:
- Run link: https://github.com/tenstorrent/tt-xla/actions/runs/19395494766
- Status: ✅
- Docker:
- no changes