// File a bug report with a reproducer against Tenstorrent repos (tt-lang, tt-metal, tt-xla)
| name | tt-bug-report |
| description | File a bug report with a reproducer against Tenstorrent repos (tt-lang, tt-metal, tt-xla) |
| argument-hint | <description of the problem> |
File a bug report for a Tenstorrent issue. Gather context, minimize the reproducer, and create a GitHub issue in the right repo.
| Component | Repo |
|---|---|
| TT-Lang (DSL, compiler, kernels) | tenstorrent/tt-lang |
| TTNN (ops, tensor APIs, device) | tenstorrent/tt-metal |
| TT-Metal (low-level C++ kernels) | tenstorrent/tt-metal |
| TT-XLA (PyTorch/JAX frontend) | tenstorrent/tt-xla |
If it's unclear which repo the bug belongs to, ask the user.
$ARGUMENTS
Determine the type of issue and gather relevant info:
For crashes or incorrect results:
For compiler errors (tt-lang):
/tmp/ttlang_initial.mlir, /tmp/ttlang_final.mlir)Ask the user if it's OK to upload the reproducer to the issue and spend some iterations trying to minimize it. If they agree, try to reduce it to the smallest script that still triggers the bug. Any reproducer is better than none -- if you can't minimize further, use what you have.
Use gh to create an issue in the appropriate repo:
gh issue create --repo <owner/repo> \
--title "[external bug report][from claude] <short title>" \
--body "$(cat <<'EOF'
<body>
EOF
)"
The body should include:
Return the issue URL to the user.
Write a ForgeModel-compatible loader for a HuggingFace model, validate it on CPU, and push the result to a branch on tenstorrent/tt-forge-models.
Install tt-forge, run the model loader from the cpu bringup branch on Tenstorrent hardware, iterate on failures, and open a PR to tenstorrent/tt-forge-models on success.
Set up and verify remote connection to Tenstorrent hardware. Provides tools for running kernels, copying files, and reading logs on remote devices.
TTNN trace capture and replay for eliminating dispatch overhead. Essential for real-time inference and multi-chip performance.
Profile and optimize TT-Lang kernels for performance. Covers auto-profiling, perf summary, signposts, and optimization workflow.
Comprehensive TT-Lang DSL reference including programming model, APIs, hardware constraints, and guides for translating CUDA, Triton, PyTorch, or TTNN kernels