Split trtllm-gen cubin pins per arch by Vinnie6167 · Pull Request #4191 · flashinfer-ai/flashinfer

Vinnie6167 · 2026-07-28T00:54:25Z

📌 Description

The BMM & GEMM trtllm-gen cubins in ToT fail on SM 100. This MR introduces a WAR, loading an older set of published cubins when targeting SM 100.

🔍 Related Issues

🚀 Pull Request Checklist

Thank you for contributing to FlashInfer! Before we review your pull request, please make sure the following items are complete.

✅ Pre-commit Checks

I have installed pre-commit by running pip install pre-commit (or used your preferred method).
I have installed the hooks with pre-commit install.
I have run the hooks manually with pre-commit run --all-files and fixed any reported issues.

If you are unsure about how to set up pre-commit, see the pre-commit documentation.

🧪 Tests

Tests have been added or updated as needed.
All tests are passing (unittest, etc.).

Reviewer Notes

gemini-code-assist · 2026-07-28T00:54:29Z

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

coderabbitai · 2026-07-28T00:54:35Z

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: efa8fee7-ca36-45f1-adb3-bebcf81e26a6

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

kahyunnam

In AOT, gen_all_modules() (aot.py) emits both when has_sm100 and has_sm107 (default arch list has 10.0a and 10.7a): each gen_* rewrites the same synlink, then all modules compile afterward. Last writer wins → Blackwell modules (enable_rubin=False, no -DTLLM_RUBIN_FEATURES) can compile against Rubin export headers → same GemmOptions ABI skew this PR is fixing.

Same issue in gen_trtllm_gen_gemm_module, gen_trtllm_low_latency_gemm_module, and gen_trtllm_gen_fused_moe_sm100_module

@kahyunnam

The Blackwell and Rubin variants shared one export-header symlink under FLASHINFER_GEN_SRC_DIR (independent of enable_rubin). In an AOT build all modules are generated before anything compiles, so the last gen_* call's ensure_symlink won, and the other variant could compile against the wrong (flag-mismatched) export headers -- the same GemmOptions/metadata ABI skew this PR fixes. Runtime JIT was unaffected (one variant per device). Give each module its own export-header root (FLASHINFER_GEN_SRC_DIR/ trtllm_export/<module_name>) and put it first on the include path so the fixed `#include "flashinfer/trtllm/.../trtllmGen_*_export/..."` resolves per-variant. Also de-conflicts from moe_utils, which shares that path. Addresses review feedback from @kahyunnam on PR flashinfer-ai#4191. AI-assisted (Claude Code).

gemini-code-assist · 2026-07-28T01:57:32Z

Caution

The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased.

flashinfer-bot · 2026-07-28T02:04:51Z

GitLab MR !1058 has been created, and the CI pipeline #59880138 is currently running. I'll report back once the pipeline job completes.

Renaming get_trtllm_gemm_module to the typed _get_trtllm_gemm_module_impl made mypy check its body, surfacing a pre-existing setup_cubin_loader(Path) call that expects str. Match the sibling MoE / low-latency-gemm loaders, which already wrap it in str(). No runtime change. AI-assisted (Claude Code).

flashinfer-bot · 2026-07-28T02:07:00Z

[CANCELING] Pipeline #59880138: canceled

Vinnie6167 · 2026-07-28T02:13:12Z

/bot run

flashinfer-bot · 2026-07-28T02:14:17Z

GitLab MR !1058 has been updated with latest changes, and the CI pipeline #59881205 is currently running. I'll report back once the pipeline job completes.

test_get_subdir_file_list was already failing on release-v0.6.16 before this branch: flashinfer-ai#4191 added TRTLLM_GEN_BMM_RUBIN and TRTLLM_GEN_GEMM_RUBIN to cubin_dirs, but the test never registered mocks for them. get_checksums() therefore tried to fetch two unmocked checksums.txt URLs, exhausted the retry budget (~50s of backoff) and died with FileNotFoundError. Verified failing at the unmodified base commit c46207c. Register directory-index and checksums.txt mocks for both Rubin pins, reusing the non-Rubin index bodies so the mocked pins share kernel filenames exactly as the published ones do, with every hash differing. Update the meta-info header count 3 -> 5 (FMHA, GEMM, BMM, GEMM_RUBIN, BMM_RUBIN). Add regression coverage for the checksum collision fixed in the previous commit: assert paths are unique and that a filename present in both a plain and a Rubin pin resolves to its own pin's hash. Against the unfixed get_checksums() this fails with AssertionError: Bmm_..._schedS_bN_clmp_dynBatch_sm100f.cubin resolved to the same checksum for both pins (bbbb111122223333) -- the per-pin hashes collided tests/test_artifacts.py: 1 failed, 3 passed -> 4 passed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

…ild break (#4200)  ## 📌 Description Fixes both failing jobs in the 0.6.16rc2 release run ([30330417436](https://github.com/flashinfer-ai/flashinfer/actions/runs/30330417436)) on `c46207c2`. Both are deterministic consequences of introducing a second, Rubin-specific cubin pin (`TRTLLM_GEN_BMM_RUBIN` / `TRTLLM_GEN_GEMM_RUBIN`) alongside the existing one — neither is flaky. **1. `build-flashinfer-cubin` — `Failed to download cubins: checksum mismatch`** `get_checksums()` flattened every pin's `checksums.txt` into one dict keyed by **bare filename**, namespacing only `.h` files. The plain and Rubin pins ship the **same** sm100f/sm103a kernel names built from different sources, so the pin processed last silently overwrote the earlier one's hashes. `get_subdir_file_list()` then looked cubins up by bare name while headers already used the full path. Measured against the live published manifests: | pin pair | shared filenames | same hash | **conflicting** | |---|---|---|---| | `TRTLLM_GEN_BMM` vs `_RUBIN` | 2534 | 0 | **2534** | | `TRTLLM_GEN_GEMM` vs `_RUBIN` | 108 | 3 | **105** | The Rubin pins come later in `cubin_dirs`, so they win and 2640 of 8055 files are verified against the wrong pin's hash. Replaying both keying schemes over the real manifests: ``` OLD (bare filename) entries= 5412 files=8055 MISMATCHED=2640 NEW (full path) entries= 8055 files=8055 MISMATCHED=0 ``` **Fix:** key every entry by full path, and look cubins up the same way headers already were. This is **not** a stale pinned hash — all 12 `checksums.txt` manifests still match their `CheckSumHash` SHA256 exactly. It is also **not** download flakiness: the log shows two transient `403 Forbidden`s on `fmhaSm100fKernel_QkvE4m3OE2m1H256...Q{8,32}Kv128StaticSwapsAbForGen.cubin`, but both recovered on retry (no `Max retries reached`) and both URLs serve `200` now. A checksum-retry loop would have re-downloaded 2640 files, failed identically, and turned a fast failure into a slow one. **2. `build-flashinfer-jit-cache` (all 6 matrix entries) — compile error** ``` csrc/trtllm_batched_gemm_runner.cu(137): error: enum "batchedGemm::trtllm::gen::CudaArch" has no member "Sm107a" ``` The `Sm107a` comparison added in #4122 is unguarded, but that enumerator exists only in the Rubin pin's generated header — the default pin's `flashinferMetaInfo.h` has `Sm100f`/`Sm103a` and **zero** occurrences of `Sm107a` (the Rubin pin has 2552). So the non-Rubin module `fused_moe_trtllm_sm100` cannot compile. This is why `main`, where #4122 was reverted, is unaffected. **Fix:** guard with `TLLM_RUBIN_FEATURES`, which `gen_trtllm_gen_fused_moe_sm100_module` already defines (`rubin_flags`) exactly when the Rubin pin is selected — so the guard is by construction in sync with the enum's availability. `sm_version == 107` is unreachable in the non-Rubin module regardless: `enable_rubin` is chosen from the same device compute capability that `getSMVersion()` reports. **3. `tests/test_artifacts.py::test_get_subdir_file_list` — already broken on the release branch** Independent of the above, #4191 added the two Rubin pins to `cubin_dirs` without registering mocks for them, so the test exhausted its retry budget (~50 s of backoff) and died with `FileNotFoundError`. Verified failing at the unmodified base `c46207c2`. Mocks added, meta-info header count corrected 3 → 5, plus regression coverage for the collision above. ## 🔍 Related Issues - Release run: https://github.com/flashinfer-ai/flashinfer/actions/runs/30330417436 - Introduced by the per-arch pin split (#4191) interacting with SM107 support (#4122) ## 🚀 Pull Request Checklist Thank you for contributing to FlashInfer! Before we review your pull request, please make sure the following items are complete. ### ✅ Pre-commit Checks - [x] I have installed `pre-commit` by running `pip install pre-commit` (or used your preferred method). - [x] I have installed the hooks with `pre-commit install`. - [x] I have run the hooks manually with `pre-commit run --all-files` and fixed any reported issues. > If you are unsure about how to set up `pre-commit`, see [the pre-commit documentation](https://pre-commit.com/). ## 🧪 Tests - [x] Tests have been added or updated as needed. - [x] All tests are passing (`unittest`, etc.). `tests/test_artifacts.py`: **1 failed, 3 passed → 4 passed**. The new regression assertion is verified to actually catch the bug — reverting only `artifacts.py` to the buggy base while keeping the new test yields: ``` AssertionError: Bmm_..._schedS_bN_clmp_dynBatch_sm100f.cubin resolved to the same checksum for both pins (bbbb111122223333) -- the per-pin hashes collided ``` ## Reviewer Notes - **Scope of local testing.** `tests/test_artifacts.py` is fully exercised locally (it is `responses`-mocked, no GPU needed). The two build jobs themselves are not reproducible locally and need CI to confirm. - **Why the jit-cache fix is sufficient for the whole matrix.** All 4 failed jit-cache jobs report **exactly one** compiler error each — the same `Sm107a` one, same target `csrc_trtllm_batched_gemm_runner.cuda.o`. The matrix builds `7.5 8.0 8.9 9.0a 10.0a 10.3a 12.0f` (no 10.7), so `has_sm107` is false and the Rubin modules are not compiled there — this change unblocks the job without pulling in any `sm_107a` codegen. - **Two alternatives considered and rejected** for the `Sm107a` guard: - A new `FLASHINFER_HAS_SM107A` macro — nothing would define it, so the `#else` branch would always compile, silently disabling Rubin kernel selection. - Falling back to `Sm103a`/`Sm100f` under `sm_version == 107` — worse than the build error: on real Rubin silicon it would select wrong-arch cubins instead of failing loudly. - **Follow-up worth considering (not in this PR):** the `checksum mismatch` error does not name the offending file. Including the path plus expected/actual hashes would have made this diagnosable from the log alone. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

…4225) ## Summary Fixes two test regressions introduced by #4191 (release-only, not on `main`): - **#4208** — `test_make_tuning_config_reuses_topk_ids_initializer` called `get_trtllm_moe_sm100_module.cache_clear()`, but #4191 moved `@functools.cache` to `_get_trtllm_moe_sm100_module_impl`. - **#4166 (re-open)** — #4187 updated the symlink test to assert under `GEN_SRC_DIR/flashinfer/...`, but #4191 moved the export symlink to `GEN_SRC_DIR/trtllm_export/fused_moe_trtllm_sm100/flashinfer/...`. Both failures appeared on all Blackwell jobs in Test Cycle 2 (pipeline 59912149, rc3). ## Test plan - [x] `pytest tests/autotuner/test_autotuner_core.py::test_make_tuning_config_reuses_topk_ids_initializer` — pass - [x] `pytest tests/utils/test_gen_module_symlink_race_condition.py` — pass - [ ] B300/GB200 CI rerun to confirm cycle-2 green on these two files

…(release port for #4107) (#4230) ## Description Release-specific port of #4177 onto `release-v0.6.16` for #4107. On SM12x (Spark, RTX Pro 6000), TRTLLM routed-MoE backends were incorrectly claiming support (`arch >= 100`) and then dispatching sm100f/sm103a cubins, causing `RuntimeError: Error occurred when running GEMM!` or segfaults in `test_split_fused_moe_kernel_vs_reference`. ## Changes - **`csrc/trtllm_batched_gemm_runner.cu`**: Replace per-SM if-chains with `isArchCompatible()`; reject unknown cubin families; guard `Sm107a` behind `#ifdef TLLM_RUBIN_FEATURES` (only exists in the Rubin cubin pin's headers). - **`csrc/trtllm_gemm_runner.cu`**: Same arch filter for the plain GEMM runner (previously had no arch filtering at all). - **`flashinfer/fused_moe/api.py`**: Tighten `Trtllm*Config.supported()` from `arch >= 100` to explicit allowlists `_TRTLLM_ROUTED_ARCHS = (100, 103, 107)` and `_TRTLLM_ROUTED_FP8_ARCHS = (100, 103)`. - **`tests/moe_ep/test_split_fused_moe_kernel_vs_reference.py`**: Gate GPU tests on `config_cls.supported(arch)`; add CPU contract tests + SM120 regression guard. ## Release-specific notes - Includes **sm107** in the allowlist (unlike the rebased #4177 for `main`, which will drop 107 after the #4122 revert). - `Sm107a` enum case is gated behind `TLLM_RUBIN_FEATURES`, matching the existing release pattern from #4122/#4191. Verified against the actual pinned headers: the default BMM/GEMM pins do not define `Sm107a`; only the Rubin pins do. ## Verification - CPU tests: 36 passed, 2 skipped (`test_split_fused_moe_kernel_vs_reference.py`) - Backend claims: sm120/sm121 now fall back to Cutlass only (no Trtllm* backends) - Compile: `isArchCompatible()` builds cleanly against both default and Rubin BMM export headers; ported batched runner compiles against default pin ## Related - Fixes #4107 - Upstream PR: #4177 (targets `main`, has merge conflicts) - Cherry-pick source commit on `main`: not yet merged ## Pre-existing issue (not in scope) `#4213` on release references `options.mDtypeSfC`, which does not exist in the default BMM cubin pin's headers (only in the Rubin pin). This is a separate release-only compile issue on the non-Rubin module, predating this port.

…ackages carry mDtypeSfC (#4235) ## 📌 Description Fixes two coupled problems on `release-v0.6.16`: **1. The branch tip cannot compile the trtllm-gen fused-MoE module — on any architecture.** #4213 checks `BatchedGemmOptions::mDtypeSfC`, but that field does not exist in either pinned cubin package's exported headers (Blackwell `b368d003`, Rubin `46d3f356` — verified 0 occurrences in both published `BatchedGemmOptions.h`, and in the internal `81a53cff` package as well). Any JIT or AOT rebuild of `fused_moe_trtllm_*` fails with: ``` trtllm_batched_gemm_runner.cu(128): error: class "BatchedGemmOptions" has no member "mDtypeSfC" ``` (The original PR #4168 shows the same failure in its CI on main.) **2. The FP4-MoE illegal-memory-access #4213 targets (#4164 / NVBug 6517914) is still live on Rubin.** The offending `bmm_E2m1xFp32_*` kernels emit linear FP32 output scale-factors into the E4M3-sized (1 B/block) buffer the MoE pipeline allocates → device-side overflow → async IMA, near-deterministic under autotune. Verified against the published catalogs: **132 such kernels, all in the sm_107a package (`46d3f356`); zero in the Blackwell package** — which is why the per-arch pin split (#4191) healed Blackwell while Rubin still crashes. ## What changed Replace the un-compilable typed check with a kernel-**name** filter carrying the same intent: reject configs whose `mFunctionName` contains the `E2m1xFp32` token. This compiles against the *currently published* packages and removes the bad candidates before autotune can run them. #4213's other changes (per-token-scaling restructure, launcher tactic-enumeration alignment) are inherited unchanged. **Scope note:** the name filter is deliberately narrower than the typed check — it rejects only the known-bad family. Against the published catalogs this is behaviorally equivalent (the `E2m1xFp32` kernels are the only offenders present). The TODO in the code ties restoring the typed `mDtypeSfC` check to publishing + pinning packages that carry the field. ## Validation - Build break reproduced (sbsa AOT wheel build, CTK 13.4) at the branch tip; gone with this patch. - VR200 (Rubin) re-run of the affected MoE test files in progress; results will be posted here before marking ready for review. Refs: #4164, #4213, #4168 · NVBug 6517914 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Vinnie6167 <Vinnie6167@users.noreply.github.com>

Select the Rubin (sm_107 / cc 10.7) deep-gemm artifact directory at runtime, mirroring the enable_rubin split added for the trtllm-gen GEMM/BMM/MoE modules in PR flashinfer-ai#4191. - Add is_rubin_arch() / get_deepgemm_artifact_path() helpers and use them in load_all(), load(), and KernelMap.init_indices() so cubins and kernel_map.json are fetched from DEEPGEMM_RUBIN on Rubin. - Add KernelMap.KERNEL_MAP_HASH_RUBIN for the Rubin kernel_map.json. - Register DEEPGEMM_RUBIN in get_subdir_file_list()'s download list. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

## Summary Follow-up to #4258 (which added `ArtifactPath.DEEPGEMM_RUBIN` and its checksum). That PR merged the path + checksum only; this PR adds the arch-based split that actually routes deep-gemm cubin loading to the Rubin directory on sm_107, mirroring the `enable_rubin` pattern from #4191. ### `flashinfer/deep_gemm.py` - New `is_rubin_arch()` / `get_deepgemm_artifact_path()` helpers selecting `DEEPGEMM_RUBIN` on arch `107a` and `DEEPGEMM` otherwise. - Used in `load_all()`, `load()`, and `KernelMap.init_indices()` so both the kernel cubins and `kernel_map.json` are fetched from the correct per-arch directory. - Added `KernelMap.KERNEL_MAP_HASH_RUBIN = f8bf2b1b…a36e27` for the Rubin `kernel_map.json` (a separate manifest from the default one), selected by arch. ### `flashinfer/artifacts.py` - Added `DEEPGEMM_RUBIN` to `get_subdir_file_list()`'s download list so `flashinfer artifacts download` fetches it, and updated the DEEPGEMM comment to reference the Rubin map hash. Both Rubin hashes were verified by downloading the artifacts directly from the cubin repository. Targets the `release-v0.6.16` release branch. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

## 📌 Description This PR relands SM 107 support to main branch (reverted in #4171) as well as some other release fixes. #### Cherry Picks - #4191 - #4189 - #4200 - #4215 - #4225 - #4230 - #4235 - #4226 - #4257 - #4258 - #4261 #### Other Changes - Rubin guards from #4252's conflict resolution (`TLLM_RUBIN_FEATURES`: SiTuGlu static_asserts + tile-192 advertisement, compiled out for the Rubin BMM pin) - Test-contract update: `test_unified_moe.py` arch assertions written post-revert (#4159) flipped to the restored contract (FP4/BF16 claim 107; FP8 stays 100/103)  ## 🔍 Related Issues  #4107, #4164, reverts #4171 ## 🚀 Pull Request Checklist Thank you for contributing to FlashInfer! Before we review your pull request, please make sure the following items are complete. ### ✅ Pre-commit Checks - [x] I have installed `pre-commit` by running `pip install pre-commit` (or used your preferred method). - [x] I have installed the hooks with `pre-commit install`. - [x] I have run the hooks manually with `pre-commit run --all-files` and fixed any reported issues. > If you are unsure about how to set up `pre-commit`, see [the pre-commit documentation](https://pre-commit.com/). ## 🧪 Tests - [ ] Tests have been added or updated as needed. - [ ] All tests are passing (`unittest`, etc.). ## Reviewer Notes   ## Summary by CodeRabbit * **New Features** * Added support for Rubin/SM107 GPUs across GEMM, MoE, attention, quantization, sampling, and DeepGEMM workflows. * Added architecture-aware kernel selection, memory sizing, compilation, and artifact handling. * **Bug Fixes** * Improved validation and error messages for incompatible GPU architectures and invalid kernel configurations. * Clearly rejects unsupported NVFP4 KV-cache operations on SM107. * **Documentation** * Updated installation guidance with the SM107 architecture target. * **Tests** * Expanded architecture coverage and compatibility checks across GPU test suites.  --------- Co-authored-by: Vinnie6167 <Vinnie6167@users.noreply.github.com> Co-authored-by: Ka-Hyun Nam <knam@nvidia.com> Co-authored-by: Alex Yang <aleyang@nvidia.com> Co-authored-by: Claude Opus 5 <noreply@anthropic.com> Co-authored-by: Jimmy Zhou <79552142+jimmyzho@users.noreply.github.com>

flashinfer-ai#4191 gave gen_trtllm_gen_gemm_module() and gen_trtllm_low_latency_gemm_module() an enable_rubin switch and used it for three things -- the cubin pin (TRTLLM_GEN_GEMM_RUBIN), the -DTLLM_RUBIN_FEATURES define, and the module name -- but left the gencode flags hardcoded: + sm100a_nvcc_flags, So on Rubin both modules were named *_sm107, pointed at the Rubin pin and compiled with the Rubin define, yet targeted compute_100a. Switch the arch flags on the same condition as everything else enable_rubin already selects. This is metadata, not kernels. csrc/trtllm_gemm_runner.cu and csrc/trtllm_low_latency_gemm_runner.cu contain no __global__, __device__ or <<<>>>, and neither do the trtllmGen_gemm_export headers; the GEMMs are precompiled cubins fetched from the pin and launched through the driver API (loadCubinData -> cuModuleGetFunction -> launchKernel). Only the device-link stub carried the wrong target, and nothing loads it, which is why Rubin has been functionally correct all along and no test caught this. PR flashinfer-ai#5096's per-provider SASS validation is the first thing to read that label. Controlled A/B on an SM107 (GR100, compute_cap 10.7), CUDA 13.5, same tree, only these lines differing, cuobjdump --list-elf on the built .so: enable_rubin=True (aot.py takes this path under has_sm107) before: GENCODE=['-gencode=arch=compute_100a,code=sm_100a'] trtllm_gemm_sm107.1.sm_100a.cubin trtllm_low_latency_gemm_sm107.1.sm_100a.cubin after: GENCODE=['-gencode=arch=compute_107a,code=sm_107a'] trtllm_gemm_sm107.1.sm_107a.cubin trtllm_low_latency_gemm_sm107.1.sm_107a.cubin enable_rubin=False (Blackwell, unchanged in both arms) trtllm_gemm.1.sm_100a.cubin trtllm_low_latency_gemm.1.sm_100a.cubin Both arms build clean, so CUDA 13.5 accepts compute_107a for these TUs. sm107a_nvcc_flags already existed in flashinfer/jit/core.py and was simply unused at these two sites. The Blackwell branch is the previous behaviour verbatim, so the only path that changes is has_sm107. Note the parenthesisation: a few lines above, gen_tgv_gemm_sm10x_module() writes this same idiom without parens, where `if/else` binds looser than `+` and silently drops the cflags list in the else branch. That is a separate pre-existing bug and is not touched here. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Split trtllm-gen cubin pins per arch

e7b0300

flashinfer-bot added op: gemm op: moe labels Jul 28, 2026

kahyunnam reviewed Jul 28, 2026

View reviewed changes

Comment thread flashinfer/jit/fused_moe.py

Comment thread flashinfer/jit/gemm/core.py Outdated

Comment thread flashinfer/jit/gemm/core.py Outdated

Vinnie6167 marked this pull request as ready for review July 28, 2026 01:57

Vinnie6167 requested review from Aneureka, IwakuraRein, aleozlx, bkryu, dhiraj113, feih-nv, jiahanc, nv-yunzheq, samuellees, sricketts and yzh119 as code owners July 28, 2026 01:57

Vinnie6167 requested review from cyx-6, jimmyzho, qiching, qsang-nv, saltyminty, yongwww and yyihuang as code owners July 28, 2026 01:57

Vinnie6167 added the run-ci label Jul 28, 2026

aleozlx merged commit 82b329b into flashinfer-ai:release-v0.6.16 Jul 28, 2026
6 of 7 checks passed

aleozlx mentioned this pull request Jul 28, 2026

release-v0.6.16: fix cubin checksum collision and unguarded Sm107a build break #4200

Merged

5 tasks

kahyunnam mentioned this pull request Jul 29, 2026

test: fix #4191 regressions in autotuner_core and symlink race test #4225

Merged

3 tasks

kahyunnam mentioned this pull request Jul 29, 2026

fix(trtllm): restrict routed-MoE backends to supported architectures (release port for #4107) #4230

Merged

Vinnie6167 mentioned this pull request Jul 29, 2026

fix(moe): name-filter output-scale-incompatible cubins until pinned packages carry mDtypeSfC #4235

Merged

jimmyzho mentioned this pull request Jul 30, 2026

deep_gemm: arch-based split for DEEPGEMM_RUBIN cubins #4261

Merged

Vinnie6167 mentioned this pull request Jul 30, 2026

SM 107 Reland + Merge Back from v0.6.16 Release Branch #4280

Merged

5 tasks

Vinnie6167 mentioned this pull request Sep 11, 2026

fix(jit): build the Rubin trtllm-gen GEMM modules for sm107a dierksen/flashinfer#2

Merged

Navigation Menu

Uh oh!

Split trtllm-gen cubin pins per arch - #4191

Split trtllm-gen cubin pins per arch#4191
aleozlx merged 3 commits into
flashinfer-ai:release-v0.6.16from
Vinnie6167:fix-rubin-features-blackwell-abi

Vinnie6167 commented Jul 28, 2026 •

edited

Loading

Uh oh!

gemini-code-assist Bot commented Jul 28, 2026

Uh oh!

coderabbitai Bot commented Jul 28, 2026 •

edited

Loading

Review skipped

Uh oh!

kahyunnam left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

gemini-code-assist Bot commented Jul 28, 2026

Uh oh!

flashinfer-bot commented Jul 28, 2026

Uh oh!

flashinfer-bot commented Jul 28, 2026

Uh oh!

Vinnie6167 commented Jul 28, 2026

Uh oh!

flashinfer-bot commented Jul 28, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Conversation

Vinnie6167 commented Jul 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

📌 Description

🔍 Related Issues

🚀 Pull Request Checklist

✅ Pre-commit Checks

🧪 Tests

Reviewer Notes

Uh oh!

gemini-code-assist Bot commented Jul 28, 2026

Uh oh!

coderabbitai Bot commented Jul 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Uh oh!

kahyunnam left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

gemini-code-assist Bot commented Jul 28, 2026

Uh oh!

flashinfer-bot commented Jul 28, 2026

Uh oh!

flashinfer-bot commented Jul 28, 2026

Uh oh!

Vinnie6167 commented Jul 28, 2026

Uh oh!

flashinfer-bot commented Jul 28, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Vinnie6167 commented Jul 28, 2026 •

edited

Loading

coderabbitai Bot commented Jul 28, 2026 •

edited

Loading