mrkeyoor.comAsk

deepseek-ai/DeepSelecttested 01 Oct 2026commit bfa4507

DeepSelect review

Trial DeepSelect only when its documented TopK contract fits your workload and you can qualify it on the intended hardware.

  • Install 17 s
  • Build 7 s
0/11passed, run exited non-zero
  • 35 packages
AI Toolsevaluationupdated 01 Nov 2026
+0stars / 7d

Consider DeepSelect when you need specialized TopK kernels for sparse attention or sampling and can validate them on your target hardware. DeepSelect supports NVIDIA CUDA and Huawei Ascend. Our run on October 1, 2026, at commit bfa4507 did not establish a working kernel installation, correctness or runtime performance.

Verdict

Trial DeepSelect only when its documented TopK contract fits your workload and you can qualify it on the intended hardware. Our run on October 1, 2026, at commit bfa4507 left installation and native compilation unverified: zero command exit codes coexist with build-backend errors, and pytest stopped during collection. Keep install time at its recorded 17 seconds and build time at its separate recorded 7 seconds; neither demonstrates runtime speed or deployment readiness.

We ran it

Lab card: what happened when we ran DeepSelect
Lab card from our sandbox run
Screenshot of DeepSelect (github.com/deepseek-ai/DeepSelect)
github.com/deepseek-ai/DeepSelect
Install✓ · 17s35 packages · 37 MB
Build✓ · 7s
Tests✗ · 6s0 passed · 0 failed · 11 errors of 11 (pytest)
Known vulns0(pip-audit)
Repo7049 files~1,020,766 lines of source · 154.4 MB · 0 CI workflows · tests dir

Answers from our run

Does DeepSelect build from source?

Dependencies installed in 17 seconds (35 packages), and the build succeeded in 7 seconds. We cloned commit bfa4507 into a clean Debian container with 3 CPUs and no project-specific setup.

Do DeepSelect's tests pass?

No. The project's own test command finished in 6 seconds and reported 0 passed, 0 failed, 11 collection errors (pytest).

Does DeepSelect have known vulnerabilities in its dependencies?

pip-audit found none in the dependency tree at the time of our run.

Who should not use DeepSelect?

Applications requiring topk values above 4096 or input types outside the documented paths.

What are the alternatives to DeepSelect?

PyTorch, CUTLASS. Trial DeepSelect only when its documented TopK contract fits your workload and you can qualify it on the intended hardware.

Setup2/5On the visible five-point rubric, substantial qualification remains: installation attempts show missing torch, test collection stops, and zero step exit codes do not verify a native build.
Docs4/5On the visible five-point rubric, detailed documentation covers supported workloads, installation, alignment and NaN behavior; execution remains unverified.
Community2/5Last push on 2026-09-30 and 546 stars.Editor’s note: The supplied primary extracts do not establish contributor activity, maintainer response times or the quality of support available to users.
Maturity2/5On the visible five-point rubric, substantial qualification remains because hardware and input constraints are narrow and our run supplies no kernel correctness or deployment result.

How we score

Who it’s for

Inference engineers evaluating specialized TopK kernels for DeepSeek Sparse Attention or sampling.
Kernel teams able to check compiler targets, tensor alignment and output correctness on their deployment hardware.
Services whose workloads fit the documented bfloat16 or CUDA float32 paths and topk limit of 4096.

Who it’s NOT for

Applications requiring topk values above 4096 or input types outside the documented paths.
Teams expecting this sandbox record to demonstrate successful native compilation or GPU execution.
Deployments that cannot accommodate aligned tensor strides or explicitly test NaN handling.
Users seeking a CPU implementation: the supplied build configuration selects CUDA or Ascend and rejects other platforms.

Setup reality

Our run on October 1, 2026, used commit bfa4507 in our Python 3.12 Debian Bookworm sandbox with 3 CPUs and 8 GB RAM. The harness recorded install exit 0 in 17 seconds and build exit 0 in 7 seconds, but those statuses do not establish a successful DeepSelect installation or native kernel build.

The installation log records build-backend failures caused by missing torch. Both editable and wheel requirement discovery reached tests/kernelkit/bench.py and failed while importing torch. The build tail contains Python SyntaxWarnings from bundled CUTLASS examples; it supplies no evidence that a CUDA or Ascend extension compiled.

The test command exited 1 after 6 seconds. Its stored parser summary is 0 passed, 0 failed, 0 skipped and 11 errors, with total recorded as 11. These are parser fields, not a verified count of unique or exhaustive tests. The raw footer reports 11 collection errors in 0.70 seconds, a different timing scope from the harness step duration.

The dependency record contains 35 packages and 37 MB. Pip-audit recorded 0 advisories, but scan coverage is unknown because scannedPackages is absent. These figures do not establish that DeepSelect installed or that its dependencies are secure.

Did our run establish a usable installation?

No. Our run on October 1, 2026, at commit bfa4507 recorded zero exit codes for the install and build steps, alongside installation errors in the captured output. The installation log records build-backend failures caused by missing torch. A successful wrapper command is insufficient evidence that the intended package installed.

Our harness stores step status, exit code, elapsed time, parsed test counts and captured log tails separately. Those fields must be read together here: the install status says success, while the tail shows failed editable and wheel build-requirement discovery. The build status also says success, but its tail shows Python warnings rather than native compilation evidence. The stored record does not explain how those command outcomes arose.

The MrKeyoor testing methodology explains these kinds of records and their limits. That page provides provenance context; its later description does not establish which procedure version produced this October 1 run. This correction uses the existing record and includes no new execution.

What stopped test collection?

The stored parser summary records 0 passed, 0 failed, 0 skipped and 11 errors. It does not establish 11 executed tests or exhaustive coverage of DeepSelect. The raw footer ends with 11 collection errors in 0.70 seconds; the harness records 6 seconds for the overall test step.

The captured tail identifies several distinct problems in bundled CUTLASS material: a missing torch import in CuTeDSL collection, missing pycute imports in Python tests, and a sharding test that cannot import running_on_ci from pytest's assertion utilities. The evidence does not support assigning every collection error to one missing dependency.

These collection errors provide no CUDA or Ascend correctness result. Our sandbox record does not establish kernel execution, workload throughput, multi-device behavior or deployment readiness.

Does the kernel contract fit your workload?

DeepSelect supports NVIDIA CUDA and Huawei Ascend. The README at the recorded commit describes specialized TopK paths for sparse attention and sampling, with topk capped at 4096 and tuning focused on 512. Float32 input is supported only on CUDA; Ascend supports bfloat16 input and int32 output indices.

Check layout before integrating it. Input rows require byte-aligned strides, and the last dimension must be contiguous. Returned tensors may be non-contiguous. Caller-owned index buffers also have alignment requirements. The interface marks begin and hint as unsupported, despite their presence in the signature.

The checked-in CUDA build targets sm_100a and sm_103a, skips SM80 and SM90 in its target list, and requires NVCC 12.9 or newer. The Ascend build imports torch_npu and requires an Ascend toolkit installation. These source requirements are separate from the harness's recorded build status.

NaN behavior also needs an integration check. The README says the default abort_when_nan_found=True invokes trap() and aborts the kernel. CUDA rows whose length is at most topk skip the NaN check. Validate the behavior your application needs before routing model outputs through it.

The README's speedup figures are upstream benchmark claims. Our run did not reproduce them, and installation timings do not measure TopK performance.

Which alternatives answer a different need?

PyTorch provides tensors that can live on the CPU or GPU. Keep its built-in torch.topk as your comparison baseline when evaluating whether a specialized extension is worth integrating. PyTorch is also the framework DeepSelect extends, so retaining PyTorch and adopting DeepSelect are compatible choices.

CUTLASS provides reusable components for building custom CUDA kernels. It is a development toolkit to consider when you intend to implement or modify kernels, rather than a demonstrated replacement for DeepSelect's Python TopK interface. DeepSelect's source build itself includes CUTLASS headers.

How to read the ratings

The assessed dimensions use a five-point rubric: 1 means a basic requirement is blocked or undocumented; 2 means substantial qualification remains; 3 means the dimension is adequate within a documented scope; 4 means detailed evidence supports it with stated limits; 5 requires demonstrated coverage across the intended use. Each rating applies only to its named dimension. Unknown evidence receives no numerical rating.

Setup ease is 2 because missing imports blocked installation attempts and test collection despite the stored success statuses. Documentation is 4 because the supplied material explains installation, supported cases, layout requirements and NaN behavior, although documentation cannot establish successful execution. Maturity is 2 because the narrow hardware and input contract still requires qualification that our run did not provide. Community is unassessed: the supplied primary extracts do not establish contributor activity, response times or support quality.

What to do before adoption

First check the documented dtype, topk, stride and compiler requirements against your actual workload and hardware. Then establish that the native extension installs and imports in that environment, run correctness checks against your baseline, and measure the workload you intend to deploy. Include short rows, NaNs and caller-owned output buffers if your application uses them. This stored sandbox run cannot substitute for those checks.

Alternatives

ProjectWhat it isPick it when
PyTorch gh↗The tensor framework DeepSelect extends, with CPU and GPU tensors.Keep its built-in torch.topk as the baseline while deciding whether your workload warrants a specialized extension. PyTorch remains a complement if you adopt DeepSelect.
CUTLASSReusable CUDA components for implementing custom kernels and related computations.Choose this development route when you need to build or modify kernels yourself. It is a complement and implementation resource, not an established drop-in replacement for DeepSelect's interface.

What people are saying

  1. [velocity-scout] deepseek-ai/DeepSelect

Sources

  1. DeepSelect README at the tested commit
  2. DeepSelect build configuration
  3. DeepSelect Python interface
  4. DeepSelect algorithm notes
  5. Open tensor-boundary issue
  6. Open wrong-device issue

More ai tools reviews