• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

kubernetes-sigs / inference-perf / 41527634063
87%

Build:
DEFAULT BRANCH: main
Ran 02 Nov 2026 09:33PM UTC
Jobs 1
Files 137
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

02 Nov 2026 09:21PM UTC coverage: 85.662%. Remained the same
41527634063

push

github

web-flow
Bump llm-d-inference-sim to v1.0.0 (#1013)

Part of #1012

The dev shell built llm-d-inference-sim 0.6.1 against zeromq and
libtokenizers. It now builds v1.0.0, which needs Go 1.27 and neither
library.

Changes:
* Bump the sim in `flake.nix` to v1.0.0.
* Add a `nixpkgs-go` input on nixos-26.05, used only for the sim's Go
toolchain. v1.0.0's go.mod asks for go 1.27.2. nixos-25.11 stops at Go
1.26. The dev shell stays on 25.11. `flake.lock` only gains the new
input.
* Drop the zeromq and libtokenizers build inputs, pkg-config, and the
libtokenizers derivation. v0.8.0 moved the sim to a pure-Go ZeroMQ
client and took tokenization out of the binary. Nothing else in the
flake used them.
* Pass the sim's TTFT and ITL flags with a unit in
`test_metrics_fidelity.py`, `test_load_shape_accuracy_sim.py` and
`test_worker_matrix_smoke.py`. Since v0.9.0 the sim parses them as Go
durations and exits on a bare number. The values are unchanged.

```
$ llm-d-inference-sim --model google/gemma-3-270m --time-to-first-token 150
"failed to read configuration" err="invalid argument \"150\" for \"--time-to-first-token\" flag: time: missing unit in duration \"150\""
```

Other changes:
* Build only `cmd/llm-d-inference-sim`. v1.0.0 adds a `dataset-tool`
binary the suite does not use. Building it would put it on the dev
shell's PATH.

The other breaking changes in the v0.7.0 to v1.0.0 release notes do not
reach the suite. Readiness polls `HEAD /`, not the removed `/ready`. No
e2e test reads the renamed cache metrics or `/fake-metrics`. No sim test
asserts the sim's own prompt token count.

`nix develop -c pdm run test:e2e -rs` locally gives 102 passed and 31
skipped on main and on this branch, with the same tests in each bucket.
The skips are the vLLM, guidellm, vllm-bench parity and live-tier cases.
Every sim test ran.

13215 of 15427 relevant lines covered (85.66%)

0.86 hits per line

Jobs
ID Job ID Ran Files Coverage
1 41527634063.1 02 Nov 2026 09:33PM UTC 137
85.66
GitHub Action Run
Source Files on build 41527634063
  • Tree
  • List 137
  • Changed 0
  • Source Changed 0
  • Coverage Changed 0
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Github Actions Build #41527634063
  • 7b37cd0f on github
  • Prev Build on main (#41105705751)
  • Next Build on main (#41527937623)
  • Delete
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc