Skip to content

[spec decoding] support inkling dspark - #31847

Merged
Qiaolin-Yu merged 8 commits into
mainfrom
dspark-integration
Aug 10, 2026
Merged

Qiaolin-Yu merged 8 commits into
mainfrom
dspark-integration

Conversation

@Qiaolin-Yu

@Qiaolin-Yu Qiaolin-Yu commented Jul 20, 2026 •

Copy link
Copy Markdown
Collaborator

Motivation

Modifications

Accuracy Tests

Speed Tests and Profiling

Checklist

Review and Merge Process

  1. Ping Merge Oncalls to start the process. See the PR Merge Process.
  2. Get approvals from CODEOWNERS and other reviewers.
  3. Trigger CI tests with comments or contact authorized users to do so.
    • Common commands include /tag-and-rerun-ci, /tag-run-ci-label, /rerun-failed-ci
  4. After green CI and required approvals, ask Merge Oncalls or people with Write permission to merge the PR.

CI States

Latest PR Test (Base): ✅ Run #31350816508
Latest PR Test (Extra): ✅ Run #31350816469

@gemini-code-assist

Copy link
Copy Markdown
Contributor

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

@Qiaolin-Yu Qiaolin-Yu changed the title wip [spec decoding] support inkling dspark Jul 27, 2026
@Qiaolin-Yu Qiaolin-Yu added the run-ci CI: run the baseline test suite on this PR label Jul 30, 2026
@ispobock ispobock added bypass-fastfail run-ci-extra CI: also run the extra suite (requires run-ci) labels Jul 30, 2026
Boreas618 added a commit to Boreas618/sglang that referenced this pull request Aug 4, 2026
…fill-graph static buffer)

The Inkling-specific serving glue PR sgl-project#31847 does not deliver: aux-layer
capture on the target via a single-buffer _AuxCaptureSink with a stash/pop
handoff, set_dflash_layers_to_capture / lm_head attach points for the
DSpark worker, tensor-form aux_hidden_states handling in LogitsProcessor,
and full prefill CUDA-graph support through a stable-address aux buffer
re-armed at replay.
Boreas618 added a commit to Boreas618/sglang that referenced this pull request Aug 4, 2026
Complement to PR sgl-project#31847, which bounds the mask token by padded_vocab_size
in the worker; this relaxes the config-side hard error to a warning for
targets whose mask token lives in a padded embedding slot (Inkling-Small:
embed rows 201024 > logical vocab 200058).
@ispobock

Copy link
Copy Markdown
Collaborator

/rerun-failed-ci

@Qiaolin-Yu
Qiaolin-Yu merged commit 733c05c into main Aug 10, 2026
299 of 316 checks passed
@Qiaolin-Yu
Qiaolin-Yu deleted the dspark-integration branch August 10, 2026 20:21
vstone-w pushed a commit to ClownBin/sglang that referenced this pull request Aug 12, 2026
saturn-acc pushed a commit to saturn-acc/sglang that referenced this pull request Aug 16, 2026
Atituiset pushed a commit to Atituiset/sglang that referenced this pull request Sep 10, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bypass-fastfail jit-kernel run-ci CI: run the baseline test suite on this PR run-ci-extra CI: also run the extra suite (requires run-ci)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants