Skip to content

[None][fix] Add KV cache V2 recompute pause path - #14816

Closed
yizhang-nv wants to merge 1 commit into
NVIDIA:feat/deepseek_v4from
yizhang-nv:feat/kv-cache-v2-cancel-request
Closed

[None][fix] Add KV cache V2 recompute pause path#14816
yizhang-nv wants to merge 1 commit into
NVIDIA:feat/deepseek_v4from
yizhang-nv:feat/kv-cache-v2-cancel-request

Conversation

@yizhang-nv

Copy link
Copy Markdown
Member

@coderabbitai summary

Description

This PR separates destructive recompute pause from ordinary KV cache V2 suspension in the Python scheduler/executor path. It adds a scheduler output field for recompute-paused requests with a default empty value so V1 scheduler construction remains unchanged, gates suspended-victim promotion on host-tier availability, and has PyExecutor terminate/replay recompute-paused requests separately from ordinary suspended requests.

Test Coverage

  • python3 -m pytest -s tests/unittest/_torch/executor/test_kv_cache_v2_scheduler.py tests/unittest/_torch/executor/test_scheduler_serializable_output.py
  • python3 -m pytest -s tests/unittest/pyexecutor/test_iter_stats_populate.py tests/unittest/_torch/executor/test_py_scheduler.py -k "paused or full_pipeline_output_structure"
  • python3 -m pytest -s tests/unittest/_torch/executor/test_py_scheduler.py -k full_pipeline_output_structure
  • git diff --check
  • pre-commit hooks on commit

PR Checklist

Please review the following before submitting your PR:

  • PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.

  • PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.

  • Test cases are provided for new code paths (see test instructions)

  • Any new dependencies have been scanned for license and vulnerabilities

  • CODEOWNERS updated if ownership changes

  • Documentation updated as needed

  • Update tava architecture diagram if there is a significant design change in PR.

  • The reviewers assigned automatically/manually are appropriate for the PR.

  • Please check this after reviewing the above items as appropriate for this PR.

GitHub Bot Help

To see a list of available CI bot commands, please comment /bot help.

@yizhang-nv
yizhang-nv requested review from a team as code owners June 1, 2026 07:35
@yizhang-nv
yizhang-nv requested review from achartier and removed request for a team June 1, 2026 07:35
@yizhang-nv

Copy link
Copy Markdown
Member Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #51334 [ run ] triggered by Bot. Commit: e9ba97c Link to invocation

@nvpohanh

nvpohanh commented Jun 1, 2026

Copy link
Copy Markdown
Collaborator

@lowsfer could you review this? thanks

@yizhang-nv
yizhang-nv force-pushed the feat/kv-cache-v2-cancel-request branch from e9ba97c to 9d6ef45 Compare June 1, 2026 08:13
@nvpohanh
nvpohanh requested a review from lowsfer June 1, 2026 08:15
@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #51334 [ run ] completed with state FAILURE. Commit: e9ba97c
/LLM/main/L0_MergeRequest_PR pipeline #40746 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

Signed-off-by: Yi Zhang <187001205+yizhang-nv@users.noreply.github.com>
@yizhang-nv
yizhang-nv force-pushed the feat/kv-cache-v2-cancel-request branch from 9d6ef45 to 1e8ba4a Compare June 1, 2026 08:36
@yizhang-nv

Copy link
Copy Markdown
Member Author

/bot run

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #51337 [ run ] triggered by Bot. Commit: 1e8ba4a Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #51337 [ run ] completed with state FAILURE. Commit: 1e8ba4a
/LLM/main/L0_MergeRequest_PR pipeline #40749 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

@yizhang-nv
yizhang-nv requested a review from lancelly June 2, 2026 03:35
@yizhang-nv

Copy link
Copy Markdown
Member Author

/bot run --disable-fail-fast

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #51492 [ run ] triggered by Bot. Commit: 1e8ba4a Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #51492 [ run ] completed with state SUCCESS. Commit: 1e8ba4a
/LLM/main/L0_MergeRequest_PR pipeline #40898 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

@yizhang-nv

Copy link
Copy Markdown
Member Author

/bot run --disable-fail-fast

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #51796 [ run ] triggered by Bot. Commit: 1e8ba4a Link to invocation

@tensorrt-cicd

Copy link
Copy Markdown
Collaborator

PR_Github #51796 [ run ] completed with state SUCCESS. Commit: 1e8ba4a
/LLM/main/L0_MergeRequest_PR pipeline #41160 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

  • Please check the failed tests and fix your PR
  • If you cannot view the failures, ask the CI triggerer to share details
  • Once fixed, request an NVIDIA team member to trigger CI again

Link to invocation

@yizhang-nv

Copy link
Copy Markdown
Member Author

closing in favor of #15252

@yizhang-nv yizhang-nv closed this Jun 16, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants