Skip to content

Navigation Menu

Appearance settings

View all features
- BY COMPANY SIZE
  Enterprises
  Small and medium teams
  Startups
  Nonprofits
- BY USE CASE
  App Modernization
  DevSecOps
  DevOps
  CI/CD
  View all use cases
- BY INDUSTRY
  Healthcare
  Financial services
  Manufacturing
  Government
  View all industries
View all solutions
- EXPLORE BY TOPIC
  AI
  Software Development
  DevOps
  Security
  View all topics
- EXPLORE BY TYPE
  Customer stories
  Events & webinars
  Ebooks & reports
  Business insights
  GitHub Skills
- SUPPORT & SERVICES
  Documentation
  Customer support
  Community forum
  Trust center
  Partners
View all resources
- COMMUNITY
  GitHub SponsorsFund open source developers
- PROGRAMS
  Security Lab
  Maintainer Community
  Accelerator
  GitHub Stars
  Archive Program
- REPOSITORIES
  Topics
  Trending
  Collections
- ENTERPRISE SOLUTIONS
  Enterprise platformAI-powered developer platform
- AVAILABLE ADD-ONS
  GitHub Advanced SecurityEnterprise-grade security features
  Copilot for BusinessEnterprise-grade AI features
  Premium SupportEnterprise-grade 24/7 support
Pricing

Search code, repositories, users, issues, pull requests...

Search

Clear

Search syntax tips

Provide feedback

We read every piece of feedback, and take your input very seriously.

Include my email address so I can be contacted

Saved searches

Use saved searches to filter your results more quickly

Name

Query

To see all available qualifiers, see our documentation.

Appearance settings

You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session.

Dismiss alert

InfiniTensor / InfiniOps Public

Notifications You must be signed in to change notification settings
Fork 3
Star 5

Code
Issues
Pull requests 18
Actions
Projects
Security and quality
Insights

Additional navigation options

Code
Issues
Pull requests
Actions
Projects
Security and quality
Insights

feat(ascend): add 9 Ascend operator kernels#47

Draft

zhangyue207 wants to merge 56 commits intomasterInfiniTensor/InfiniOps:masterfrom

feat/ascend-operatorsInfiniTensor/InfiniOps:feat/ascend-operatorsCopy head branch name to clipboard

Conversation Commits56 (56)Checks Files changed

Draft

feat(ascend): add 9 Ascend operator kernels#47
zhangyue207 wants to merge 56 commits intomasterfrom
feat/ascend-operators

Commits

Commits on Apr 17, 2026

fix(ascend): refine framework layer — caching, naming, build fixes
zhangyue
committed
feat(base): add new operator base classes and refine existing ones
zhangyue
committed
feat(cpu): add CPU implementations for Cast, Cat, Linear, Mul
zhangyue
committed
feat(ascend): add Ascend operator kernels for all operators
zhangyue
committed
feat(ascend): add multi-implementation variants and ATB operators
zhangyue
committed
feat(ascend): add custom AscendC kernels for RmsNorm and AddRmsNorm
zhangyue
committed
test(ascend): add comprehensive tests for all Ascend operators
zhangyue
committed
style: apply clang-format and fix code convention violations (round 1)
zhangyue
committed
style: fix code convention violations (round 2)
zhangyue
committed
style: fix code convention violations (round 3)
zhangyue
committed
fix(ci): add Ascend toolkit environment variables to CI Dockerfile
zhangyue
committed
style: apply ruff format and clang-format to all modified files
zhangyue
committed
style: fix ruff F401 lint errors for side-effect imports
zhangyue
committed
refactor(test): remove duplicate rms_norm test files and unify kernel loading
zhangyue
committed
feat(ascend): sync latest operators from feat/ascend-operators
zhangyue
committed
fix(ascend): re-upload cos_sin_cache when operator cache reuses stale data
zhangyue
committed
refactor: framework-level clear_cache() and skip 910B-unsupported tests
zhangyue
committed
test(ascend): enable ATB RoPE bfloat16 tests
zhangyue
committed
feat(ascend): add D2H-free paged_attention host tensor support
zhangyue
committed
style: apply clang-format to all modified C++ files
zhangyue
committed
style: apply ruff format to test and utility files
zhangyue
committed
fix(ascend): prevent double-free in operator destructors at process exit
zhangyue
committed
refactor(ascend): consolidate custom kernel macros into INFINI_HAS_CUSTOM_KERNELS
zhangyue
committed
docs(ascend): fix misleading destructor and rope restriction comments
zhangyue
committed
test(ascend): broaden rope impl/dtype coverage, add padding-slot case, narrow PA skip probe
zhangyue
committed
docs(perf): add e2e baseline reports and host-side gap findings
zhangyue
committed
docs(perf): update e2e progress with env-flag sweep + graph-mode root cause
zhangyue
committed
docs(perf): host-side cProfile analysis identifies current_stream_ptr as top delta
zhangyue
committed
docs(perf): record 0.5B host profile + stream-cache correctness regression
zhangyue
committed
docs(perf): record stream-ptr cache results — both models clear 80% eager
zhangyue
committed
docs(perf): design note — dispatch-count reduction (F1/F2) replaces fused-attention plan
zhangyue
committed
docs(perf): dispatch-count mystery resolved — vllm-ascend uses fx fusion passes
zhangyue
committed
docs(perf): capture-replay sanity — all 4 infini.ops pass npugraph capture
zhangyue
committed
docs(perf): record G2 FX-rewrite throughput (3B graph 63.8% -> 71.5%)
zhangyue
committed
docs(perf): scoped G1 fusion design — P-1 split_rope + P-3 noop_elimination
zhangyue
committed
docs(perf): update G1 design with FX dump — P-1 pattern confirmed 36x/forward
zhangyue
committed
docs(perf): record fusion-pass scaffolding commit (zero passes, P-3 deferred)
zhangyue
committed
docs(perf): record #28 split_rope_collapse as measured-within-noise opt-in
zhangyue
committed
docs(perf): mission final report — eager met, graph capped at 71%
zhangyue
committed
docs(perf): rewrite mission_final per team-lead review — honest framing, ratio table first, decision matrix
zhangyue
committed
docs(perf): reflect #29 GatherV3-at-parity finding in mission final
zhangyue
committed
chore(pr47): drop mission perf docs from PR scope + default AUTO_DETECT_BACKENDS=OFF
zhangyue
committed
revert(pr47): restore AUTO_DETECT_BACKENDS=ON default
zhangyue
committed
build(pr47): add torch.libs/ to rpath-link for bundled libgfortran
zhangyue
committed
fix(ascend): adopt PR #63/#60 master API — GetWorkspacePool/Ensure rename + drop registry.h (SFINAE autodetect)
zhangyue
committed
fix(scripts): py::arg order in bindings generator must match C++ param order (impl before stream)
zhangyue
committed
fix(ci): treat exit 137 as success when pytest junit XML reports no failures
zhangyue
committed
fix(ascend): remove stale 910B skip guard from PagedAttention test
zhangyue
committed
feat(ascend): support non-neox rotaryMode via ATB RopeParam rotaryCoeff
zhangyue
committed
feat(ascend/rotary_embedding): add impl=2 via `aclnnRopeWithSinCosCache`
zhangyue
committed
docs(paged_attention): explain why `seq_lens_host` / `block_table_host` exist
zhangyue
committed
perf(ascend/reshape_and_cache): replace int64 slot_mapping D2H with async `aclnnCast`
zhangyue
committed
feat(rotary_embedding): make `query_out` / `key_out` optional (inplace-default)
zhangyue
committed
feat(flash_attention): add vLLM-style `sliding_window` entry (additive)
zhangyue
committed
style: apply clang-format to recent API-alignment changes
zhangyue
committed
style: apply clang-format to silu_and_mul/causal_softmax/swiglu kernel files
zhangyue
committed

Footer

© 2026 GitHub, Inc.

Footer navigation

Terms
Privacy
Security
Status
Community
Docs
Contact

You can’t perform that action at this time.