Skip to content
View LOGO127's full-sized avatar
🌴
On vacation
🌴
On vacation
  • Hangzhou, Zhejiang, China

Block or report LOGO127

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
LOGO127/README.md
LOGO127 — AI Systems, LLM Training, and Inference

Zijian Luo · LOGO127

AI systems first · AI for Science alongside

Master's student at Zhejiang University. I build reproducible systems to understand language models, with a growing focus on reliable inference and stateful agent APIs.

Projects · Contribution log · Upstream PRs


✅ Merged contributions · 8

External PRs authored by me and merged upstream. Verified September 12, 2026.

Agentic API #257 adds bounded transient response sessions, continuation history, retention accounting, and cancellation cleanup. Approved and merged September 11, 2026 (UTC). The final version includes maintainer-contributed WebSocket integration.

Agentic API #258 rejects unsupported message-file inputs on typed Responses paths. Merged September 9, 2026 (UTC), including maintainer-contributed schema and validation updates.

WebCodex #372 waits for npm installer download streams to close before cleanup. Merged September 10, 2026 (UTC).

vLLM-Omni #7065 fixes Higgs Audio v3 voice-clone token validation while preserving reference-audio placement across chunked prefill. Merged after collaborator review on September 4, 2026 (UTC).1

DeePMD-kit #6010 excludes virtual atoms from force, Hessian and spin-force error metrics while preserving padded detail outputs. Approved and merged by project member njzjz on September 7, 2026 (UTC).2

Speculators #1096 fixes multi-engine evaluation undercount by summing per-position acceptance counters across engines. Approved and merged by fynnsu on September 8, 2026 (UTC).3

MACE #1712 rejects HTML responses before caching model downloads. Approved and merged by aacostadiaz on September 9, 2026 (UTC).4

ONNX Runtime #32435 prevents identity transposes from being incorrectly fused into Gemm, with regression coverage for input, output and mixed transpose paths. Approved and merged by xadupre on September 9, 2026 (UTC).5

🎯 Current contribution focus

Track Engineering focus Public work
Agentic API Bounded response sessions; continuation and storage contracts #257, #258 — merged; replay-catalog collaboration in Zheng-Lu's #261 — open6
vLLM-Omni Audio efficiency; duplex lifecycle; reproducible multi-session validation Mimi #7399, cancellation #7429, sample-rate snapshots #7448 — open; validation #7428 — draft7
Vime Post-training data ordering, resume correctness, and numerical precision cursor #416, entropy #419, RoPE #420 — open
XGrammar Structured-generation tool input validation #881 — draft; local regression verified, awaiting review8
DeePMD-kit Ragged graph batching; charge/spin and numerical correctness #6008 — open at last check9; #6010 — merged2
MACE Scientific model downloads and TorchScript export #1712 — merged4; #1717 — draft at last check10

These are contribution areas I am working toward maintaining, not assigned module ownership. Open and draft PRs are not accepted contributions. The dated contribution log separates merged work, review candidates, and unpublished experiments.

📦 Personal projects

Project Focus
wechat-ai-memory Local-first, traceable conversation context
CS336 Assignment 1 From-scratch language-model foundations
cs336.2026 Systems-first learning and implementation notes

🔍 Engineering approach

Case study: honoring Router queue deadlines under contention — token, queue and HTTP regressions; CI-pinned Rust validation; PR #248 submitted after personal review and sign-off; awaiting upstream review, not merged.

Case study: preserving heterogeneous conversation data in GuideLLM — identical baseline/candidate regressions and a real CLI-to-localhost HTTP check; issue evidence shared, patch unpublished and personal review pending.

Case study: preserving Vime's sample stream across epochs — matched baseline/candidate tests and real cursor save/load; PR #416 is open after personal verification and sign-off; not an accepted fix.

Case study: validating an XGrammar parser-history regression — attributed fork candidate with deterministic tests and bytewise mask comparisons; upstream coordination pending, not merged.

Case study: checkpoint-save failure and distributed recovery — real CPU/Gloo failure reproduction with a matched successful-save control; PR #3151 closed without merging; upstream acceptance and GPU/NCCL coverage are not claimed.

  • Reproduce failures and preserve passing controls before changing behavior.
  • Keep patches scoped, with explicit dependency and compatibility boundaries.
  • Separate local tests, upstream CI, and actual model/hardware validation.
  • Explain design decisions and respond to review with reproducible evidence.

Python · Rust · PyTorch · pytest · Git · uv


Build the smallest useful version. Test the assumptions. Explain what happened.

Footnotes

  1. vLLM-Omni. Merged PR #7065.

  2. DeePMD-kit. Merge commit 28b7d068. Merged September 7, 2026, 22:14 UTC (September 8, 06:14 Asia/Shanghai); verified directly on September 8. This is an accepted contribution, not an assigned module-maintainer role. 2

  3. Speculators. Merge commit 04e08de9. PR authored by LOGO127, merged September 8, 2026, 21:19 UTC (September 9, 05:19 Asia/Shanghai); verified September 9. This is an accepted contribution, not a module-maintainer appointment or a throughput benchmark.

  4. MACE. Merge commit 52302120. PR authored by LOGO127, merged September 9, 2026, 10:37 UTC (18:37 Asia/Shanghai); verified directly. No module-maintainer appointment is implied. 2

  5. ONNX Runtime. Merge commit b905e946. PR authored by LOGO127, merged September 9, 2026, 11:48 UTC (19:48 Asia/Shanghai); verified directly. This is a graph-optimizer correctness contribution, not a performance benchmark or a maintainer appointment.

  6. Agentic API. Sessions #257 and typed Responses file validation #258 are merged, verified September 12, 2026. Final integration includes other contributors' work; sole authorship of every merged line, complete #238 coverage, and module ownership are not claimed. #261 is another author's PR and is not included in my authored merge count.

  7. vLLM-Omni. Current statuses checked September 12, 2026. Reference-audio diagnostics #7098 and shared RPC deadline #7151 also remain open. Passing build/documentation checks are not full GPU/model validation.

  8. XGrammar. Malformed function-tool validation #881, head a79fd29, checked September 8, 2026. Draft publication is not upstream acceptance; local tests are not upstream CI.

  9. DeePMD-kit. Ragged charge/spin batching #6008. Open at the September 6 check; #6010's later merge is recorded separately below.

  10. MACE. Gated-block TorchScript export #1717 was an open draft at the September 8, 2026 check, not counted as merged. #1712's later merge is recorded in footnote 9.

Pinned Loading

  1. vllm-project/vllm-omni vllm-project/vllm-omni Public

    A framework for efficient model inference with omni-modality models

    Python 6.8k 1.7k

  2. vllm-project/speculators vllm-project/speculators Public

    A unified library for building, evaluating, and storing speculative decoding algorithms for LLM inference in vLLM

    Python 826 218

  3. deepmodeling/deepmd-kit deepmodeling/deepmd-kit Public

    A deep learning package for many-body potential energy representation and molecular dynamics

    Python 2k 649