Skip to content
View LeslieWylie's full-sized avatar
🎯
Focusing
🎯
Focusing
7 followers · 12 following
  • Beijing University of Posts and Telecommunications
  • Beijing

Achievements

Achievement: Pair ExtraordinaireAchievement: Pull Sharkx2Achievement: YOLO

Achievements

Achievement: Pair ExtraordinaireAchievement: Pull Sharkx2Achievement: YOLO

Highlights

  • Pro

Block or report LeslieWylie

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
LeslieWylie/README.md

Lele — Agent Engineering & Inference Systems

Lele · Open-source contributor

I contribute to inference runtimes and agent tooling, focusing on multimodal generation, LoRA correctness, and agent memory.

我关注 Agent 工程与模型推理优化，通过上游 PR 贡献模型接入、推理优化和运行时修复。

Merged PRs · Open PRs · Agent projects · All upstream PRs ↗

Merged PRs

Project Contribution PR
vLLM [LoRA] Keep failed activations retryable #61618
Sana [Sol-H3] Support custom resident text encoders #505
vLLM-Omni Preserve CUDA toolkit header precedence in a2a JIT builds #7599
FastVideo Import MiniMax H3 Comfy NVFP4-AWQ text encoder #1865

All merged upstream PRs →

Open PRs

Selected open, non-draft proposals. Titles and status sync from GitHub daily.

Project Proposed change PR
vLLM-Omni Fuse H3 final affine and FP32 promotion #8677
vLLM-Omni Load Alibaba PDD adapters for MiniMax H3 #7425
vLLM [Frontend] Reject unsupported nested Anthropic tool-result content #63404
SGLang anthropic: forward inline MP4 user content to Chat #46119
SGLang anthropic: reject unsupported nested tool-result blocks #46118
huggingface/diffusers [MiniMax-H3] Postprocess decoded video in frame batches #15070
huggingface/diffusers [MiniMax-H3] Reuse Ulysses shard lengths across transformer blocks #15064
vLLM [LoRA] Account for CLIP CLS rows in LLaVA tower mappings #61799

All open upstream PRs →

Automatic PR sync · Daily at 08:17 Beijing time.

Agent projects

Project What it does
repository-memory Source-backed memory for coding agents, with Git citations, a local store, CLI, and MCP server.
dsh-ops-kit Evidence-driven memory, orchestration, benchmark checks, and release workflows for DeepSeek Harness.
dsh-session-search-pro Search past and live DeepSeek Harness sessions through the runtime's own session service.
review-workflow Structured review with independent panelists, disagreement resolution, and a separate critic.

Also exploring coding-agent evaluation and verifiable distributed inference.

Tools I use: Python · TypeScript / JavaScript · PyTorch · MCP · FastAPI · Docker

Based in Beijing · WeChat: yaole516

Pinned Loading

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

    A framework for efficient model inference with omni-modality models

    Python 7.4k 2k

  2. hao-ai-lab/FastVideo hao-ai-lab/FastVideo Public

    A unified inference and post-training framework for accelerated video generation.

    Python 4.7k 507

  3. vllm-project/vllm vllm-project/vllm Public

    A high-throughput and memory-efficient inference and serving engine for LLMs

    Python 95.6k 24k

  4. sgl-project/sglang sgl-project/sglang Public

    SGLang is a high-performance serving framework for large language models and multimodal models.

    Python 38.2k 9.9k

  5. repository-memory repository-memory Public

    Source-backed memory for AI agents: Git citations, local retrieval, CLI, and MCP.

    Python

  6. dsh-ops-kit dsh-ops-kit Public

    Evidence-driven memory, orchestration, benchmarks, and release workflows for DeepSeek Harness.

    TypeScript 2