Wei.Wang's repositories
daily_check_in
每天签到
ai-pr-reviewer
AI-based Pull Request Summarizer and Reviewer with Chat Capabilities.
AI-System
System for AI Education Resource.
AISystem
AISystem 主要是指AI系统，包括AI芯片、AI编译器、AI推理和训练框架等AI全栈底层技术
baby-llama2-chinese
用于从头预训练+SFT一个小参数量的中文LLaMa2的仓库；24G单卡即可运行得到一个具备简单中文问答能力的chat-llama2.
BentoML
The most flexible way to serve AI/ML models in production - Build Model Inference Service, LLM APIs, Inference Graph/Pipelines, Compound AI systems, Multi-Modal, RAG as a Service, and more!
cube-studio
cube studio开源云原生一站式机器学习/深度学习AI平台，支持sso登录，多租户/多项目组，大数据平台对接，notebook在线开发，拖拉拽任务流pipeline编排，多机多卡分布式训练，超参搜索，推理服务VGPU，边缘计算，serverless，标注平台，自动化标注，数据集管理，大模型微调，vllm大模型推理，llmops，私有知识库，AI模型应用商店，支持模型一键开发/推理/微调，支持国产cpu/gpu/npu芯片，支持RDMA，支持pytorch/tf/mxnet/deepspeed/paddle/colossalai/horovod/spark/ray/volcano分布式
duckduckgo-api
免费的无限制的搜索接口
gpu_poor
Calculate token/s & GPU memory requirement for any LLM. Supports llama.cpp/ggml/bnb/QLoRA quantization
HAMi
OpenAIOS vGPU scheduler for Kubernetes is originated from the OpenAIOS project to virtualize GPU device memory.
HAMi-core
HAMi-core compiles libvgpu.so, which ensures hard limit on GPU in container
hd_write_verify
hd_write_verify & hd_write_verify_dump is a tool for testing disk stability and verifying data consistency, for example: physical disk: ide/sata/scsi/ssd/iscsi/fc/raid. virtual disk: loop/nbd/lvm/soft raid. virtual machine disk: ide/sata/scsi/virtio-blk/virtio-scsi.
light-hf-proxy
A light proxy solution for HuggingFace hub.
metahuman-stream
Real time streaming digital human based on nerf
olah
Self-hosted huggingface mirror service.
pfconductor
Management system for PureFlash, the server SAN designed for flash device.
PureFlash
A ServerSAN storage system designed for flash device
qcow2-defrag
qemu-img convert command(eg: qemu-img convert -f qcow2 -o preallocation=off -O qcow2 src.qcow2 dst.qcow2) will lost snapshots of qcow2 image. qcow2-defrag is a tool for defragging qcow2 image which can solve this problem.
qcow2-delta
qcow2镜像合并工具、虚拟机磁盘克隆工具
qcow2-dump
qcow2-dump is a useful tool for checking and repairing damaged qcow2 image, it has some improvements compare with qemu-img check command (qcow2-dump has all functions which qemu-img check command has).
virtopt
Virtualization Optimization
vllm
A high-throughput and memory-efficient inference and serving engine for LLMs
volcano-vgpu-device-plugin
Device-plugin for volcano vgpu which support hard resource isolation
zstack-utility
Agents and tools for project ZStack http://zstack.org
zstack-vyos
ZStack virtual router agent