Configure and troubleshoot RCCL-based distributed training across Hygon DCUs.
npx skills add https://github.com/dongg622/china-ai-chip-skill --skill dtk-rccl
Please help me install this Agent Skill. Skill: dtk-rccl Source: https://github.com/dongg622/china-ai-chip-skill/tree/main/Hygon/dtk-rccl Command: npx skills add https://github.com/dongg622/china-ai-chip-skill --skill dtk-rccl
This Skill facilitates the deployment and troubleshooting of RCCL-based distributed training setups on Huawei Hygon DCUs, ensuring reliable multi-node communication for AI workloads.
Load the environment, verify RCCL library availability, and follow the step-by-step commands to initialize and test multi-node communications with detailed diagnostics.
High-intent search queries and answers about installing and using this skill.
Multi-node distributed training on Hygon DCUs requires configuring RCCL environments across multiple nodes. This Skill provides step-by-step commands to initialize environments, verify RCCL library availability, and test multi-node communications for AI training workflows.
RCCL communication failures across DCUs often stem from network configuration or XGMI connection issues. This Skill assists in diagnosing multi-node communication problems and provides troubleshooting steps to resolve network connections for reliable distributed training.
Yes, you can optimize multi-node communication performance for Hygon DCU clusters. This Skill provides guidance for diagnosing communication issues, optimizing performance, and configuring network connections like XGMI to ensure efficient AI training.
RCCL is used for efficient multi-node communication in distributed AI training. This Skill facilitates RCCL-based setup and troubleshooting across multiple Hygon DCUs, ensuring reliable communication and performance optimization for machine learning workloads.
RCCL distributed training on DCUs requires specific network configurations like XGMI. This Skill streamlines environment setup, persists configuration details, and offers troubleshooting steps to ensure proper network connections for multi-node training.
The best way to troubleshoot RCCL environment setup is to load the environment, verify library availability, and run detailed diagnostics. This Skill provides step-by-step commands to initialize and test multi-node communications effectively.
kubonyProvision Google Cloud Platform virtual machines with firewall configuration.
pwarnockSet up and configure Bun development environments with TypeScript and hot-reloading workflows.
ovachieverApply PyTorch FSDP for sharded, mixed-precision distributed training of large models.
nateaEnable multi-node synchronization and hybrid vector+metadata search in AgentDB.
