Debug distributed training errors with structured investigation protocols.
npx skills add https://github.com/ByteDance-Seed/VeOmni --skill veomni-debug
Please help me install this Agent Skill. Skill: veomni-debug Source: https://github.com/ByteDance-Seed/VeOmni/tree/main/.agents/skills/veomni-debug Command: npx skills add https://github.com/ByteDance-Seed/VeOmni --skill veomni-debug
This Skill provides comprehensive guidance and structured protocols to identify, analyze, and fix complex bugs, errors, and unexpected behaviors in large-scale distributed model training environments.
Read the troubleshooting guide to systematically diagnose a training hang or crash caused by distribution misconfiguration.
High-intent search queries and answers about installing and using this skill.
Debug distributed training hangs by following structured investigation protocols to isolate distribution misconfigurations. It guides you through root cause analysis and hypothesis testing to systematically resolve stalls in complex training environments.
Troubleshoot FSDP2 OOM errors by applying structured debugging protocols to isolate memory issues. The approach guides you through investigating sharding mismatches and verifying fixes to efficiently resolve out-of-memory crashes in large-scale model training.
The best way to investigate sharding mismatches is using a structured approach that guides you through root cause investigation and fix verification. This ensures accurate diagnosis of sharding errors in distributed training scenarios.
Yes, this debugging approach works for MoE and sequence parallelism errors by providing structured protocols applicable to large-scale model training scenarios. It ensures accurate diagnosis and resolution of issues across these specific distributed environments.
You need a structured protocol for error analysis when troubleshooting complex bugs, errors, or unexpected behaviors in large-scale distributed model training. It provides the necessary steps to systematically investigate, diagnose, and resolve intricate training failures.
tangzheng202202Guide PyTorch FSDP parameter sharding, mixed precision, and CPU offloading.
henrycashe26Diagnose machine learning training runs by interpreting loss curves and validation metrics.
NVIDIA-NeMoOffload activations and optimizer states to CPU during large-scale model training.
Yangel-hideAutomate PyTorch training workflows with distributed strategies like DDP, FSDP, and DeepSpeed.
