Choose a training strategy and tune your MoE VLM runs.
Guides selection and tuning of parallel training strategies for mixture-of-experts vision-language models in Megatron Bridge.
When to use it
Use when training MoE vision-language models or investigating a commit that caused training failure or an out-of-memory error.
Give it a MoE VLM training question; it returns analysis and configuration instructions.
No setup requirements listed.
No additional actions listed.