vLLM GRPO GPU Memory Utilization Estimator

Paste your HuggingFace model config URL (ending in config.json), and enter experiment details. Model parameters are automatically extracted and estimated from the config.

Note: This is a general recommendation and may not be optimal for your specific environment. Always verify your actual training GPU requirements. For example, if you're using DeepSpeed, consider utilizing their memory estimation tool: https://deepspeed.readthedocs.io/en/latest/memory.html

If you encounter "not enough memory" errors, try increasing the GPU memory utilization setting. If you experience out-of-memory (OOM) errors, lower the utilization value and/or reduce your batch size.