Fine-tuning model memory usage

[ShengYang-pixel]

Thank you for your excellent work.
Just one question about fine-tuning:
When full-parameter fine-tuning the Infinity-2B model, are four 3090 GPUs (with a memory of 24GB each * 4) sufficient? When fine-tuning the Infinity-8B model, are eight A6000 GPUs (with memory of 48GB each * 8) sufficient?