Discrepancy Between Reproduced and Reported Results in CIFAR-10 Experiment

[zzpp-11]

Thank you for your excellent work, which has benefited me greatly. While reproducing the code, I followed the parameter settings in Project/experiments/cifar10/run_cifar10_repvgg_a0.sh and conducted experiments on the CIFAR-10 dataset. However, the experimental results differ significantly from those reported in the paper. Could you please let me know if I have made any mistakes in setting the parameters, or if there are any additional settings I need to pay attention to in the implementation of this method? I would greatly appreciate it if you could provide an answer or any more detailed information about the code.
（In the figure below, the left side shows the reproduced results, while the right side shows the experimental results reported in the paper.）

[gfyddha]

Hi, thank you for your interest in our work. The result on the right side is based on CIFAR-100, not CIFAR-10, so there might be some mistakes in your reproduction. We suggest double-checking whether your configurations align with the description in our paper. If the issue still exists, please provide more details so we can help you.

[229250919]

Hi, thank you for your interest in our work. The result on the right side is based on CIFAR-100, not CIFAR-10, so there might be some mistakes in your reproduction. We suggest double-checking whether your configurations align with the description in our paper. If the issue still exists, please provide more details so we can help you.

The scores for CIFAR-10 across the three models in the chart cannot be reproduced, and there are also quite unusual results for RepVGG-A0.