Results on different slices

[Shrimp-99]

I run the purchase data and set shard=1, slice=1 and epoch=1, the output model is trained on 0.25 million data for 1 epoch and the accuracy is around 0.955. Then I set shard=1, slice=16, and epoch=1. The train always take places on the 8th slice and epoch for it is 1 while for other slices is 0, according to the 'avg_epochs_per_slice' and 'slice_epochs' calculated. But the output model still uses 0.14 million data and the accuracy is around 0.945. I didn't see significant difference between them as Fig5 in paper shows. Is there any mistake using the above settings?

[cnhbww]

I run the purchase data and set shard=1, slice=1 and epoch=1, the output model is trained on 0.25 million data for 1 epoch and the accuracy is around 0.955. Then I set shard=1, slice=16, and epoch=1. The train always take places on the 8th slice and epoch for it is 1 while for other slices is 0, according to the 'avg_epochs_per_slice' and 'slice_epochs' calculated. But the output model still uses 0.14 million data and the accuracy is around 0.945. I didn't see significant difference between them as Fig5 in paper shows. Is there any mistake using the above settings?

Hello, I am a newbie who has just entered this field. I tried to directly clone the author's project for experiments, but encountered many problems. Can you share your experimental project with me? Thank you very much if you can.

[Sushant-kumar-pal]

I run the purchase data and set shard=1, slice=1 and epoch=1, the output model is trained on 0.25 million data for 1 epoch and the accuracy is around 0.955. Then I set shard=1, slice=16, and epoch=1. The train always take places on the 8th slice and epoch for it is 1 while for other slices is 0, according to the 'avg_epochs_per_slice' and 'slice_epochs' calculated. But the output model still uses 0.14 million data and the accuracy is around 0.945. I didn't see significant difference between them as Fig5 in paper shows. Is there any mistake using the above settings?

Hello, I am a newbie who has just entered this field. I tried to directly clone the author's project for experiments, but encountered many problems. Can you share your experimental project with me? Thank you very much if you can.

do u got the experimental results?Can you share with me please

[cnhbww]

我运行购买数据并设置 shard=1、slice=1 和 epoch=1，输出模型在 1 个 epoch 的 25 万条数据上训练，准确率约为 0.955。然后我设置 shard=1、slice=16 和 epoch=1。根据计算的“avg_epochs_per_slice”和“slice_epochs”，火车总是发生在第 8 个切片和纪元上，它的纪元为 1，而其他切片的纪元为 0。但输出模型仍然使用 0.14 万个数据，准确率约为 0.945。我没有看到它们之间的显着差异，如论文中的图 5 所示。使用上述设置有什么错误吗？

您好，我是刚进入这个领域的新手。我尝试直接克隆作者的项目进行实验，但遇到了很多问题。你能和我分享你的实验项目吗？如果可以的话，非常感谢你。

你有实验结果吗？你能和我分享吗

no, I didn't even get through the code，Can you give me some advice or help? pls，I will also share my progress with you later