Questions about model performance reproduction

[jolin830]

Dear Author, I am trying to reproduce your code, but the current run is not optimal for performance. Here is what I have done:

1. I set use_adapter, info_nce, use_cl, use_prefix_p, prefix_projection to True in the config and left the rest unchanged.
2. I am using the dataset as shown in the picture. I did not run preprocess.py on the mosi,mosei dataset because I did not find the corresponding train/dev/test csv.
3. I have only run 50 epochs so far.

I think this is because I didn't run enough epochs or my dataset processing is wrong. I wonder if you can provide some suggestions? Much appreciated!

[Lingy12]

Hi, I am also trying to reproduce the experiments result. But I have some minor difference with the file path.

I created a fork here, if you are interested, https://github.com/Lingy12/UniMSE.

Do you mind share you data processing step and running step? I am able to run the main.py, but also not getting the desired result.

Maybe we can help each other. Thank you.

[LeMei]

I am sorry for that. Recently, I am busy with my graduation stuff. I am kind of forgetting the details. I will explain that tomorrow after I check this work.

[jolin830]

I am sorry for that. Recently, I am busy with my graduation stuff. I am kind of forgetting the details. I will explain that tomorrow after I check this work.

Thank you for your reply! I would also like to ask, how should the corresponding config be set up for optimal performance of your model? Is it the same as the config provided in your code? Looking forward to your reply!

[jolin830]

Hi, I am also trying to reproduce the experiments result. But I have some minor difference with the file path.

I created a fork here, if you are interested, https://github.com/Lingy12/UniMSE.

Do you mind share you data processing step and running step? I am able to run the main.py, but also not getting the desired result.

Maybe we can help each other. Thank you.

Hello! My data processing is like this: I first ran the preprocess.py for each dataset, followed by the data_precessor.py in src. I hope this will help you!

[Lingy12]

May I ask what's the parameters for main.py? I tried to run the pipeline, however, the performance for MOSEI can only reach ~20%. I use the default setting for main.py, but tweak to batch size to 8 to fit the GPU memory.

[huigeStudent]

你好，我也在尝试重现实验结果。但我的文件路径有一些细微的差别。

我在这里创建了一个分支，如果您有兴趣，https://github.com/Lingy12/UniMSE。

您介意分享您的数据处理步骤和运行步骤吗？我能够运行 main.py，但也没有得到所需的结果。

也许我们可以互相帮助。谢谢。

同学你复现出来了嘛 可以参考一下吗

[NoW0NDER]

Hi, I am also trying to reproduce the experiments result. But I have some minor difference with the file path.

I created a fork here, if you are interested, https://github.com/Lingy12/UniMSE.

Do you mind share you data processing step and running step? I am able to run the main.py, but also not getting the desired result.

Maybe we can help each other. Thank you.

Hi, I am also trying to reproduce this project. I think we can work together.

[huigeStudent]

Hi, I am also trying to reproduce the experiments result. But I have some minor difference with the file path. I created a fork here, if you are interested, https://github.com/Lingy12/UniMSE. Do you mind share you data processing step and running step? I am able to run the main.py, but also not getting the desired result. Maybe we can help each other. Thank you.

Hi, I am also trying to reproduce this project. I think we can work together.

这是我的结果，可以问问数据咋处理的嘛，作者的方式

[NoW0NDER]

哈喽，这是我的微信：If_Im_WonderingOn 20 Nov 2023, at 11:57, 2361806400 @.***> wrote:

Hi, I am also trying to reproduce the experiments result. But I have some minor difference with the file path. I created a fork here, if you are interested, https://github.com/Lingy12/UniMSE. Do you mind share you data processing step and running step? I am able to run the main.py, but also not getting the desired result. Maybe we can help each other. Thank you.

Hi, I am also trying to reproduce this project. I think we can work together.

这是我的结果，可以问问数据咋处理的嘛，作者的方式

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: @.***>

[wuxy07]

I'm having a KeyError problem in preprocess.py while reproducing it, could you please tell me about it? @jfightyr

[Mi7914]

@2361806400 你好，可以留个联系方式请教一下复现过程吗，感谢感谢