Request for Complete Code Release and Weight Files for CMITM

[zoule41]

Dear Dr.Chen,

I am writing to express my interest in your work on CMITM, which I found extremely insightful and relevant to my current research. I have gone through your published code on this GitHub repository and appreciate your effort in making this work accessible to the community.

However, I noticed that the current version of the open-sourced code does not fully reflect the methodology mentioned in your paper, particularly the two-stage training process - the initial training with reconstruction loss followed by training with contrastive loss. This aspect of your work is particularly intriguing, and a detailed implementation would be highly beneficial for a thorough understanding.

Moreover, there seems to be a lack of details on how text reconstruction loss is implemented in the code. As someone keenly interested in learning from your work, these details are crucial for me and likely for other researchers who are following your work.

Therefore, I kindly request if you could release the complete code along with the weight files that encompass these critical aspects of your methodology. Having access to the full implementation would greatly aid in comprehending and potentially building upon your valuable research.

Thank you very much for considering this request. I am looking forward to the complete code and learning more from your work.

Best regards,
Le Zou from NUDT