Kriti Aggarwal
Kriti Aggarwal
Thanks Lucas. I have another question regarding the sampling for the data. Since, the data distribution for BiT-L is heavily tailed. Was the distribution converted to uniform or other to...
use the following command to install opencv: pip install opencv-python
Any updates on this issue?
Yes, tf version 1.3 works.
Was anyone able to reproduce the results by pretraining from scratch. Athough I can see better accuracy numbers on all three testsets while pretraining but My results are still 1...
I was comparing the pretrained model you provided with the logfile you provided in another issue: [https://github.com/airsplay/lxmert/issues/22](issue22) The logfile states that the object loss and attribute loss are ``` The...
Hi @airsplay Here is the steps I followed: - Downloaded the code and data for pretraining and finetuning as mentioned in your repo. - Downloaded the [pretrained model](http://nlp.cs.unc.edu/data/model_LXRT.pth) and put...
Hi I was only comparing the train losses between the one from the logfile and the one I printed. But to further bring the point home, I have also printed...
Hi @airsplay I used the command : `bash run/lxmert_pretrain.bash 0,1,2,3 --multiGPU --tiny --loadLXMERT snap/pretrained/model` And That loads the model properly, since Masked LM loss and feature regression are similar to...