dandelin / ViLT

Code for the ICML 2021 (long talk) paper: "ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision"
Apache License 2.0
1.36k stars 209 forks source link

What could be the reason that the model weights are not updating while finetuning? #82

Open DDXDaniel opened 1 year ago

DDXDaniel commented 1 year ago

I used this command "python run.py with data_root=/data2/dsets/dataset num_gpus=8 num_nodes=1 task_finetune_irtr_f30k_randaug per_gpu_batchsize=4 load_path="weights/vilt_200k_mlm_itm.ckpt"" to fine-tuned VILT on Flickr30k dataset, but I noticed that the model weights are not updating. Below are the VILT partial weight parameters output at each step. What could be the possible reasons causing this issue?

屏幕快照 2023-04-16 下午2 30 03 屏幕快照 2023-04-16 下午2 30 26
DDXDaniel commented 1 year ago

@dandelin I would be very grateful if you could help me with this issue.

dahalbishu commented 8 months ago

@DDXDaniel did you solve this problem? i m facing same problem