search

YushengZhao / TD-STP

Code for MM 22 "Target-Driven Structured Transformer Planner for Vision-Language Navigation"
14 stars 1 forks source link

Do you inialize your designed network with the pretrained weights? #2

Closed CrystalSixone closed 2 years ago

CrystalSixone commented 2 years ago

Hi, Dear YuSheng! Thanks for your amazing work. I am interested in your work and I'm doing some research based on the VLN-HAMT as well. I am wondering if you initialize your designed network by using model_step_22000.pt provided by HAMT? And if so, did you do experiments directly on the fine-tune stage? It would be my great pleasure if you could reply to me. Thank you again. Best wishes. :)

YushengZhao commented 2 years ago

Yes, we use the pretrained weights of HAMT as initialization in our experiments.