callsys / textvr Goto Github PK

Hello, i am interested in your excellent work(TextVR). But I can't run the code successfully

Failed in Inference TextVR with given weights ckpt/textvr.pth

the result is below:
[t2v_metrics]TextVR epoch 0, R@1: 0.0, R@5: 0.1, R@10 0.2, R@50 1.1MedR: 1390, MeanR: 1369.0
[v2t_metrics]TextVR epoch 0, R@1: 0.0, R@5: 0.0, R@10 0.3, R@50 1.7MedR: 1376.5, MeanR: 1366.4

Failed in Training TextVR with given config file configs/TextVR_fusion.json

I try to train the model, but failed again.
[t2v_metrics]TextVR epoch 0, R@1: 0.0, R@5: 0.1, R@10 0.2, R@50 1.1MedR: 1390, MeanR: 1369.0
[v2t_metrics]TextVR epoch 0, R@1: 0.0, R@5: 0.0, R@10 0.3, R@50 1.7MedR: 1376.5, MeanR: 1366.4
The loss_0 is decrease to 0.75, and the val_loss_0 is 10.2, i wonder if my config is wrong or other reasons(over fitting?).

Problem about the given example of sim_matrix.npy

i use the given sim_matrix.npy to get the result, but the result is below:
[t2v_metrics]TextVR epoch 0, R@1: 0.0, R@5: 0.1, R@10 0.2, R@50 1.1MedR: 1390, MeanR: 1369.0
[v2t_metrics]TextVR epoch 0, R@1: 0.0, R@5: 0.0, R@10 0.3, R@50 1.7MedR: 1376.5, MeanR: 1366.4
the code is below:

sims = np.load('/path/sim_matrix.npy')# 本地读取,using the given example of sim_matrix.npy
res = metric(sims) # , query_masks=mask
verbose(epoch=0, metrics=res, name="", mode=metric_name)

Thanks for your patience,i wonder if my config is wrong or other reasons.