You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In this experiment, I used the default configuration. Theoretically, num-shot should be 1, but the log is recorded as -1.DATASET: ALL_AS_UNLABELED: False CIFAR_C_LEVEL: 1 CIFAR_C_TYPE: NAME: UCF101 NUM_LABELED: -1 NUM_SHOTS: -1 ROOT: /path/to/CoOp-main/data SOURCE_DOMAINS: () STL10_FOLD: -1 SUBSAMPLE_CLASSES: all TARGET_DOMAINS: () VAL_PERCENT: 0.1 So what is the true value of num-shot in this experiment?
How to set up multi card training? Where is the entry?
The text was updated successfully, but these errors were encountered:
I want to reproduction lpclip experiment. extract feature by
sh feat_extractor.sh
. I met the following questions,num-shot
? I should use the option--num-shot
orCoOp/lpclip/feat_extractor.py
Line 181 in ff61507
DATASET.NUM_SHOTS
CoOp/scripts/coop/main.sh
Line 30 in ff61507
num-shot
should be 1, but the log is recorded as -1.DATASET: ALL_AS_UNLABELED: False CIFAR_C_LEVEL: 1 CIFAR_C_TYPE: NAME: UCF101 NUM_LABELED: -1 NUM_SHOTS: -1 ROOT: /path/to/CoOp-main/data SOURCE_DOMAINS: () STL10_FOLD: -1 SUBSAMPLE_CLASSES: all TARGET_DOMAINS: () VAL_PERCENT: 0.1
So what is the true value ofnum-shot
in this experiment?The text was updated successfully, but these errors were encountered: