0xTaki.eth

icon indicating a website link https://github.com/KAndHisC icon indicating an email [email protected]

icon company UAlberta icon location Edmonton, Alberta, CA icon location Coder&Hacker。

Results 3 comments of 0xTaki.eth

Influence of batch size of training convergence

> @KeremTurgutlu I found the same question when I want to train CLIP with 8 GPUs. I found that if my total batch size is 128, then pytorch will split...

Implementation details in few-shot ImageNet evaluation

I tried run testing on CIFAR100 following by `README.md (Zero-Shot Prediction)` and also can't achieve the performance in paper. did you solve this problem?

How is the dataset collected?

> This is mentioned in paper: "The base query list is all words occurring at least 100 times in > the English version of Wikipedia. This is augmented with bi-grams...