Sometimes, there are as much as 10% validation accuracy difference between different seeds. Here’s how I am setting the seeds.
torch.manual_seed(1)
np.random.seed(1)
I am also keeping sklearn’s random state identical.
Sometimes, there are as much as 10% validation accuracy difference between different seeds. Here’s how I am setting the seeds.
torch.manual_seed(1)
np.random.seed(1)
I am also keeping sklearn’s random state identical.
Hello,
What is the size of your validation set and what’s the number of classes you have?