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
Hi, I noticed in the paper of BLIP-2 you provided(in Table 2.) the Total Params metric of different models, but I wonder if these values derive from specific statistics? If so, how?
Actually, I used the following code to calculate the param counts, but I got 3218346496 for BLIP-2 ViT-L FlanT5XL, a little less than 3.4B that you reported.
num = 0
for param in model.parameters():
num += param.numel()
print(num)
The text was updated successfully, but these errors were encountered:
Hi, I noticed in the paper of BLIP-2 you provided(in Table 2.) the
Total Params
metric of different models, but I wonder if these values derive from specific statistics? If so, how?Actually, I used the following code to calculate the param counts, but I got
3218346496
for BLIP-2 ViT-L FlanT5XL, a little less than 3.4B that you reported.The text was updated successfully, but these errors were encountered: