Hello, I enjoyed reading your insightful work.
During my experiments, I noticed that the FLOPs I calculated differ significantly from the values reported in your paper.
In the paper, you mention 3.1G FLOPs, but when I measured it using fvcore’s flop_count_table, I obtained about 3.8G. I am wondering whether I might have miscalculated the FLOPs, or if a different measurement protocol was applied. Could you please clarify the method you used to arrive at 3.1G FLOPs in the paper?
#2
When I applied the method you suggested in #2, it resulted in an even larger value.
Thank you.
Hello, I enjoyed reading your insightful work.
During my experiments, I noticed that the FLOPs I calculated differ significantly from the values reported in your paper.
In the paper, you mention 3.1G FLOPs, but when I measured it using fvcore’s flop_count_table, I obtained about 3.8G. I am wondering whether I might have miscalculated the FLOPs, or if a different measurement protocol was applied. Could you please clarify the method you used to arrive at 3.1G FLOPs in the paper?
#2
When I applied the method you suggested in #2, it resulted in an even larger value.
Thank you.