Why can you achieve such high training speed?
starYellow966 opened this issue · 7 comments
Excuse me, After reading your thesis and code, i have benefited a lot. Again, there are some doubts, I hope you can help answer
1）Is the training speed displayed by REAME.md (shown below) based on CAMRa2011?
AGREE at embedding size 32, run Iteration:30, NDCG and HR at 5
...
User Iteration 10 [449.8 s]: HR = 0.6216, NDCG = 0.4133, [1.0 s]
Group Iteration 10 [471.9 s]: HR = 0.5910, NDCG = 0.4005, [23.0 s]
2）If so, do you run with gpu or cpu? I use your code directly and find that I can't reach this speed.
Looking forward for your response, thank you
Hi, I'm glad to help you.
For the Question 1, the answer is yes. I displayed our result on the CAMRa2011. And the speed is our training speed.
For the Question 2, the answer is cpu, we run our code on the CPU I7-6800K.
And, I think it would help you run the code more quickly if you use the gpu.
Thank you very much for your answer.
Hi, I'm glad to help you.
For the Question 1, the answer is yes. I displayed our result on the CAMRa2011. And the speed is our training speed.
For the Question 2, the answer is cpu, we run our code on the CPU I7-6800K.
And, I think it would help you run the code more quickly if you use the gpu.
Hello, I use my computer train this model, but it can not arrive this speed, what should I to do to solve this problem. (my computer cpu is I7 7700HQ, NVIDIA 1060)
It may be the os version? I'm not sure, but our code run in the ubuntu 16.04. BTW, It would speed your training time by increasing the size of batch and utilizing the GPU.
It may be the os version? I'm not sure, but our code run in the ubuntu 16.04. BTW, It would speed your training time by increasing the size of batch and utilizing the GPU.
When I used GPU to train in windows10 system, the same situation still occurred, and I was stuck in AGREE at embedding size 32, run Iteration:30, NDCG and HR and RECALL at top-5. What can I do to try to solve this problem? (change system or use better GPU )
I'm not sure the accurate time, but I remembered the first epoch would cost about 2000s (maybe, I'm not sure...). This is mainly because we need some time to process the data for feed into the model. And when the data preprocess is over, the run time may be 1000s each epoch.
I have the same problem. The code runs very slowly, and there is no result after running for a day. How did you solve it later?