add googlenet results?

[hengck23]

Thank you very much for the results. Is there any plan to add results of gogglenet v1,v2,v3,v4? or even squeeze and caffenet (alexnet with lrn layer replaced)?

[ottolu]

+1.

I think GoogleNet is still very well used in many cases. It would be great if Justin could bring them in.

[jcjohnson]

CaffeNet would be pretty easy to add since it can be loaded into torch with loadcaffe.

GoogLeNet v1 is doable using soumith/inception.torch and v3 from Moodstocks/inception-v3.torch; I'm not aware of any Torch port of v2 or v4.

I will try and add these benchmarks this weekend.

[vimalthilak]

@jcjohnson Have you had a chance to benchmark Inception v1 (GoogLeNet) and Inception v3? I'm looking forward to your results if/when they become available. Thanks

[jcjohnson]

@vimalthilak GoogleNet v1 is easy, but I ran into a snag with v3; they don't provide dumped weights and their weight-dumping script depends on TensorFlow; I opened an issue but haven't gotten a response (https://github.com/Moodstocks/inception-v3.torch/issues/7).

I wasted several hours trying to build TensorFlow from source to support CUDA 8.0 RC so I could use my Pascal Titan X, but couldn't get it to work and gave up.

If someone can provide the dumped weights for GoogLeNet v3 then I can add both benchmarks tomorrow.

[jcjohnson]

@vimalthilak I've added Inception-v1 results.

[lymhust]

@jcjohnson Hi, I have the dumped Torch file of inception v3 (https://drive.google.com/drive/folders/0B-XjOBwS-l8sRXRxRW45U3kybjQ). Hope that you can add it as the benchmark.