chenusc11
Hi, Just wondering do you still have a plan to share the pre-trained model? Thx @HRLTY
@yuanyuanli85 Hi, thanks for your code. I'm wondering when you said "We have tried vgg16 and alexnet, only 1% loss in test accuracy (fc'weights to 16bits and conv's weights to...
Hi, I encountered similar issue here. Sometimes it randomly crushes. However, I never met the same issue without using AMP+tiny-cuda-nn. Even if the architecture is the same. My environment is...
@ashawkey Thanks for the reply. No I didn't set detect anomaly myself. It seems to be happening when using tcnn's sigma_net. WIth the exactly same architecture of PyTorch, I didn't...
Thanks for the suggestion. Will do.
> @chenusc11 Have you solved the problem? I meet the same issue now. Hi, unfortunately I couldn't solve the problem.
@taylorlu @ashawkey I have solved my issue by removing the torch AMP package. I talked to the tiny-cuda-nn's author on one occasion recently, and he said the AMP is not...
The reason is probably you have encountered overflow/underflow in your pipeline somewhere in the next steps of tcnn. The float16’s max value from the tcnn library is probably 65504-ish. If...
@chaoxianluo I tried to reimplement the VLocNet myself for the past few weeks and I found the same phenomenon with sx and sq kept dropping. After some experiment and comparison...
> I also experienced the error-accumulating issue and according to my simple mind, this couldn't be avoided in the VLocNet. I haven't found any module in the model which might...