alexhex7 / drml_pytorch Goto Github PK

I think if the output of sigmoid() function >0.5 , the sample can be classified into 1. So Could you tell me why you use thresh = 0.8.

Hi,

Thanks very much for sharing your codes. May I ask why the test loss shown in your logs folder is increasing while the training loss is decreasing? Thanks!

Could you tell me how to select the pictures and what information is included in .txt file?

Really Sorry but just want to ask that how to use checkpoints for this model? as i try to run the model with disfa or other bigger dataset it will came out the error about "CUDA run out of memory" ,QAQ, i seach the way to solve it but i have no idea to use the supported way of checkpoints in this model.
Any ways, thanks!!

Hello and thank you very much for sharing this code.

I have a Cuda error in:

                          ```     File: DRML_pytorch-master\lib\region_layer.py", line 44, in forward
                                 grid = self.region_layers[module_name](grid.contiguous()) + grid     ```

The error is:

RuntimeError: CUDA out of memory. Tried to allocate 184.00 MiB (GPU 0; 6.00 GiB total capacity; 3.79 GiB already allocated; 100.36 MiB free; 241.12 MiB cached)

I decreased the batch size in the config file but still get this error. could you please give me some guides?

thanks again and greetings

What should the data look like? Should it only contain the face_images folder, train_info.txt, and test_info.txt? If there is no AU information available, what would be the default setting, and which AUs will be used respectively?