Ev

when the system is idle and not processing, shouldn't somehow python not use the whole GPU memory ? it is a useful feature

you are probably trying to process video level features whereas his code I think is for frame level features, I don't know if it would work by changing --feature_names="rgb,audio" to...

I am using Ubuntu 16.04 LTS but I traced the problem to be that I tried to evaluate and inference tfrecords of different level features than the model. I wanted...

did you choose little-endian or big-endian for the audioset feature extraction ? and why ?

``` class Net(nn.Module): def __init__(self, video_modality_dim, text_dim, audio_cluster=8, text_cluster=32): super(Net, self).__init__() self.audio_pooling = NetVLAD(feature_size=video_modality_dim['audio'][1], cluster_size=audio_cluster) self.text_pooling = NetVLAD(feature_size=text_dim, cluster_size=text_cluster) self.mee = MEE(video_modality_dim, self.text_pooling.out_dim) def forward(self, text, video, ind, conf=True): aggregated_video...

How do you perform exactly the max pooling ? I guess I have to do it myself (after trying all day to see what's wrong with the model )

the MEE block accepts input of size N X M , where N is the batch size and M is the product of (feature size X cluster size) derived from...

I also have the same problem when I try to execute the inference code, to be precise the output csv file only has 1 or 2 videos with labels then...

the question is why is it trying to look for tfrecords in that particular folder "/YT8M/youtube-8m/features/"? I can't trace where this happens so I can delete it

num_examples_processed initializes with 0. out_file.write("VideoId,LabelConfidencePairs\n") doesn't need to be printed, check only if the CSV file contains those strings in two columns. you need the threads to perform the evaluations...