Anything wrong about the result?

[liulaha]

Here is the result of GZSL on CUB: [0.0] novel=0.0190, seen=0.0024, h=0.0042 , loss=5.0547 [1.0] novel=0.0067, seen=0.0037, h=0.0048 , loss=4.8157 [2.0] novel=0.0057, seen=0.0051, h=0.0054 , loss=4.6095 [3.0] novel=0.0040, seen=0.0123, h=0.0060 , loss=4.4239 [4.0] novel=0.0040, seen=0.0143, h=0.0063 , loss=4.2417 [5.0] novel=0.0037, seen=0.0199, h=0.0062 , loss=4.0880 [6.0] novel=0.0040, seen=0.0214, h=0.0067 , loss=3.9427 [7.0] novel=0.0040, seen=0.0220, h=0.0068 , loss=3.7959 [8.0] novel=0.0037, seen=0.0220, h=0.0063 , loss=3.6719 [9.0] novel=0.0047, seen=0.0231, h=0.0078 , loss=3.5571 [10.0] novel=0.0047, seen=0.0240, h=0.0078 , loss=3.4314 [11.0] novel=0.0047, seen=0.0250, h=0.0079 , loss=3.3595 [12.0] novel=0.0047, seen=0.0249, h=0.0079 , loss=3.2751 [13.0] novel=0.0043, seen=0.0254, h=0.0074 , loss=3.1913 [14.0] novel=0.0043, seen=0.0260, h=0.0074 , loss=3.0842 [15.0] novel=0.0040, seen=0.0275, h=0.0070 , loss=3.0107 [16.0] novel=0.0030, seen=0.0281, h=0.0054 , loss=2.9473 [17.0] novel=0.0027, seen=0.0281, h=0.0049 , loss=2.8864 [18.0] novel=0.0027, seen=0.0286, h=0.0049 , loss=2.8150 [19.0] novel=0.0027, seen=0.0286, h=0.0049 , loss=2.7325 [20.0] novel=0.0030, seen=0.0281, h=0.0054 , loss=2.7147 [21.0] novel=0.0030, seen=0.0281, h=0.0054 , loss=2.6553 [22.0] novel=0.0027, seen=0.0292, h=0.0049 , loss=2.5960 0.00489604612812 I think something must be wrong so that I get such result.

[edgarschnfld]

Which setup did you use? The one below? _python single_experiment.py --dataset CUB --numshots 0 --generalized True If you use the data and packages as described in the readme, you should not get such a result.

[liulaha]

Thanks for your reply. Here is the setup: parser.add_argument('--dataset',default='CUB') parser.add_argument('--num_shots',type=int,default=0) parser.add_argument('--generalized',type=bool,default=True)

[liulaha]

Should I use the same version of the required packages in the readme? For example, my torch is 0.4.0 while torch==0.4.1 in the readme.

[Hanzy1996]

I come across the same problem. How do you fix it?

[liulaha]

@Hanzy1996 try to configure your environment as the same as the author proposed exactly

[Hanzy1996]

Thanks so much! I will try.