zjut-multimediaplus/dadh

source code for "Deep adversarial discrete hashing for cross-modal retrieval"

dadh's Issues

Error when testing

I am getting an error when testing the mir dataset.
These are the contents of my data directory, with data downloaded as per the README

> ls data
FLICKR-25K.mat  README.md  imagenet-vgg-f.mat

After training (successfully) with

python3 main.py train --flag 'mir' --bit 16

I tried testing with

python3 main.py test --flag 'mir' --bit 16

but got the following error:

Configuration:
        load_model_path: None
        pretrain_model_path: ./data/imagenet-vgg-f.mat
        vis_env: main
        vis_port: 8097
        flag: mir
        batch_size: 128
        image_dim: 4096
        hidden_dim: 8192
        modals: 2
        valid: True
        valid_freq: 1
        max_epoch: 300
        bit: 16
        lr: 5e-05
        device: cuda:0
        alpha: 10
        gamma: 1
        beta: 1
        mu: 1e-05
        lamb: 1
        margin: 0.4
        dropout: False
Traceback (most recent call last):
  File "main.py", line 335, in <module>
    fire.Fire()
  File "/home/clusterusers/mceccarello/.local/lib/python3.8/site-packages/fire/core.py", line 138, in Fire
    component_trace = _Fire(component, args, parsed_flag_args, context, name)
  File "/home/clusterusers/mceccarello/.local/lib/python3.8/site-packages/fire/core.py", line 463, in _Fire
    component, remaining_args = _CallAndUpdateTrace(
  File "/home/clusterusers/mceccarello/.local/lib/python3.8/site-packages/fire/core.py", line 672, in _CallAndUpdateTrace
    component = fn(*varargs, **kwargs)
  File "main.py", line 260, in test
    generator = GEN(opt.image_dim,
TypeError: __init__() missing 1 required positional argument: 'class_dim'

Is there any way I can fix it?

about paper

hello sir,Thanks for your code,I want to know more about your paper,but i can't download,would you mind provide the full paper link or something?thank you very much.and my email is [email protected]

如何运行代码？急

你好，我在拥有3090显卡的服务器上跑这份代码，每次都是显示“已杀死”，为什么进程会被杀死呢？

Problem with testing

Hi,

I came across an error when running "python main.py test --flag 'nus' --bit 16" as below:

RuntimeError: Error(s) in loading state_dict for GEN:
Missing key(s) in state_dict: "image_module.3.weight", "image_module.3.bias", "image_module.4.running_mean", "image_module.4.running_var", "image_module.6.weight", "image_module.6.bias", "image_module.7.weight", "image_module.7.bias", "image_module.7.running_mean", "image_module.7.running_var", "text_module.3.weight", "text_module.3.bias", "text_module.4.running_mean", "text_module.4.running_var", "text_module.6.weight", "text_module.6.bias", "text_module.7.weight", "text_module.7.bias", "text_module.7.running_mean", "text_module.7.running_var".
Unexpected key(s) in state_dict: "image_module.9.weight", "image_module.9.bias", "image_module.9.running_mean", "image_module.9.running_var", "image_module.9.num_batches_tracked", "image_module.5.weight", "image_module.5.bias", "image_module.5.running_mean", "image_module.5.running_var", "image_module.5.num_batches_tracked", "image_module.8.weight", "image_module.8.bias", "text_module.9.weight", "text_module.9.bias", "text_module.9.running_mean", "text_module.9.running_var", "text_module.9.num_batches_tracked", "text_module.5.weight", "text_module.5.bias", "text_module.5.running_mean", "text_module.5.running_var", "text_module.5.num_batches_tracked", "text_module.8.weight", "text_module.8.bias".
size mismatch for image_module.4.weight: copying a param with shape torch.Size([4096, 8192]) from checkpoint, the shape in current model is torch.Size([4096]).
size mismatch for text_module.4.weight: copying a param with shape torch.Size([4096, 8192]) from checkpoint, the shape in current model is torch.Size([4096]).

This occurs after I train with "bash run_nus.sh 16" script.
Testing on "mir" dataset is OK without this error.

Could you help to have a check?

Thanks.