search

potterhsu / SVHNClassifier-PyTorch

A PyTorch implementation of Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks (http://arxiv.org/pdf/1312.6082.pdf)
MIT License
198 stars 45 forks source link

Can I have a trained SVHNClassifier? #1

Open changkaizhao opened 6 years ago

changkaizhao commented 6 years ago

Hi, I have no powerful machine to train SVHNClassifier. Can I have a trained SVHNClassifier (model-100.tar)?

fmcalcagno commented 5 years ago

I would love to try the pre-trained model for another project I have, would you be so kind to deliver the weights?

selous123 commented 5 years ago

Author has released the pretrained model in Readme.md [Result]

|steps| |54000|

Batush123 commented 4 years ago

Hi, i had some problem during load_state_dict using the pre trained weights, do you have any solution?

ghost commented 4 years ago

Try this code , resolved the problem for me

    device = torch.device('cpu')
    pretrained_model = torch.load(path_to_checkpoint_file, map_location=device)
    model = Model()
    model.load_state_dict(pretrained_model,strict=False)
BuceaGeorgia commented 6 months ago

I have this problem when trying to run inference on the example image, I used the above code to load the model and the inference script:

  File "torch/nn/functional.py", line 2478, in batch_norm
        _verify_batch_size(input.size())

    return torch.batch_norm(
           ~~~~~~~~~~~~~~~~ <--- HERE
        input, weight, bias, running_mean, running_var, training, momentum, eps, torch.backends.cudnn.enabled
    )
RuntimeError: running_mean should contain 54 elements not 48