Error on loading pre-trained model

[robi56]

I am trying to work on transfer learning using Mobilenet Pretrained model

model = nets.MobileNet35v2(inputs, is_training=True, classes=2)
with tf.Session() as sess:
nets.pretrained(model)

But this gives the following error

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-9-ba2d4690d272> in <module>
      8 
      9 with tf.Session() as sess:
---> 10     nets.pretrained(model)
     11 
     12     for i in range(0,len(epochs)):

~/.conda/envs/tf-gpu/lib/python3.6/site-packages/tensornets/pretrained.py in assign(scopes)
     56         model_name = parse_scopes(scope)[0]
     57         try:
---> 58             __load_dict__[model_name](scope)
     59         except KeyError:
     60             try:

~/.conda/envs/tf-gpu/lib/python3.6/site-packages/tensornets/pretrained.py in load_mobilenet35v2(scopes, return_fn)
    550         file_hash='cf758f7f8024d39365e553ec924bb395')
    551     values = parse_weights(weights_path)
--> 552     return return_fn(scopes, values)
    553 
    554 

~/.conda/envs/tf-gpu/lib/python3.6/site-packages/tensornets/pretrained.py in _assign(scopes, values)
    108 
    109     for scope in scopes:
--> 110         sess.run(pretrained_initializer(scope, values))
    111 
    112 

~/.conda/envs/tf-gpu/lib/python3.6/site-packages/tensornets/utils.py in pretrained_initializer(scope, values)
    324                       'Never mind if you are trying to load stem layers only.')
    325 
--> 326     if scope.dtype == tf.float16:
    327         ops = [weights[0].assign(np.asarray(values[0], dtype=np.float16))]
    328         for (w, v) in zip(weights[1:-2], values[1:-2]):

AttributeError: 'str' object has no attribute 'dtype'

[taehoonlee]

@robi56, The nets.pretrained() will be depreciated. Please use sess.run(model.pretrained()) instead.