Incubator-mxnet: UserWarning: input name is not found in symbol.list_arguments()

Created on 11 Apr 2017 · 4Comments · Source: apache/incubator-mxnet

I ran NN on raspberrypi. Please ignored OOM prolem(std::bad_alloc), I wanna fix userWarning. Thanks a lot!

Error Messages

(py-perfblas) yuanshuai@raspberrypi:~/code/mxnet_inference/vgg-16-official-mxnet $ python run_inference.py                                                                                                                                                                                                             
[16:42:22] src/nnvm/legacy_json_util.cc:190: Loading symbol saved by previous version v0.8.0. Attempting to upgrade...
[16:42:22] src/nnvm/legacy_json_util.cc:198: Symbol successfully upgraded!
/home/yuanshuai/pyEnv/py-perfblas/local/lib/python2.7/site-packages/mxnet-0.9.5-py2.7.egg/mxnet/module/base_module.py:52: UserWarning: You created Module with Module(..., label_names=['softmax_label']) but input with name 'softmax_label' is not found in symbol.list_arguments(). Did you mean one of:
        data
        prob_label
  warnings.warn(msg)
/home/yuanshuai/pyEnv/py-perfblas/local/lib/python2.7/site-packages/mxnet-0.9.5-py2.7.egg/mxnet/module/base_module.py:64: UserWarning: Data provided by label_shapes don't match names specified by label_names ([] vs. ['softmax_label'])
  warnings.warn(msg)
terminate called after throwing an instance of 'std::bad_alloc'
  what():  std::bad_alloc
Aborted

Code

import os, sys 
import urllib
import mxnet as mx
import time

# Added by yuanshuai
sys.path.append('/mnt/sdcard/lizhihao/install/opencv-3.1.0/lib/python2.7/dist-packages')

with open('synset.txt', 'r') as f:
    synsets = [l.rstrip() for l in f]

# Edited by yuanshuai
sym, arg_params, aux_params = mx.model.load_checkpoint('caffenet-upgrade', 1)
#arg_params['prob_label'] = mx.nd.array([0])
#mx.model.bind(data_shapes=[('data', (1, 3, 224, 224))], label_shapes=[('prob_label', (1,))] )

mod = mx.mod.Module(symbol=sym, context=mx.cpu(0))
input_shape = (224, 224)
predict_times = 1 
mod.bind(for_training=False, data_shapes=[('data', (1, 3, input_shape[0], input_shape[1]))])
mod.set_params(arg_params, aux_params)

import cv2 
import numpy as np
from collections import namedtuple
Batch = namedtuple('Batch', ['data'])

def get_image(url, show=True):
    get_image_start = time.time()
    filename = url.split("/")[-1]
    img = cv2.imread(filename)
    if img is None:
        print('failed to download ' + url)
    print "get image take {0}".format((time.time()-get_image_start)/100)
    return filename

def predict(filename, mod, synsets, input_shape, predict_times=1):
    predict_start = time.time()
    img = cv2.cvtColor(cv2.imread(filename), cv2.COLOR_BGR2RGB)
    if img is None:
        return None
    img = cv2.resize(img, (224, 224))
    img = np.swapaxes(img, 0, 2)
    img = np.swapaxes(img, 1, 2)
    img = img[np.newaxis, :]

    start = time.time()
    for i in range(predict_times):
         mod.forward(Batch([mx.nd.array(img)]))
         #print 'take time {0}'.format(time.time()-start)
         prob = mod.get_outputs()[0].asnumpy()
         prob = np.squeeze(prob)
         a = np.argsort(prob)[::-1]
    print "each image take {0}".format((time.time()-start)/predict_times)
    #for i in a[0:5]:
    #    print('probability=%f, class=%s' % (prob[i], synsets[i]))

url = './Cat-hd-wallpapers.jpg'
predict(url, mod, synsets, input_shape, predict_times = predict_times)

Source

ysh329

Most helpful comment

i'm having the same issue loading vgg16 model and doing a prediction. Where do you suggest I provide the dummy 'softmax_label'? In the model json file? I'm totally new so not familiar with that format.

temojin on 3 Jun 2017

👍2

All 4 comments

By Default MxNet network looks for an input with the name 'softmax_label' as the target variable for training. When you built your network by loading the check-point, it was expecting an input 'softmax_label' which you did not provide that hence the warning.
Take a look here:
https://github.com/dmlc/mxnet/blob/master/python/mxnet/module/module.py#L31

A probable remedy could be to just supply a dummy variable with the name 'softmax_label'