Loading parameters and architecture

Gluon
1

Hi,
I have trained a pre-trained model “Resnet50_v2” using different dataset.
I got model architecture in json file and PARAMS file.
Now do I need to create the whole model and pass these parameters into it
or just “gluon.nn.SymbolBlock.imports” will work.
Please help

Here is what I am doing:

[quote=“yatharth, post:1, topic:6032, full:true”]
Hi,
I have trained a pre-trained model “Resnet50_v2” using different dataset.
I got model architecture in json file and PARAMS file.
Now do I need to create the whole model and pass these parameters into it
or just “gluon.nn.SymbolBlock.imports” will work.
Please help

Here is what I am doing:

from mxnet import nd

from mxnet import gluon

from mxnet.gluon import nn

import mxnet as mx

from mxnet.gluon.data.vision import datasets, transforms

from IPython import display

from matplotlib import image

import matplotlib.pyplot as plt

import warnings

import cv2

num_gpus = 0

ctx = [mx.gpu(i) for i in range(num_gpus)] if num_gpus > 0 else [mx.cpu()]

transformer = transforms.Compose([

transforms.ToTensor(),

transforms.Normalize(0.13, 0.31)])

with warnings.catch_warnings():

warnings.simplefilter("ignore")

deserialized_net = gluon.nn.SymbolBlock.imports(r"E:\desktop\x_ray\resnet-symbol.json", ['data'], r"E:\desktop\x_ray\resnet-0001.params", ctx=ctx)

#mnist_valid = datasets.FashionMNIST(train=False)

sample_data = image.imread(r’E:\desktop\x_ray\predict\NORMAL2-IM-1436-0001.jpeg’)

if (len(sample_data.shape))!=3:

sample_data = cv2.cvtColor(sample_data,cv2.COLOR_GRAY2RGB)

#X = sample_data

preds =

#for x in X:

x = transformer(mx.nd.array(sample_data)).expand_dims(axis=0)

pred = deserialized_net(x).argmax(axis=1)

preds.append(pred.astype(‘int32’).asscalar())

_, figs = plt.subplots(1, 1, figsize=(15, 15))

text_labels = [‘normal’,‘pneumonia’]

display.set_matplotlib_formats(‘svg’)

print(text_labels[preds[0]])