Step 8: Parallel Computing
Running computations and processing data in parallel.
Hub enables you to easily run computations in parallel and significantly accelerate your data processing workflows. This example primarily focuses on parallel dataset uploading.
Parallel compute using Hub has two core steps:
- 1.Define a function or pipeline that will run in parallel and
- 2.Evaluate the function using the appropriate inputs and outputs.
Defining the parallel computing function
The first step is to define a function that will run in parallel by decorating it using
@hub.compute. In the example below, file_to_hub converts data from files into hub format, just like in Step 2: Creating Hub Datasets Manually. If you have not completed Step 2, please download and unzip the example image classification dataset below:animals.zip
338KB
Binary
animals dataset
1
import hub
2
from PIL import Image
3
import numpy as np
4
import os
5
6
@hub.compute
7
def file_to_hub(file_name, sample_out, class_names):
8
## First two arguments are always default arguments containing:
9
# 1st argument is an element of the input iterable (list, dataset, array,...)
10
# 2nd argument is a dataset sample
11
# Other arguments are optional
12
13
# Find the label number corresponding to the file
14
label_text = os.path.basename(os.path.dirname(file_name))
15
label_num = class_names.index(label_text)
16
17
# Append the label and image to the output sample
18
sample_out.labels.append(np.uint32(label_num))
19
sample_out.images.append(hub.read(file_name))
20
21
return sample_out
Copied!
In all functions decorated using
@hub.compute, the first argument must be a single element of any input iterable that is being processed in parallel. In this case, that is a filename file_name, because file_to_hub reads image files and populates data in the dataset's tensors. The second argument is a dataset sample
sample_out, which can be operated on using similar syntax to dataset objects, such as sample_out.append(...), sample_out.extend(...), etc.The function decorated using
@hub.compute must return sample_out, which represents the data that is added or modified by that function.Executing the parallel computation
To execute the parallel computation, you must define the dataset that will be modified.
1
ds = hub.empty('./animals_hub_transform') # Creates the dataset
Copied!
Next, you define the input iterable that describes the information that will be operated on in parallel. In this case, that is a list of files
files_list:1
# Find the class_names and list of files that need to be uploaded
2
dataset_folder = './animals'
3
4
class_names = os.listdir(dataset_folder)
5
6
files_list = []
7
for dirpath, dirnames, filenames in os.walk(dataset_folder):
8
for filename in filenames:
9
files_list.append(os.path.join(dirpath, filename))
Copied!
You can now create the tensors for the dataset and run the parallel computation using the
.eval syntax. Pass the optional input arguments to file_to_hub and skip the first two default arguments file_name and sample_out. The input iterable
files_list and output dataset ds is passed to the .eval method as the first and second argument respectively.1
with ds:
2
ds.create_tensor('images', htype = 'image', sample_compression = 'jpeg')
3
ds.create_tensor('labels', htype = 'class_label', class_names = class_names)
4
5
file_to_hub(class_names=class_names).eval(files_list, ds, num_workers = 2)
Copied!
Additional parallel computing use cases such as dataset transformations can be found in this tutorial.
1
Image.fromarray(ds.images[0].numpy())
Copied!
Congrats! You just created a dataset using parallel computing! 🎈