configure8

Startups of The Year 2023 Winners: North America 

the global IT spending is expected to grow to a staggering $3.8 trillions in 2019. So, it’s no wonder if your great idea of an app or a software product is a part of this statistics. And if want to execute it quickly, efficiently, you’re probably looking into IT outsourcing companies to find your perfect match. According to Gartner, And from here the world is your oyster! The options to choose from are more than enough. In fact, the choice is so vast that you might feel overwhelmed. But don’t worry. Essentially, the are two main questions you need to answer — where? and who? Let’s start from where. Today, you can find a web developer virtually in any country or a part of the world. Let’s compare the main destinations for software outsourcing in terms of billing rates and the ease of cooperation. And see why you should consider Eastern European partner for your next project. Top destinations for IT outsourcing companies The image above shows top IT destinations and key countries in each region as well as a rough hourly rate. Now let’s look more closely into global software outsourcing rates of the last year for each job title and how the prices compare to the United States. Eastern Europe here looks like a golden middle. It has been a popular IT outsourcing destination for many years. Which is not surprising, it has a rich talent pool, an impressive engineering expertise and a solid educational background. On top of that, this region seems appealing because of its proximity to Western Europe, decent level of English and a good cultural compatibility. This is a sound choice for companies that are after a good quality of work rather than looking for the cheapest solution. Now that we’ve established the “Where”, let’s look into the ‘Who’ Top 10 IT Outsourcing companies in Eastern Europe Below is the list of reliable software development partners that have developing centers in Eastern Europe and have been recognized by as top firms in Eastern Europe. Clutch.co Perfectial https://perfectial.com/ Lviv, Kyiv, Ivano-Frankivsk, Ukraine Location: Cloud & SaaS applications, Android & iOS mobile applications, Artificial Intelligence & Machine Learning, R&D services, Blockchain development, Enterprise development Services: Perfectial is a business-oriented product design and development company committed to helping businesses kickstart, develop, and succeed. Perfectial has implemented innovative ideas that allowed businesses — large and small — to scale, evolve, and prosper in a range of business domains. Perfectial’s competence in working with modern platforms and frameworks is recognized and highly valued within the U.S. and European markets. The company has over 200 professionals in-house, including project managers, solutions architects, .Net, Java, Ruby on Rails, Python and Node.js developers, UI/UX design and Quality Assurance teams. 2. Ingemark https://www.ingemark.com/ Zagreb, Croatia Location: Software development, Cloud consulting Services: Ingemark is a custom software development agency trusted by some of the largest companies in Southeast Europe, as well as a number of companies from Western Europe, Middle East and the U.S. Founded in 1990., custom software development has been company’s core business since 2006.Ingenmark can help with everything from software design and development to post-launch support and maintenance. 3. SoftwareMill https://softwaremill.com/ Warsaw, Poland Location: Big Data & Machine Learning, Web development, Enterprise solutions, UI/UX design Services: SoftwareMill is a Poland-based software and consultancy company that develops customized IT systems, especially big data, business intelligence, and machine learning solutions. The company specializes in Java, Scala, Akka, the Lightbend Reactive Platform, big data / ML frameworks such as Kafka, Spark, and TensorFlow, SQL & NoSQLdatabases: PostgreSQL, Cassandra, Mongo. 4. CodeWave https://codewave.eu/ Wroclaw, Poland Location: Solutions for creative agencies, Web-based systems and mobile apps, Cloud solutions, Services: CodeWave has been trading since 2008, at first delivering web development and server administration services to local advertising agencies in Poland, later expanding to global markets (EU & US) and delivering complete IT project outsourcing services — providing roles of dev-ops, database administrators, backend and frontend web and mobile applications developers. 5. Sloboda Studio https://sloboda-studio.com/ : Kharkiv, Ukraine Location Back-end development, Front-end development, Project management, Quality assurance Services: Sloboda Studio is a company providing custom web development services. Founded in 2010, they’ve completed 92 projects for the clients from Western European countries. Among company’s clients are Marketplaces, Fintech and Adtech products. Working with them, they apply Agile methodologies in order to ensure the timely delivery and the high quality of the product. 6. GP Solutions https://gpsolutions.com/ Minsk, Belarus Location: Dedicated development team, Web application development, Enterprise software development, Mobile application development, UI/UX design Services: GP solutions is an established European software development house with a proven track record of delivering exceptional quality in custom software development, IT consulting and support. The Headquarters of GP Solutions is located in Munich, Germany, the Sales Office is in Miami Beach, USA and the Development Center is in Minsk, Belarus. GP Solutions has earned the trust of some of the world’s leading brands and most discerning clients in the , , . e-banking travel & hospitality information technology 7. Code Runners https://www.code-runners.com/ Sofia, Bulgaria Location: Software development, Data management, Product development, Machine Learning & AI Services: Code Runners build online products and help implement digital transformation projects around the globe. The company has been in cooperation with many small startups and large business clients. They have experience in web design, web and mobile applications, database management systems and various specific projects aligned with our customers’ wishes. The company is using the newest technologies on the market, which guarantee that the products they build will look and feel equally good on different devices and will remain useful in years to come, besides being easy to maintain and configure. 8. Pulilab http://www.pulilab.com/ Budapest, Hungary Location: Hardware prototyping, Ionic App Development, iOT integration, Drone Technology, App design & development, Digital product creation & conceptualization, Tech specification, Architecture & software design, Web & mobile application development, Big-data API integrations Services: Pulilab is an international creative software and digital media agency that offers innovative solutions crafted by talented developers and designers in Budapest. Their mission is to help companies design and build software products for web, mobile and connected devices. 9. Vega IT Sourcing https://www.vegaitsourcing.rs/ Novi Sad, Serbia Location: Software product development, Website development, Dedicated development team, Embedded software development, Mobile apps development, Front-end development Services: Vega IT is a company that has been developing software solutions for 10 years for foreign markets. The company was founded in 2008 and is still led by its original founders. Besides Software Product Development, their team of more than 150 experienced developers can help you build Enterprise software solutions, CMS Websites, Mobile applications. Moreover, they can help with maintenance and support, Salesforce and Research and Development. 10. Itera https://itera.no/en/ Kyiv, Ukraine, Bratislava, Slovakia Location: Services: Application management, UX design, Business design, DevOps Itera is a specialist in creating digital business, with communication, technology and innovation as core competency tools. Itera delivers projects and services in cross-functional teams to Nordic organisations that see the instrumental contribution that innovation, efficient communication and smart utilization of technology can make to achieve their goals. Itera’s core sectors are banking and insurance, public, healthcare, the service industry, energy and utility. Itera is headquartered in Oslo, and has about 500 employees across the Nordics and in Ukraine and Slovakia. All things considered, Eastern Europe does look like a golden mean. You can find a reliable and transparent partner in any and every country of the region. The billing rates are quite sensible while companies offer a wide range of software development services that will suit everyone’s taste.

Top IT Outsourcing Companies in Europe. Why You Should Consider Outsourcing to Europe

I previously used the Coze platform (with zero coding experience) to build a Chat Translator AI that can be used on Telegram, WhatsApp, Discord, and others. Here, I will show you How to Configure/Integrate the Chatbot built with Coze into the Telegram chat. Requirements Telegram account
BotFather Step 1 Create a Telegram account. And if you already have an account with Telegram, you can use your existing account. Step 2 Search BotFather in your Telegram account search bar, and click Start as shown here. Type this command /newbot and send. You will see a message prompt to choose/set a name for the bot (which can be modified later) and a username. Note: Copy and save the token from the BotFather conversation. Step 3 Go back to your Coze account, and click on the exact chatbot agent from Workspace to open it. On the Develop page, click on Publish. After you have "Publish", it will display a page to select any platform to "publish to". 'Click' on Configure for Telegram since we are integrating the Translator Bot into Telegram. Enter the Access Token copied from BotFather on the pop-up page, and 'click' Save. Finally, now 'click' on Chat Now or Copy Agent Link, and open it in your Telegram account to use the Chat Translator agent/bot, which can also be used in group chat (when you add it as admin). Here Are a Few Translations I asked KuzarBot (Created With Coze), and It Responded Perfectly in Less Than a Second. I previously used the Coze platform (with zero coding experience) to build a Chat Translator AI that can be used on Telegram, WhatsApp, Discord, and others. Coze platform (with zero coding experience) build a Chat Translator AI Here, I will show you How to Configure/Integrate the Chatbot built with Coze into the Telegram chat. How to Configure/Integrate the Chatbot built with Coze into the Telegram chat. Requirements Requirements Telegram account BotFather Telegram account BotFather Step 1 Step 1 Create a Telegram account. And if you already have an account with Telegram, you can use your existing account. Step 2 Step 2 Search BotFather in your Telegram account search bar, and click Start as shown here. BotFather BotFather Start Start Type this command /newbot and send. /newbot /newbot You will see a message prompt to choose/set a name for the bot (which can be modified later) and a username. (which can be modified later) Note: Copy and save the token from the BotFather conversation. Note: Copy and save the token from the BotFather conversation. Note: Step 3 Step 3 Go back to your Coze account , and click on the exact chatbot agent from Workspace to open it. Coze account Workspace On the Develop page, click on Publish . Publish After you have "Publish", it will display a page to select any platform to "publish to". After you have "Publish", it will display a page to select any platform to "publish to". 'Click' on Configure for Telegram since we are integrating the Translator Bot into Telegram. Configure Enter the Access Token copied from BotFather on the pop-up page, and 'click' Save . Save Finally, now 'click' on Chat Now or Copy Agent Link, and open it in your Telegram account to use the Chat Translator agent/bot, which can also be used in group chat (when you add it as admin). Chat Now Copy Agent Link, (when you add it as admin). Here Are a Few Translations I asked KuzarBot (Created With Coze), and It Responded Perfectly in Less Than a Second.

How to Configure/Integrate a Chatbot AI In-chat Translator (Built With Coze) Into Telegram Chat

Machine learning experiments require extensive parametrization, including optimizer parameters, network architecture, and data augmentation. However, we strive for concise, readable code instead of a cumbersome 200 lines dedicated to argparse. Our goal is to focus on programming logic rather than threading new parameters through function signatures. Additionally, we seek a structure that is easily expandable without burdening the project, ensuring the reproducibility of experiments. Hydra offers a solution to these challenges. Below, you will find a basic guide on how to use it. What is Hydra? Hydra is a library with rich capabilities for managing configurations. The main site describes the name like this: “The key feature is the ability to dynamically create a hierarchical configuration by composition and override it through config files and the command line.The name Hydra comes from its ability to run multiple similar jobs - much like a Hydra with multiple heads.” But I have my own interpretation of its name. It is just a combination of some tools that work together impressively well! Hydra offers a seamless solution to the common headaches faced by ML engineers when attempting to replicate experiments. It elegantly replaces the need for argparse or YAML configurations, allowing access to parameters both from the command line and YAML files. Consider the pain points: Replicating experiments using argparse forces reliance on string inputs and complicates YAML config launches. Relying solely on YAML files for configuration leads to duplication and potential errors when only a single parameter needs alteration. Hydra addresses these issues by enabling dynamic configuration adjustments without the need for multiple bulky files or rigid command-line arguments. Furthermore, it simplifies the process of passing complex configurations, such as model architectures or functions, directly from the config file to the model. This capability eliminates the tedious task of manually feeding parameters into the model, streamlining the workflow and reducing the margin for error. Basic Setup Let's imagine the simplest setup: multiclass classification on MNIST using an MLP. We have a configuration and a training script. .
├── configs
│   └── config.yaml
└── main.py The main script can look like this. import torch
import torch.nn as nn
import torch.optim as optim
from torchvision import datasets, transforms
from torch.utils.data import DataLoader

import hydra
from omegaconf import DictConfig

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.model = nn.Sequential(
            nn.Linear(28*28, 128),
            nn.ReLU(),
            nn.Linear(128, 64),
            nn.ReLU(),
            nn.Linear(64, 10)
        )
    
    def forward(self, x):
        return self.model(x.view(-1, 28*28))

@hydra.main(version_base=None, config_path=None, config_name="config")
def main(cfg: DictConfig):
    # Load MNIST dataset
    transform = transforms.Compose([
        transforms.ToTensor(), 
        transforms.Normalize((0.5,), (0.5,))
    ])
    train_dataset = datasets.MNIST(root='./data', train=True, download=True, transform=transform)
    train_loader = DataLoader(train_dataset, batch_size=cfg.batch_size, shuffle=True)

    # Initialize the network, loss function, and optimizer
    model = Net()
    criterion = nn.CrossEntropyLoss()
    optimizer = optim.SGD(model.parameters(), lr=cfg.lr, momentum=cfg.momentum)

    # Train the network
    for epoch in range(cfg.epochs):  # loop over the dataset multiple times
        for i, (inputs, labels) in enumerate(train_loader, 0):
            optimizer.zero_grad()  # zero the parameter gradients
            outputs = model(inputs)  # forward pass
            loss = criterion(outputs, labels)  # calculate loss
            loss.backward()  # backward pass
            optimizer.step()  # optimize

    print('Finished Training')

if __name__ == "__main__":
    main() The configuration has the following structure. It is the default, and parameters from it will be used by default if not specified otherwise. # conf/config.yaml
batch_size: 64
lr: 0.01
momentum: 0.9
epochs: 1 To start training you can use: python main.py And for changing parameters, you can do it not only from YAML but also from cli: python main.py lr=0.03 If you use it, parameters from cli override the first parameters. As you can see, there is no argparse or additional middleware. It's easy to change parameters from the command line. Pass Class Object With Hydra We see that in class Net, the architecture can be highly customizable. Even more, we can use also another net for this, e.g. CNN. But we do not want to change our pipeline. Hydra can construct almost any Python object with specified parameters. Let’s describe our net in another way, using YAML. model:
  _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Flatten
    - _target_: torch.nn.Linear
      in_features: 784  # 28x28 images are flattened into 784
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 64
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 64
      out_features: 10 and for CNN _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Conv2d
      in_channels: 1  # MNIST images are grayscale, so 1 input channel
      out_channels: 32  # Number of output channels
      kernel_size: 3  # Size of the convolutional kernel
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2  # Pooling window size
      stride: 2
    - _target_: torch.nn.Conv2d
      in_channels: 32
      out_channels: 64
      kernel_size: 3
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2
      stride: 2
    - _target_: torch.nn.Flatten  # Flatten the output for the fully connected layer
    - _target_: torch.nn.Linear
      in_features: 7*7*64  # Size after convolutions and pooling
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Dropout
      p: 0.5  # Dropout rate
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 10  # Number of classes in MNIST Pass Function Through Config Sometimes, our parameters can also function. For example, How to Access Config Without Decorating Main - e.g., in ipython Notebook. So, you can use Compose API. from hydra import compose, initialize from omegaconf import OmegaConf

with initialize(version_base=None, config_path="conf", job_name="run_0001"):
  cfg = compose(config_name="config", overrides= ["parameter=value"])

print (OmegaConf.to_yaml(cfg)) Multi Runs For example, you want to find some good hyperparam and Hydra allows a lot for that. Basically, you can launch several experiments sequentially like. python train_model.py -m "batch_size=16,32,64" It launches sequentially, but it is easy to extend to parallel if turned on some launcher like joblib. python train_model.py -m "batch_size=16,32,64 hydra/launcher=joblib" Run All Experiments From the Folder for config_file in configs/*.yaml; do python main.py --config-name="${config_file}"; done Resources hydra - https://hydra.cc/
good template for DL - https://github.com/ashleve/lightning-hydra-template/tree/main Machine learning experiments require extensive parametrization, including optimizer parameters, network architecture, and data augmentation. However, we strive for concise, readable code instead of a cumbersome 200 lines dedicated to argparse. Our goal is to focus on programming logic rather than threading new parameters through function signatures. Additionally, we seek a structure that is easily expandable without burdening the project, ensuring the reproducibility of experiments. Hydra offers a solution to these challenges. Below, you will find a basic guide on how to use it. What is Hydra? Hydra is a library with rich capabilities for managing configurations. The main site describes the name like this: “The key feature is the ability to dynamically create a hierarchical configuration by composition and override it through config files and the command line.The name Hydra comes from its ability to run multiple similar jobs - much like a Hydra with multiple heads.” “The key feature is the ability to dynamically create a hierarchical configuration by composition and override it through config files and the command line. The name Hydra comes from its ability to run multiple similar jobs - much like a Hydra with multiple heads.” But I have my own interpretation of its name. It is just a combination of some tools that work together impressively well! Hydra offers a seamless solution to the common headaches faced by ML engineers when attempting to replicate experiments. It elegantly replaces the need for argparse or YAML configurations, allowing access to parameters both from the command line and YAML files. Consider the pain points: Replicating experiments using argparse forces reliance on string inputs and complicates YAML config launches. Replicating experiments using argparse forces reliance on string inputs and complicates YAML config launches. Relying solely on YAML files for configuration leads to duplication and potential errors when only a single parameter needs alteration. Relying solely on YAML files for configuration leads to duplication and potential errors when only a single parameter needs alteration. Hydra addresses these issues by enabling dynamic configuration adjustments without the need for multiple bulky files or rigid command-line arguments. Furthermore, it simplifies the process of passing complex configurations, such as model architectures or functions, directly from the config file to the model. This capability eliminates the tedious task of manually feeding parameters into the model, streamlining the workflow and reducing the margin for error. Basic Setup Let's imagine the simplest setup: multiclass classification on MNIST using an MLP. We have a configuration and a training script. .
├── configs
│   └── config.yaml
└── main.py .
├── configs
│   └── config.yaml
└── main.py The main script can look like this. import torch
import torch.nn as nn
import torch.optim as optim
from torchvision import datasets, transforms
from torch.utils.data import DataLoader

import hydra
from omegaconf import DictConfig

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.model = nn.Sequential(
            nn.Linear(28*28, 128),
            nn.ReLU(),
            nn.Linear(128, 64),
            nn.ReLU(),
            nn.Linear(64, 10)
        )
    
    def forward(self, x):
        return self.model(x.view(-1, 28*28))

@hydra.main(version_base=None, config_path=None, config_name="config")
def main(cfg: DictConfig):
    # Load MNIST dataset
    transform = transforms.Compose([
        transforms.ToTensor(), 
        transforms.Normalize((0.5,), (0.5,))
    ])
    train_dataset = datasets.MNIST(root='./data', train=True, download=True, transform=transform)
    train_loader = DataLoader(train_dataset, batch_size=cfg.batch_size, shuffle=True)

    # Initialize the network, loss function, and optimizer
    model = Net()
    criterion = nn.CrossEntropyLoss()
    optimizer = optim.SGD(model.parameters(), lr=cfg.lr, momentum=cfg.momentum)

    # Train the network
    for epoch in range(cfg.epochs):  # loop over the dataset multiple times
        for i, (inputs, labels) in enumerate(train_loader, 0):
            optimizer.zero_grad()  # zero the parameter gradients
            outputs = model(inputs)  # forward pass
            loss = criterion(outputs, labels)  # calculate loss
            loss.backward()  # backward pass
            optimizer.step()  # optimize

    print('Finished Training')

if __name__ == "__main__":
    main() import torch
import torch.nn as nn
import torch.optim as optim
from torchvision import datasets, transforms
from torch.utils.data import DataLoader

import hydra
from omegaconf import DictConfig

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.model = nn.Sequential(
            nn.Linear(28*28, 128),
            nn.ReLU(),
            nn.Linear(128, 64),
            nn.ReLU(),
            nn.Linear(64, 10)
        )
    
    def forward(self, x):
        return self.model(x.view(-1, 28*28))

@hydra.main(version_base=None, config_path=None, config_name="config")
def main(cfg: DictConfig):
    # Load MNIST dataset
    transform = transforms.Compose([
        transforms.ToTensor(), 
        transforms.Normalize((0.5,), (0.5,))
    ])
    train_dataset = datasets.MNIST(root='./data', train=True, download=True, transform=transform)
    train_loader = DataLoader(train_dataset, batch_size=cfg.batch_size, shuffle=True)

    # Initialize the network, loss function, and optimizer
    model = Net()
    criterion = nn.CrossEntropyLoss()
    optimizer = optim.SGD(model.parameters(), lr=cfg.lr, momentum=cfg.momentum)

    # Train the network
    for epoch in range(cfg.epochs):  # loop over the dataset multiple times
        for i, (inputs, labels) in enumerate(train_loader, 0):
            optimizer.zero_grad()  # zero the parameter gradients
            outputs = model(inputs)  # forward pass
            loss = criterion(outputs, labels)  # calculate loss
            loss.backward()  # backward pass
            optimizer.step()  # optimize

    print('Finished Training')

if __name__ == "__main__":
    main() The configuration has the following structure. It is the default, and parameters from it will be used by default if not specified otherwise. # conf/config.yaml
batch_size: 64
lr: 0.01
momentum: 0.9
epochs: 1 # conf/config.yaml
batch_size: 64
lr: 0.01
momentum: 0.9
epochs: 1 To start training you can use: python main.py python main.py And for changing parameters, you can do it not only from YAML but also from cli: python main.py lr=0.03 python main.py lr=0.03 If you use it, parameters from cli override the first parameters. As you can see, there is no argparse or additional middleware. It's easy to change parameters from the command line. Pass Class Object With Hydra We see that in class Net, the architecture can be highly customizable. Even more, we can use also another net for this, e.g. CNN. But we do not want to change our pipeline. Hydra can construct almost any Python object with specified parameters. Let’s describe our net in another way, using YAML. model:
  _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Flatten
    - _target_: torch.nn.Linear
      in_features: 784  # 28x28 images are flattened into 784
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 64
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 64
      out_features: 10 model:
  _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Flatten
    - _target_: torch.nn.Linear
      in_features: 784  # 28x28 images are flattened into 784
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 64
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Linear
      in_features: 64
      out_features: 10 and for CNN _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Conv2d
      in_channels: 1  # MNIST images are grayscale, so 1 input channel
      out_channels: 32  # Number of output channels
      kernel_size: 3  # Size of the convolutional kernel
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2  # Pooling window size
      stride: 2
    - _target_: torch.nn.Conv2d
      in_channels: 32
      out_channels: 64
      kernel_size: 3
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2
      stride: 2
    - _target_: torch.nn.Flatten  # Flatten the output for the fully connected layer
    - _target_: torch.nn.Linear
      in_features: 7*7*64  # Size after convolutions and pooling
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Dropout
      p: 0.5  # Dropout rate
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 10  # Number of classes in MNIST _target_: torch.nn.Sequential
  layers:
    - _target_: torch.nn.Conv2d
      in_channels: 1  # MNIST images are grayscale, so 1 input channel
      out_channels: 32  # Number of output channels
      kernel_size: 3  # Size of the convolutional kernel
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2  # Pooling window size
      stride: 2
    - _target_: torch.nn.Conv2d
      in_channels: 32
      out_channels: 64
      kernel_size: 3
      stride: 1
      padding: 1
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.MaxPool2d
      kernel_size: 2
      stride: 2
    - _target_: torch.nn.Flatten  # Flatten the output for the fully connected layer
    - _target_: torch.nn.Linear
      in_features: 7*7*64  # Size after convolutions and pooling
      out_features: 128
    - _target_: torch.nn.ReLU
    - _target_: torch.nn.Dropout
      p: 0.5  # Dropout rate
    - _target_: torch.nn.Linear
      in_features: 128
      out_features: 10  # Number of classes in MNIST Pass Function Through Config Sometimes, our parameters can also function. For example, How to Access Config Without Decorating Main - e.g., in ipython Notebook. So, you can use Compose API . Compose API from hydra import compose, initialize from omegaconf import OmegaConf

with initialize(version_base=None, config_path="conf", job_name="run_0001"):
  cfg = compose(config_name="config", overrides= ["parameter=value"])

print (OmegaConf.to_yaml(cfg)) from hydra import compose, initialize from omegaconf import OmegaConf

with initialize(version_base=None, config_path="conf", job_name="run_0001"):
  cfg = compose(config_name="config", overrides= ["parameter=value"])

print (OmegaConf.to_yaml(cfg)) Multi Runs For example, you want to find some good hyperparam and Hydra allows a lot for that. Basically, you can launch several experiments sequentially like. python train_model.py -m "batch_size=16,32,64" python train_model.py -m "batch_size=16,32,64" It launches sequentially, but it is easy to extend to parallel if turned on some launcher like joblib. python train_model.py -m "batch_size=16,32,64 hydra/launcher=joblib" python train_model.py -m "batch_size=16,32,64 hydra/launcher=joblib" Run All Experiments From the Folder for config_file in configs/*.yaml; do python main.py --config-name="${config_file}"; done for config_file in configs/*.yaml; do python main.py --config-name="${config_file}"; done Resources hydra - https://hydra.cc/ good template for DL - https://github.com/ashleve/lightning-hydra-template/tree/main hydra - https://hydra.cc/ https://hydra.cc/ good template for DL - https://github.com/ashleve/lightning-hydra-template/tree/main https://github.com/ashleve/lightning-hydra-template/tree/main

How to Configure Experiments With Hydra - From an ML Enginner Perspective

Introduction The article demonstrates the use of ConfigureAwait(false) efficiently to add deadlock-free asynchronous code. Learning Objectives How to use ConfigureAwait(false) instead of traditional async await programming Why ConfigureAwait(false) is better Prerequisites for Developers Basic understanding of C# programming language Basic understanding of asynchronous programming using async await Getting Started Consider an example where the user wants to load data asynchronously within a method. /// <summary>
    /// Old approach with classic async await 
    /// </summary>
    /// <returns></returns>
    public async static Task OldApproach()
    {
        await ReadDataAsync();
    } In this approach, the await operator waits for ReadDataAsync and then proceeds with execution in the same synchronization context from where it began. The aforementioned approach is used when the developer ensures that UI updates are executed in a separate thread. However, it may introduce potential deadlock risks. Optimized approach with ConfigureAwait(false) Let’s transform the above method using ConfigureAwait(false) /// <summary>
    /// Optimized approach with ConfigureAwait
    /// </summary>
    /// <returns></returns>
    public static async Task OptimizedApproachAsync()
    {
        await ReadDataAsync().ConfigureAwait(false);
    } By adding this, the compiler doesn't add the execution in the same synchronization context, which reduces the chances of deadlocks. The aforementioned optimization is beneficial in non-UI applications like library code, etc. Why ConfigureAwait(false) is better Please find below the benefits of using ConfigureAwait(false) method Improved Performance As the optimized approach doesn’t add the execution to the same synchronization context, it saves on extra overhead and helps create scalable applications. Reduce chances of deadlocks ConfigureAwait(false) method mitigates the risk of deadlocks when the synchronization context is blocked. Conclusion The ConfigureAwait(false) method in C# aims to craft efficient, deadlock-avoidant asynchronous code. Its advantages are particularly beneficial in non-UI applications and in library projects. Complete Code on GitHub GitHub — ssukhpinder/30DayChallenge.Net C# Programming🚀 Thank you for being a part of the C# community! Before you leave: If you’ve made it this far, please show your appreciation with a clap and follow the author! 👏️️ Follow us: | | | | | X LinkedIn Dev.to Hashnode Newsletter Tumblr Visit our other platforms: | | | | GitHub Instagram Tiktok Quora Daily.dev More content at C# Programming Also published . here

The 30-Day .NET Challenge Day 13: ConfigureAwait(false)

Your Updated Guide to Building a Multi-language Asp.Net 8 MVC Application 

Visualizing PostgreSQL Data With Angular To Analyze What Passengers Find Rude to Do During a Flight

Building Scalable E-commerce Infrastructure on Magento

AWS Athena: Everything You Need To Know

Crypto Tool or Data Thief? How Meme-Token-Hunter-Bot and Its Clones Steal from macOS Users

ክሪፕቶ መሣሪያ ወይስ ዳታ ሌባ? Meme-Token-Hunter-Bot እና ክሎኖቹ ከማክሮስ ተጠቃሚዎች እንዴት እንደሚሰርቁ

¿Herramienta criptográfica o ladrón de datos? Cómo el Meme-Token-Hunter-Bot y sus clones roban a los usuarios de macOS

כלי קריפטו או גנב נתונים? כיצד Meme-Token-Hunter-Bot והשיבוטים שלו גונבים ממשתמשי macOS

暗号ツールかデータ泥棒か？Meme-Token-Hunter-BotとそのクローンがmacOSユーザーから盗む方法

Крипто алатка или крадец на податоци? Како Meme-Token-Hunter-Bot и неговите клонови крадат од корисниците на macOS

Alat Kripto atau Pencuri Data? Bagaimana Meme-Token-Hunter-Bot dan Klonnya Mencuri daripada Pengguna macOS

د کریپټو وسیله یا ډیټا چور؟ څنګه Meme-Token-Hunter-Bot او د هغې کلونونه د macOS کاروونکو څخه غلا کوي

¿Cripto Herramientachu icha Datos Suwachu? Imaynatachus Meme-Token-Hunter-Bot chaymanta Clonesnin macOS Usuariokunamanta suwakunku

Kryptoverktyg eller datatjuv? Hur Meme-Token-Hunter-Bot och dess kloner stjäl från macOS-användare

Асбоби крипто ё дузди маълумот? Чӣ тавр Meme-Token-Hunter-Bot ва клонҳои он аз корбарони macOS медузданд

Kripto Aracı mı yoksa Veri Hırsızı mı? Meme-Token-Hunter-Bot ve Klonları macOS Kullanıcılarından Nasıl Çalıyor?

Ithuluzi le-Crypto noma Isela Ledatha? Indlela i-Meme-Token-Hunter-Bot nama-Clones ayo antshontsha ngayo kubasebenzisi be-macOS

Stories, News & Mentions about CONFIGURE8

Configure8’s Stories on HackerNoon