WizardCoder 34B: The New Frontier in Code Generation with Language Models

October 3, 2023


The era of digital transformation has ushered in a wave of advancements in the tech industry, including the burgeoning field of Artificial Intelligence (AI). Large Language Models (LLMs) have turned into game-changers, especially in the domain of software development. LLMs can read, understand, and even generate code, bridging the gap between natural language and computer languages. Their versatility has made them an indispensable tool for developers, automating a multitude of tasks ranging from code completion and bug fixing to even generating whole code bases.

As the technology progresses, we're seeing more specialized language models designed for very specific tasks. WizardCoder 34B, a model created with a focus on code generation, is one of the latest LLMs that is being hailed by experts. Unlike generic LLMs that are good at a range of tasks, WizardCoder 34B excels at understanding and generating code. This specialization has already made WizardCoder a noteworthy competitor to other established models like GPT-4 and ChatGPT-3.5. 

WizardCoder 34B is built on Code Llama, a large language model (LLM) developed by Meta. Code Llama serves as the foundational architecture upon which WizardCoder 34B has been fine-tuned and optimized for coding tasks. The adaptation and specialization of WizardCoder 34B make it distinct from the general-purpose capabilities of Code Llama, focusing primarily on code generation, debugging, and other development-related activities.

Developed by Wizard LM, it has been fine-tuned to perform exceptionally well in coding tasks, boasting impressive results on coding benchmarks like HumanEval. In summary, WizardCoder 34B is not just another language model; it is a targeted solution for software development tasks [1]. As we explore its capabilities and features in this blog, we'll see why it's drawing the attention of developers and tech companies alike.

WizardCoder 34B vs the Competition

The world of large language models is teeming with contenders. Models like GPT-4 and GPT-3.5 have been around for some time and are celebrated for their wide range of capabilities, including code generation. Then there is Code Llama, a model developed by Meta, which serves as the base for WizardCoder and is specifically aimed at coding tasks. But what makes WizardCoder 34B stand apart?

  • GPT-4: While GPT-4 is a general-purpose model proficient at various tasks, it is not specialized in code generation. When it comes to more complex coding tasks, GPT-4 may fall short.
  • GPT-3.5: Similar to GPT-4 but slightly older, ChatGPT-3.5 offers broad capabilities. However, when the stakes are coding-specific, WizardCoder 34B has the edge.
  • Code Llama: Developed by Meta, this is more of a cousin to WizardCoder 34B. While Code Llama lays the groundwork, WizardCoder builds upon it, refining and optimizing the model specifically for code generation tasks.

Performance Metrics: HumanEval Benchmarks

An important benchmark test for these models' coding capabilities lies in their performance on HumanEval, a widely-recognized benchmark for evaluating the coding prowess of LLMs [2].

  • GPT-4: The August release of GPT-4 scored an impressive 67% on the HumanEval benchmarks in the month of march, and 82% in the month of August
  • GPT-3.5: Managed to secure a respectable score but lagged behind the specialized models with a score around 65%.
  • Code Llama: Scored pass rates of 67.6% and 69.5% on its fine-tuned versions, making it almost on par with GPT-4 [3].
  • WizardCoder 34B: Steals the spotlight by scoring a 73.2% on the HumanEval Benchmarks, outdoing not just its predecessor, Code Llama, but also more generalized models like ChatGPT-3.5.

In March 2023, GPT-4 scored 67% on the HumanEval benchmarks. However, by August, this figure had jumped to 82%. This significant increase in just a few months highlights the rapid development and fine-tuning that GPT-4 has undergone. WizardCoder 34B has an impressive score of 73.2% on the HumanEval benchmarks. While this is a strong showing, it's essential to note that this score was compared to GPT-4's March version, not the more recent August update.

However, it's important to consider the scope and context. WizardCoder 34B is specialized for coding tasks, particularly in Python, and might offer more nuanced and specialized outputs for such tasks compared to a generalist model like GPT-4. WizardCoder 34B is also fine-tuned for code generation, which may make it better for very specific coding scenarios that are not captured by the HumanEval benchmark.

Under the Hood: How WizardCoder Works

Let's take a look at the technical aspects that make WizardCoder 34B tick. We'll focus on the architecture, underlying technologies, and any special features like the Evol-Instruct method that set this model apart.

Architecture and Technologies

WizardCoder 34B is built on Code Llama, a large language model developed by Meta. Code Llama itself is a code-specialized version of Llama 2. It comes with 34 billion parameters, making it one of the most substantial models focused on code generation. The model is optimized for Python code and leverages Transformer architecture, similar to other leading large language models like GPT-4.

Evol-Instruct Method

One of the standout features of WizardCoder 34B is the Evol-Instruct method. This is a fine-tuning process that evolves the model's instruction-following capabilities through iterative training. The method allows WizardCoder to better understand the context of the coding task and generate more accurate and optimized code. Here's how Evol-Instruct works:

  • Initial Training: The model is trained with a large dataset of code snippets and corresponding natural language instructions.
  • Evaluation: It is then evaluated using HumanEval or other benchmarks to assess its capability in following instructions to generate code.
  • Iterative Fine-tuning: Based on the evaluations, the model is fine-tuned iteratively to enhance its ability to understand and generate task-specific code.

Special Features

  • Context-Awareness: WizardCoder 34B has a better understanding of the code's context, making it more effective in generating coherent and functional code.
  • Language Support: Although optimized for Python, WizardCoder 34B is designed to adapt to other programming languages as well.
  • Quantization Levels: To manage computational needs, WizardCoder 34B offers different quantization levels. The higher the number, the more accurate the model is, although it might require more computational power.

Understanding these architectural nuances and features can help users gain a fuller understanding of the WizardCoder 34B's capabilities and limitations, useful to developers, data scientists, and AI enthusiasts.

Installation and Setup

In this section, we will walk you through the installation process for WizardCoder, specifically the 34B variant, and list the system requirements and dependencies you need to run the model smoothly.

System Requirements

  • RAM: At least 32GB for the 34B model
  • Python 3.6 or higher

Installation Steps

Step 1: Install the required Python packages.

pip install transformers deepspeed

Here, transformers are used for utilizing the Hugging Face model hub, and Deepspeed for optimization and running models more efficiently.

Step 2: Import the WizardCoder Model from the Hugging Face library.

from transformers import AutoModelForCausalLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained('WizardLM/WizardCoder-Python-34B-V1.0')
model = AutoModelForCausalLM.from_pretrained('WizardLM/WizardCoder-Python-34B-V1.0')

Step 3: Verify the Installation.

After installation, it's always good to verify that everything is working as expected. You can do this by running a sample instruction to generate code.

instruction = 'Write a Python function to calculate the nth Fibonacci number.'
input_ids = tokenizer(instruction, return_tensors='pt').input_ids
output = model.generate(input_ids)
decoded_output = tokenizer.decode(output[0], skip_special_tokens=True)

Now that the installation procedure for WizardCoder is discussed, along with the system requirements, developers can make use of this advanced language model for their coding tasks.

Demo and Inference

Once WizardCoder 34B is installed and configured, it may be necessary to run some quick demos or inferences to test its capabilities. This section provides a straightforward guide on how to achieve this. Inference Demos can be run by the following:

With Ollama CLI

Start Ollama Server: Run the command ollama serve to start the Ollama server.

Run the Model: Open a new terminal and run the following command:

ollama run wizardcoder:34b-python

With API Calls

Start Ollama Server: If not already running, initiate it using ollama serve.

Run the Model: Use a curl command to run the model. Here's an example:

curl -X POST http://localhost:11434/api/generate -d '{
  'model': 'wizardcoder:34b-python',
  'prompt':'Write a Python function to return the Fibonacci sequence up to n'

Getting Started: Coding with WizardCoder

This section shows how to start coding with WizardCoder 34B. Some example use-cases are explored, including generating DevOps scripts, building machine learning pipelines, and more. 

Example Use Cases

Automating DevOps Scripts

WizardCoder can assist you in writing shell or Python scripts to automate various DevOps tasks like environment setup, server provisioning, and so on.

WizardCoder can assist you in writing shell or Python scripts to automate various DevOps tasks like environment setup, server provisioning, and so on.

instruction = 'Write a Python script to automate the setup of a new Linux environment.'
input_ids = tokenizer(instruction, return_tensors='pt').input_ids
output = model.generate(input_ids)
decoded_output = tokenizer.decode(output[0], skip_special_tokens=True)

Other Examples Include

  • Data Analysis - instruction = 'Generate Python code for data analysis using pandas and matplotlib for a sample dataset'
  • Machine Learning Pipelines - instruction = 'Generate a Python code snippet for a machine learning pipeline using scikit-learn.'
  • Boilerplate code for building RESTful APIs. - instruction = 'Generate Python code for a simple RESTful API using Flask.'

Memory and Performance Tuning

Working with a large language model for code generation like WizardCoder 34B requires careful planning around memory usage and performance. In this section, we’ll delve into the memory requirements, optimization techniques, model variants, and quantization levels to help you get the best out of WizardCoder 34B. Generally require at least 32GB of RAM for smooth operation. If you're running this model, make sure to allocate sufficient memory to prevent crashes or slowdowns.

Model Variants

WizardCoder 34B is the most parameter-rich variant, but there are lighter versions available with fewer parameters. These can be useful for tasks that do not require the full power of the 34B model.

  • 13B Model: Requires at least 16GB of RAM.
  • 7B Model: Suited for less memory-intensive tasks.

Quantization Levels

WizardCoder 34B allows you to choose different levels of quantization to balance accuracy against speed and memory usage.

  • 4-bit Quantization (q4): Faster but less accurate.
  • 8-bit Quantization (q8): Slower but more accurate. Requires more memory.

You can select the quantization level when running your model by specifying the corresponding tag. For example, to use 8-bit quantization, you can append -q8 to the model name during initialization.

ollama run wizardcoder:34b-python-q8

Here, Ollama is an application to run open-source large language models, such as LLaMA2, locally. Ollama bundles model weights, configuration, and data into a single package, defined by a Modelfile. It optimizes setup and configuration details, including GPU usage.

Future Direction

As the field of machine learning and natural language processing evolves, so does WizardCoder 34B. Here are some insights into the future directions the developers might take and the opportunities for synergies with other technologies.

Upcoming Features and Version Updates

  • Better Code Quality Metrics: To address the limitations of current benchmarks like HumanEval, WizardCoder developers must consider incorporating more holistic code quality metrics, like code readability, maintainability, and even automated testing scores.
  • Advanced Fine-Tuning Capabilities: The future versions may offer more customizable fine-tuning options, enabling users to adapt the model to highly specialized tasks.
  • Real-Time Collaboration: An exciting possibility is the integration of WizardCoder into IDEs for real-time coding assistance and automated code review.

Possible Collaborations with Other Technologies

  • Integrated Developer Environments (IDEs): One of the most obvious and useful collaborations could be with IDEs. A WizardCoder plugin could provide real-time coding assistance, significantly speeding up the development process.
  • Continuous Integration Tools: Integration with CI/CD pipelines could make automated code generation and review a seamless part of the software development lifecycle.
  • AI in Edge Computing: Combining WizardCoder with edge computing technologies could enable powerful code generation and analysis tasks to be performed locally, reducing the need for cloud resources.
  • Data Science Platforms: A collaboration with data science platforms could mean automated script generation for data preprocessing, analysis, and visualization, making the data science workflow more efficient.
  • Educational Software: WizardCoder can also serve as an educational tool, helping learners understand coding practices, algorithms, and data structures more effectively.


WizardCoder 34B has emerged as a powerful tool in the realm of automated coding and software development. Built on Code Llama by Meta, it has shown promising results in tasks evaluated using the HumanEval benchmark. While it lags behind GPT-4 in its August 2023 iteration, the model shows significant promise and is subject to rapid iterations. 

WizardCoder 34B is also marked by its ease of installation, advanced features like the Evol-Instruct method, and options for memory and performance tuning.

While the data and metrics cited in this blog are accurate as of the time of writing, it's crucial to acknowledge that this is a rapidly evolving field. The performance and features of WizardCoder 34B could change. Always make sure to consult the most recent documentation for the latest information.

If you need to run WizardCoder 34B, E2E cloud has a large selection of GPUs to select from. NVIDIA H100 is a good fit, as it is highly compatible for LLMs.


[1] Z. Luo et al., 'WizardCoder: Empowering Code Large Language Models with Evol-Instruct,' Jun. 2023, [Online]. Available: http://arxiv.org/abs/2306.08568.

[2] M. Chen et al., 'Evaluating Large Language Models Trained on Code,' Jul. 2021, [Online]. Available: http://arxiv.org/abs/2107.03374.

[3] L. Tomassi, 'All About Code Llama: Meta’s New Coding AI,' CodeMotion, 2023. https://www.codemotion.com/magazine/ai-ml/all-about-code-llama-metas-new-coding-ai/.

Latest Blogs
This is a decorative image for: A Complete Guide To Customer Acquisition For Startups
October 18, 2022

A Complete Guide To Customer Acquisition For Startups

Any business is enlivened by its customers. Therefore, a strategy to constantly bring in new clients is an ongoing requirement. In this regard, having a proper customer acquisition strategy can be of great importance.

So, if you are just starting your business, or planning to expand it, read on to learn more about this concept.

The problem with customer acquisition

As an organization, when working in a diverse and competitive market like India, you need to have a well-defined customer acquisition strategy to attain success. However, this is where most startups struggle. Now, you may have a great product or service, but if you are not in the right place targeting the right demographic, you are not likely to get the results you want.

To resolve this, typically, companies invest, but if that is not channelized properly, it will be futile.

So, the best way out of this dilemma is to have a clear customer acquisition strategy in place.

How can you create the ideal customer acquisition strategy for your business?

  • Define what your goals are

You need to define your goals so that you can meet the revenue expectations you have for the current fiscal year. You need to find a value for the metrics –

  • MRR – Monthly recurring revenue, which tells you all the income that can be generated from all your income channels.
  • CLV – Customer lifetime value tells you how much a customer is willing to spend on your business during your mutual relationship duration.  
  • CAC – Customer acquisition costs, which tells how much your organization needs to spend to acquire customers constantly.
  • Churn rate – It tells you the rate at which customers stop doing business.

All these metrics tell you how well you will be able to grow your business and revenue.

  • Identify your ideal customers

You need to understand who your current customers are and who your target customers are. Once you are aware of your customer base, you can focus your energies in that direction and get the maximum sale of your products or services. You can also understand what your customers require through various analytics and markers and address them to leverage your products/services towards them.

  • Choose your channels for customer acquisition

How will you acquire customers who will eventually tell at what scale and at what rate you need to expand your business? You could market and sell your products on social media channels like Instagram, Facebook and YouTube, or invest in paid marketing like Google Ads. You need to develop a unique strategy for each of these channels. 

  • Communicate with your customers

If you know exactly what your customers have in mind, then you will be able to develop your customer strategy with a clear perspective in mind. You can do it through surveys or customer opinion forms, email contact forms, blog posts and social media posts. After that, you just need to measure the analytics, clearly understand the insights, and improve your strategy accordingly.

Combining these strategies with your long-term business plan will bring results. However, there will be challenges on the way, where you need to adapt as per the requirements to make the most of it. At the same time, introducing new technologies like AI and ML can also solve such issues easily. To learn more about the use of AI and ML and how they are transforming businesses, keep referring to the blog section of E2E Networks.

Reference Links




This is a decorative image for: Constructing 3D objects through Deep Learning
October 18, 2022

Image-based 3D Object Reconstruction State-of-the-Art and trends in the Deep Learning Era

3D reconstruction is one of the most complex issues of deep learning systems. There have been multiple types of research in this field, and almost everything has been tried on it — computer vision, computer graphics and machine learning, but to no avail. However, that has resulted in CNN or convolutional neural networks foraying into this field, which has yielded some success.

The Main Objective of the 3D Object Reconstruction

Developing this deep learning technology aims to infer the shape of 3D objects from 2D images. So, to conduct the experiment, you need the following:

  • Highly calibrated cameras that take a photograph of the image from various angles.
  • Large training datasets can predict the geometry of the object whose 3D image reconstruction needs to be done. These datasets can be collected from a database of images, or they can be collected and sampled from a video.

By using the apparatus and datasets, you will be able to proceed with the 3D reconstruction from 2D datasets.

State-of-the-art Technology Used by the Datasets for the Reconstruction of 3D Objects

The technology used for this purpose needs to stick to the following parameters:

  • Input

Training with the help of one or multiple RGB images, where the segmentation of the 3D ground truth needs to be done. It could be one image, multiple images or even a video stream.

The testing will also be done on the same parameters, which will also help to create a uniform, cluttered background, or both.

  • Output

The volumetric output will be done in both high and low resolution, and the surface output will be generated through parameterisation, template deformation and point cloud. Moreover, the direct and intermediate outputs will be calculated this way.

  • Network architecture used

The architecture used in training is 3D-VAE-GAN, which has an encoder and a decoder, with TL-Net and conditional GAN. At the same time, the testing architecture is 3D-VAE, which has an encoder and a decoder.

  • Training used

The degree of supervision used in 2D vs 3D supervision, weak supervision along with loss functions have to be included in this system. The training procedure is adversarial training with joint 2D and 3D embeddings. Also, the network architecture is extremely important for the speed and processing quality of the output images.

  • Practical applications and use cases

Volumetric representations and surface representations can do the reconstruction. Powerful computer systems need to be used for reconstruction.

Given below are some of the places where 3D Object Reconstruction Deep Learning Systems are used:

  • 3D reconstruction technology can be used in the Police Department for drawing the faces of criminals whose images have been procured from a crime site where their faces are not completely revealed.
  • It can be used for re-modelling ruins at ancient architectural sites. The rubble or the debris stubs of structures can be used to recreate the entire building structure and get an idea of how it looked in the past.
  • They can be used in plastic surgery where the organs, face, limbs or any other portion of the body has been damaged and needs to be rebuilt.
  • It can be used in airport security, where concealed shapes can be used for guessing whether a person is armed or is carrying explosives or not.
  • It can also help in completing DNA sequences.

So, if you are planning to implement this technology, then you can rent the required infrastructure from E2E Networks and avoid investing in it. And if you plan to learn more about such topics, then keep a tab on the blog section of the website

Reference Links



This is a decorative image for: Comprehensive Guide to Deep Q-Learning for Data Science Enthusiasts
October 18, 2022

A Comprehensive Guide To Deep Q-Learning For Data Science Enthusiasts

For all data science enthusiasts who would love to dig deep, we have composed a write-up about Q-Learning specifically for you all. Deep Q-Learning and Reinforcement learning (RL) are extremely popular these days. These two data science methodologies use Python libraries like TensorFlow 2 and openAI’s Gym environment.

So, read on to know more.

What is Deep Q-Learning?

Deep Q-Learning utilizes the principles of Q-learning, but instead of using the Q-table, it uses the neural network. The algorithm of deep Q-Learning uses the states as input and the optimal Q-value of every action possible as the output. The agent gathers and stores all the previous experiences in the memory of the trained tuple in the following order:

State> Next state> Action> Reward

The neural network training stability increases using a random batch of previous data by using the experience replay. Experience replay also means the previous experiences stocking, and the target network uses it for training and calculation of the Q-network and the predicted Q-Value. This neural network uses openAI Gym, which is provided by taxi-v3 environments.

Now, any understanding of Deep Q-Learning   is incomplete without talking about Reinforcement Learning.

What is Reinforcement Learning?

Reinforcement is a subsection of ML. This part of ML is related to the action in which an environmental agent participates in a reward-based system and uses Reinforcement Learning to maximize the rewards. Reinforcement Learning is a different technique from unsupervised learning or supervised learning because it does not require a supervised input/output pair. The number of corrections is also less, so it is a highly efficient technique.

Now, the understanding of reinforcement learning is incomplete without knowing about Markov Decision Process (MDP). MDP is involved with each state that has been presented in the results of the environment, derived from the state previously there. The information which composes both states is gathered and transferred to the decision process. The task of the chosen agent is to maximize the awards. The MDP optimizes the actions and helps construct the optimal policy.

For developing the MDP, you need to follow the Q-Learning Algorithm, which is an extremely important part of data science and machine learning.

What is Q-Learning Algorithm?

The process of Q-Learning is important for understanding the data from scratch. It involves defining the parameters, choosing the actions from the current state and also choosing the actions from the previous state and then developing a Q-table for maximizing the results or output rewards.

The 4 steps that are involved in Q-Learning:

  1. Initializing parameters – The RL (reinforcement learning) model learns the set of actions that the agent requires in the state, environment and time.
  2. Identifying current state – The model stores the prior records for optimal action definition for maximizing the results. For acting in the present state, the state needs to be identified and perform an action combination for it.
  3. Choosing the optimal action set and gaining the relevant experience – A Q-table is generated from the data with a set of specific states and actions, and the weight of this data is calculated for updating the Q-Table to the following step.
  4. Updating Q-table rewards and next state determination – After the relevant experience is gained and agents start getting environmental records. The reward amplitude helps to present the subsequent step.  

In case the Q-table size is huge, then the generation of the model is a time-consuming process. This situation requires Deep Q-learning.

Hopefully, this write-up has provided an outline of Deep Q-Learning and its related concepts. If you wish to learn more about such topics, then keep a tab on the blog section of the E2E Networks website.

Reference Links



This is a decorative image for: GAUDI: A Neural Architect for Immersive 3D Scene Generation
October 13, 2022

GAUDI: A Neural Architect for Immersive 3D Scene Generation

The evolution of artificial intelligence in the past decade has been staggering, and now the focus is shifting towards AI and ML systems to understand and generate 3D spaces. As a result, there has been extensive research on manipulating 3D generative models. In this regard, Apple’s AI and ML scientists have developed GAUDI, a method specifically for this job.

An introduction to GAUDI

The GAUDI 3D immersive technique founders named it after the famous architect Antoni Gaudi. This AI model takes the help of a camera pose decoder, which enables it to guess the possible camera angles of a scene. Hence, the decoder then makes it possible to predict the 3D canvas from almost every angle.

What does GAUDI do?

GAUDI can perform multiple functions –

  • The extensions of these generative models have a tremendous effect on ML and computer vision. Pragmatically, such models are highly useful. They are applied in model-based reinforcement learning and planning world models, SLAM is s, or 3D content creation.
  • Generative modelling for 3D objects has been used for generating scenes using graf, pigan, and gsn, which incorporate a GAN (Generative Adversarial Network). The generator codes radiance fields exclusively. Using the 3D space in the scene along with the camera pose generates the 3D image from that point. This point has a density scalar and RGB value for that specific point in 3D space. This can be done from a 2D camera view. It does this by imposing 3D datasets on those 2D shots. It isolates various objects and scenes and combines them to render a new scene altogether.
  • GAUDI also removes GANs pathologies like mode collapse and improved GAN.
  • GAUDI also uses this to train data on a canonical coordinate system. You can compare it by looking at the trajectory of the scenes.

How is GAUDI applied to the content?

The steps of application for GAUDI have been given below:

  • Each trajectory is created, which consists of a sequence of posed images (These images are from a 3D scene) encoded into a latent representation. This representation which has a radiance field or what we refer to as the 3D scene and the camera path is created in a disentangled way. The results are interpreted as free parameters. The problem is optimized by and formulation of a reconstruction objective.
  • This simple training process is then scaled to trajectories, thousands of them creating a large number of views. The model samples the radiance fields totally from the previous distribution that the model has learned.
  • The scenes are thus synthesized by interpolation within the hidden space.
  • The scaling of 3D scenes generates many scenes that contain thousands of images. During training, there is no issue related to canonical orientation or mode collapse.
  • A novel de-noising optimization technique is used to find hidden representations that collaborate in modelling the camera poses and the radiance field to create multiple datasets with state-of-the-art performance in generating 3D scenes by building a setup that uses images and text.

To conclude, GAUDI has more capabilities and can also be used for sampling various images and video datasets. Furthermore, this will make a foray into AR (augmented reality) and VR (virtual reality). With GAUDI in hand, the sky is only the limit in the field of media creation. So, if you enjoy reading about the latest development in the field of AI and ML, then keep a tab on the blog section of the E2E Networks website.

Reference Links




Build on the most powerful infrastructure cloud

A vector illustration of a tech city using latest cloud technologies & infrastructure