A Comparative Study of Jina Embeddings vs. Llama Model for Computing Textual Semantic Similarity

January 16, 2024

Text similarity holds substantial importance in diverse natural language processing applications, including but not limited to search engines, recommendation systems, and chatbots. This article will examine two cutting-edge approaches for measuring text similarity: Jina Embeddings and the Llama Model. The exploration will encompass an in-depth analysis of their fundamental mechanisms and practical implementation utilizing the Hugging Face Transformer. Let's proceed with our investigation.

Requirements for Initiating a GPU Node on E2E Cloud

Account and Access

  • E2E Cloud Account: An active E2E Cloud account is a necessity to access the platform and initiate your GPU node. If you haven't created an account yet, the process is straightforward and can be completed through the website.
  • Billing Information: Ensure that your billing information is current and contains sufficient funds to cover the expenses associated with launching and operating your GPU node.

Technical Requirements

  • Operating System: Choose the operating system that aligns with your preferences for the GPU node. E2E Cloud provides a range of Linux distributions and Windows Server versions to cater to diverse needs. Consider compatibility with your software and tools when making your selection.
  • Software Dependencies: Check if your application or workflow requires specific software libraries or dependencies pre-installed on the node. If so, compile a list of these requirements to specify during the configuration of the node.
  • Network Connectivity: Confirm that your local internet connection can accommodate the bandwidth demands of running applications on a remote GPU node. E2E Cloud offers various network bandwidth options, allowing you to choose the one best suited for your expected data transfer and processing requirements.

Knowledge and Preparation

  • Basic Cloud Computing Understanding: Acquaint yourself with fundamental cloud computing concepts, including virtual machines, instances, and resource allocation. This familiarity will facilitate your interaction with the E2E Cloud platform.
  • Security Credentials: Have your SSH key or preferred security credentials ready for accessing your launched GPU node remotely.
  • Application and Script Preparation: If you intend to run specific applications or scripts on the node, ensure they are prepared and compatible with the chosen operating system and GPU environment.

By fulfilling these prerequisites, you can confidently embark on launching your GPU node on E2E Cloud, unlocking the remarkable potential of accelerated computing for your projects. Remember, meticulous planning and preparation form the bedrock of a successful and fruitful cloud computing experience.

Jina Embeddings

Within this integration, we utilize the robust Jina Embeddings, a text embedding model seamlessly combined with the Hugging Face Transformers library. Jina Embeddings, known as JinaBert, is a specialized embedding model grounded in the Bert architecture, specifically tailored to accommodate English text with a maximum sequence length of 8192 tokens. The model undergoes pre-training on the C4 dataset and subsequent fine-tuning on a meticulously curated set of over 400 million sentence pairs and challenging negatives from diverse domains. This thorough training regimen ensures that the embeddings effectively capture intricate semantic relationships, rendering them indispensable for applications demanding a profound comprehension of text.

Importing Libraries and Defining Cosine Similarity Function

from transformers import AutoModel 
from numpy.linalg import norm

# Defining a cosine similarity function using lambda expression
cos_sim = lambda a, b: (a @ b.T) / (norm(a) * norm(b))

In this section, the code includes the essential libraries. The use of AutoModel from the transformers library facilitates the loading of a pre-trained transformer model. The cos_sim function is employed to calculate cosine similarity between two vectors, utilizing the dot product and normalization.

Loading the Pre-Trained Transformer Model

model = AutoModel.from_pretrained("jinaai/jina-embeddings-v2-base-en", trust_remote_code=True)

This line of code loads a pre-trained transformer model named "jinaai/jina-embeddings-v2-base-en". The parameter trust_remote_code=True is specified to guarantee the trustworthiness of the associated remote code for the model.

Generating Embeddings for Sentences

similarity = model.encode(["This is me", "A 2nd sentence"])

The encode method of the model accepts a list of sentences and produces their respective embeddings. In this context, embeddings for two sample sentences are calculated.

Calculating Cosine Similarity

cosine_similarity_score = cos_sim(similarity[0], similarity[1])

Defining compute_similarity Function

def compute_similarity(sentence1, sentence2):
    embeddings = model.encode([sentence1, sentence2])
    result = cos_sim(embeddings[0], embeddings[1])
    return result

This function receives two sentences as input, generates their embeddings using the loaded model, and subsequently determines their cosine similarity using the cos_sim function. The outcome is then returned as the similarity score between the input sentences.

Example Usages of compute_similarity Function

similarity1 = compute_similarity("I love cricket.", "I like football.")
similarity2 = compute_similarity("I like basketball.", "I like basketball.")
similarity3 = compute_similarity("I like football.", "I don't like football.")

These lines exemplify the application of the compute_similarity function with various pairs of sentences. The obtained similarity scores serve as indicators of the semantic similarity between the corresponding sentence pairs.


from transformers import AutoModel
from numpy.linalg import norm

# Define cosine similarity function
cos_sim = lambda a, b: (a @ b.T) / (norm(a) * norm(b))

# Load Jina Embeddings model from Hugging Face Transformers
model = AutoModel.from_pretrained("jinaai/jina-embeddings-v2-base-en", trust_remote_code=True)

# Encode sentences and compute embeddings
similarity = model.encode(["This is me", "A 2nd sentence"])

# Calculate cosine similarity between the embeddings
similarity = cos_sim(embeddings[0], embeddings[1])
print("Cosine Similarity:", similarity)


Cosine Similarity: 0.7132004

To summarize, this code snippet illustrates the process of loading a pre-trained transformer model, producing sentence embeddings, computing cosine similarity, and encapsulating these steps into a reusable function for comparing the semantic similarity of arbitrary sentences.

Llama 2

The Llama Model, accessible via the Hugging Face Transformers library, provides cutting-edge generative text capabilities. Created by Meta, this model is available in multiple sizes, spanning from 7 billion to 70 billion parameters, thereby facilitating a diverse range of applications in natural language processing. A specialized version, Llama 2-Chat, fine-tuned for dialogue scenarios, surpasses numerous open-source chat models and demonstrates competitive performance against well-known closed-source models.

Importing Libraries and Loading Pre-Trained Llama Model

from transformers import LlamaTokenizer, LlamaForCausalLM
import torch

# Load the pre-trained model and tokenizer
model_base_name = "meta-llama/Llama-2-7b-hf"
model = LlamaForCausalLM.from_pretrained(model_base_name)
tokenizer = LlamaTokenizer.from_pretrained(model_base_name)

Within this code snippet, the necessary libraries are imported, and a pre-trained Llama model along with its associated tokenizer are loaded. The variable model_base_name is used to specify the name of the pre-trained model.

Checking Vocabulary Size and Maximum Sequence Length

vocab_size = tokenizer.vocab_size
max_seq_length = model.config.max_position_embeddings
print("Vocabulary Size:", vocab_size)
print("Max Sequence Length:", max_seq_length)

The provided code outputs the vocabulary size and the maximum sequence length permitted by the loaded model. Gaining insights into these values is essential for tokenization and processing the input data effectively.

Modifying Tokenizer for Padding and Special Tokens

tokenizer.add_special_tokens({'pad_token': '[PAD]'})

To manage variable-length sequences, the code includes a padding token in the tokenizer. Special tokens such as [PAD] play a crucial role in ensuring the proper functioning of the model during the tokenization process.

Tokenizing and Preprocessing Input Sentences

sentences = ["This is me", "A 2nd sentence"]
input_ids = tokenizer(sentences, return_tensors='pt', padding=True, truncation=True, max_length=max_seq_length)['input_ids']
input_ids = input_ids.clamp(max=vocab_size - 1)

The Llama tokenizer is employed to tokenize the input sentences. The ensuing input_ids undergo further processing: padding is incorporated, sequences exceeding the specified max_seq_length are truncated, and token IDs are clamped to guarantee they fall within the vocabulary range of the model.

Obtaining Model Outputs (Logits) and Extracting Embeddings

with torch.no_grad():
    outputs = model(input_ids)

# Extract hidden states from the base model
hidden_states = outputs.logits

# Extract embeddings for [CLS] tokens
cls_embeddings = hidden_states[:, 0, :]

The tokenized input IDs are fed through the Llama model, producing outputs in the form of logits. From these logits, embeddings for the [CLS] tokens are extracted. The [CLS] token conventionally encapsulates a condensed representation of the entire input sequence.

Computing Cosine Similarity

import torch.nn.functional as F
similarity = F.cosine_similarity(cls_embeddings[0].unsqueeze(0), cls_embeddings[1].unsqueeze(0))
print("Cosine Similarity:", similarity.item())


from transformers import LlamaTokenizer, LlamaForCausalLM
import torch

# Load Llama Model and tokenizer from Hugging Face Transformers
model_base_name = "meta-llama/Llama-2-7b-hf"
model = LlamaForCausalLM.from_pretrained(model_base_name)
tokenizer = LlamaTokenizer.from_pretrained(model_base_name)

# Specify input sentences
sentences = ["This is me", "A 2nd sentence"]

# Tokenize the input sentences with padding and truncation
input_ids = tokenizer(sentences, return_tensors='pt', padding=True, truncation=True, max_length=4096)['input_ids']

# Ensure token IDs are within the vocabulary range
input_ids = input_ids.clamp(max=tokenizer.vocab_size - 1)

# Get model outputs (logits)
with torch.no_grad():
    outputs = model(input_ids)

# Extract hidden states from the base model
hidden_states = outputs.logits

# Extract embeddings for [CLS] tokens
cls_embeddings = hidden_states[:, 0, :]

# Compute cosine similarity using torch.nn.functional.cosine_similarity
similarity = torch.nn.functional.cosine_similarity(cls_embeddings[0].unsqueeze(0), cls_embeddings[1].unsqueeze(0))
print("Cosine Similarity:", similarity.item())


Loading checkpoint shards: 100%|██████████| 2/2 [01:55<00:00, 57.98s/it]
Vocabulary Size: 32000
Max Sequence Length: 4096
Cosine Similarity: 0.9999995419367911

Process finished with exit code 0

By leveraging PyTorch's torch.nn.functional.cosine_similarity, the code calculates the cosine similarity between the [CLS] embeddings of the two input sentences. The outcome serves as an indicator of the semantic similarity between the sentences, where a value close to 1 signifies high similarity.

The resulting output presents the cosine similarity score for the given input sentences, showcasing their semantic relatedness. This code snippet illustrates the procedure of extracting embeddings from a pre-trained Llama model and assessing sentence similarity through cosine similarity computation.

Unpacking the Cosine Similarity Discrepancy

The Notable Contrast in Cosine Similarity Scores

The significant difference in cosine similarity scores, specifically 0.7132 for Jina and 0.9999 for Llama2, when evaluating the sentences "This is me" and "A 2nd sentence," prompts a closer examination. While it's essential to acknowledge that drawing definitive conclusions from a single data point is limited, it underscores the importance of investigating potential reasons for this divergence.

Potential Explanations

Model Focus

  • Jina: Primarily focuses on capturing nuanced semantic relationships between words and phrases, potentially penalizing the absence of shared vocabulary and semantic connections between the two sentences.
  • Llama2: A more expansive language model adept at handling intricate language tasks, potentially prioritizing the inherent self-referential nature of "This is me" and overlooking the lack of direct semantic overlap with "A 2nd sentence."

Training Data

  • Jina: Trained on extensive text corpora specifically emphasizing semantic relationships and contextual understanding, making it more attuned to subtle semantic differences.
  • Llama2: Trained on a diverse dataset covering various text formats, potentially prone to generalizing from simple self-referential statements, resulting in higher similarity scores even with limited overlap.


In the ever-evolving realm of natural language processing, the fusion of cutting-edge models like Jina Embeddings and the Llama Model with the user-friendly and versatile Hugging Face Transformers opens up avenues for groundbreaking applications. Jina Embeddings, rooted in the robust Bert architecture and refined through the ALiBi variant, provides developers with an opportunity to explore the intricacies of textual semantics. With its capacity for extended sequence lengths and meticulous curation of training data, it becomes a potent tool for tasks such as long document retrieval and semantic textual similarity. The seamless integration with Hugging Face Transformers ensures accessibility, enabling developers to effortlessly leverage the capabilities of this sophisticated model.

On another front, the Llama Model family, particularly Llama 2, showcases the capabilities of generative language models. Trained on extensive corpora and optimized for a variety of dialogue applications, Llama 2 models empower developers to create intelligent virtual assistants, customer support bots, and interactive dialogue systems. Its integration with Hugging Face Transformers simplifies the tokenization process, allowing developers to concentrate on crafting engaging conversations without the complexity of intricate model interactions

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