Clusters
At Tensorium, we understand that the heart of any AI-related project is a powerful and reliable computational infrastructure. That's why we have developed a range of high-performance clusters designed to provide optimal power and efficiency for your tasks
Our clusters are equipped with the latest generation of GPUs from leading chip manufacturer Nvidia, ensuring the highest data processing and analysis speeds. We offer various cluster configurations tailored to specific needs and tasks
CPU Basic
Our basic solution for small teams virtual processors deployed on Intel Xeon processors
CPU
2 vCPU
Memory
16 Gb
GPU Memory
-
Rental
$0.03/hour
Scalability
Our clusters are easily scalable, allowing you to increase computational power according to the growing demands of your project
Flexibility
Our clients can choose the optimal cluster configuration based on their unique needs and budget
Efficiency
Thanks to hardware-level optimization, our clusters provide maximum performance with minimal energy consumption
Enterprise level
Our entire ecosystem is built from the ground up and optimized for enterprise workloads
Innovations at Tensorium
Our team pays special attention to developing and implementing the latest advancements in the world of AI, ensuring our clients are always at the forefront of technological progress.
Images Generated Using Stable Diffusion Model on Nvidia A100 Clusters
AI Models
Learn about the latest AI models that have been successfully trained and tested on our clusters
Pixboost
With Pixboost, you can efficiently upscale input images. The chain of hidden pipelines significantly enhances noise reduction parameters compared to other models
Model description
First, the base Pixboost model is used to generate hidden data of the desired output size. In the second stage, we use a specialized high-resolution model and apply a method called PixboostEdit to the hidden objects generated in the first stage to produce the final output
35 mln. parameters
Cluster: Nvidia T4
Documentary
Documentary was developed to understand document images (e.g., invoices). This model is refined based on DocVQA, a dataset for visual question answering on documents
Model description
Documentary consists of an image encoder (Swin Transformer) and a text decoder (BART). Upon receiving an image, the encoder first converts it into an embedding tensor (shape: batch_size, seq_len, hidden_size), after which the decoder autoregressively generates text based on the encoding created by the encoder
56 mln. parameters
Cluster: Nvidia A40
MaskCraft
MaskCraft creates high-quality object masks from input prompts such as points or rectangles and can be used to generate masks for all objects in an image
Model description
MaskCraft consists of an image encoder that performs cross-attention between image embeddings and point embeddings, and between point embeddings and contextualized masks. The mask decoder then predicts the output masks based on the contextualized masks produced by the image encoder
92 mln. parameters
Cluster: Nvidia A100
Meshwork
Meshwork transforms a single image into a 3D textured mesh using a convolutional reconstruction model. It allows for the creation of 3D models of objects from an image
Model description
Meshwork includes a diffusion model to create multi-view images from a single input image, another diffusion model to create CCM, and a reconstruction model based on UNet to obtain the final textured mesh
140 mln. parameters
Cluster: Nvidia H100
Security
At Tensorium, we take pride in our commitment to the highest standards of quality and security. Our company is certified to the international ISO 27001 standard, which confirms our dedication to ensuring the security of our clients' information and data. We securely protect:
User Data: Personal information, transaction history, and all user data are protected in accordance with international security standards
Infrastructure: The physical and virtual infrastructure of Tensorium is reliably protected against external threats and attacks, including DDoS attacks and malware
Transactions: All financial operations and transactions are conducted through encrypted communication channels, ensuring their confidentiality and integrity
Connections and networks: All network connections used by Tensorium are secured with modern encryption protocols, including SSL/TLS, to protect data during transmission
Coming soon
Tensorium whitepaper
In our whitepaper, we delve deeply into the technical aspects and strategies underpinning Tensorium's operations. The most important topics covered in the whitepaper include:
Technological foundations: An overview of our clusters, including architecture, hardware, and software used
Innovative developments: Insights into the latest AI innovations that we integrate into our systems
Use cases: Analysis of how our technologies are applied across various industries, from healthcare to financial services
Future of Tensorium: Our vision for the future of AI and plans for further development of our technologies and expansion of our service offerings
