Choose an off-the-shelf model, or upload your own and receive a cross-hardware performance analysis to your inbox, for free!
The methods to grant us secure access to benchmark your model:
We respect your privacy and confidentiality:
Model weights can be random.
We never keep a copy of the model after analysis
The inference performance of your deep learning model will determine the viability of your solution in a production environment. An inefficient and uneconomical model could determine the entire fate of your product. That said, estimating the performance impact is far from simple. The hardware's compute power in your production environment, as well as the influence of inefficient code, can obscure this critical estimation process.
We currently analyze model latency, throughput, cost per 100K samples on the cloud, and memory usage — all compared across multiple batch sizes and hardware types.
During our analysis, the DIPS engine runs a forward pass of the model and automatically checks its performance on several environments. It then provides you with a comprehensive analysis of the model’s performance along with actionable insights and recommendations. In the report, we specify the exact conditions of all measurements to ensure standardization and provide reproducible results.
We currently support:
Your framework isn’t listed? Let us know.
We currently support the following AWS instances:
Interested in a different cloud provider or other hardware? Let us know
Completing this wizard will take less than 2 minutes and you’ll get the report by mail within 24 hours.
Yes. This service and all costs associated with model inferencing are absorbed by Deci.
You have 3 options:
1. Share the model with an open link
2. Choose an existing off-the-shelf model from our repository
3. Be contacted by one of our specialists – we’ll contact you to take this process offline
These are some of the most known deep neural networks served from our repository, so that you don’t have to upload or share anything.
When choosing this option, we’ll provide you with a report of how one of the following models will perform across different HW:
Last updated: March, 15, 2020
DECI.AI LTD. (“DECI”, “We“, “Us” or “Our“) WELCOMES YOU (“CUSTOMER”) TO OUR SERVICE. THE FOLLOWING TERMS (THE “AGREEMENT”) STIPULATE THE TERMS AND CONDITIONS OF YOUR USE OF THE SERVICE. THE SERVICE IS PROVIDED SOLELY FOR USE IN YOUR OWN BUSINESS. BY ACCESSING OR USING OUR SERVICE, YOU AGREE TO THESE TERMS. IF YOU ARE ENTERING INTO THESE TERMS ON BEHALF OF A COMPANY OR ANOTHER LEGAL ENTITY, YOU AFFIRM THAT YOU HAVE THE AUTHORITY TO BIND SUCH ENTITY AND ITS AFFILIATES TO THESE TERMS, IN WHICH CASE THE TERM “YOU” OR “CUSTOMER” WILL REFER TO SUCH ENTITY AND ITS AFFILIATES. IF THE LEGAL ENTITY THAT YOU REPRESENT DOES NOT AGREE WITH THESE TERMS, YOU MUST NOT ACCEPT THESE TERMS OR USE THE SERVICE.