How to reduce document intelligence latency ?

Question

How to reduce document intelligence latency ?

AMROUN Lysa 396

Hello,

I employ Azure Document Intelligence for document recognition using template/neural models. I've noticed that it takes approximately 5 to 6 seconds to recognize all fields. I'm interested in reducing this latency and would like to know if it's possible and how to achieve it.

thank you

Lysa

santoshkc 15,615 Reputation points Microsoft External Staff Moderator

2023-11-23T10:00:53.7233333+00:00
Hi @AMROUN Lysa ,

Thank you for reaching out to Microsoft Q&A forum!

To reduce the latency of document recognition using Azure Document Intelligence, you can try the following:

Use a smaller model: Smaller models have fewer parameters, which means they can be loaded faster and run faster.

Use a pre-built model: Try using a pre-built model. Pre-built models are optimized for performance and can be faster than custom models.

Use a GPU: Try using GPUs, which can accelerate the processing of models and reduce the latency of document recognition.

Use batch processing: If you are processing multiple documents, try using batch processing. Batch processing can reduce the overall processing time by processing multiple documents at once.

I hope this helps. Feel free to inquire for additional details or queries.
santoshkc 15,615 Reputation points Microsoft External Staff Moderator

2023-11-24T10:32:31.74+00:00

Hi @AMROUN Lysa ,

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution, please do share that same with the community as it can be helpful to others. Thank you!
AMROUN Lysa 396 Reputation points

2023-11-24T10:41:13.5966667+00:00

Hi @santoshkc

Thank's for your answer , it's ok for me :)

Answer accepted by question author

0 additional answers

Your answer

santoshkc 15,615 Reputation points Microsoft External Staff Moderator

2023-11-23T10:00:53.7233333+00:00

Hi @AMROUN Lysa ,

Thank you for reaching out to Microsoft Q&A forum!

To reduce the latency of document recognition using Azure Document Intelligence, you can try the following:

Use a smaller model: Smaller models have fewer parameters, which means they can be loaded faster and run faster.

Use a pre-built model: Try using a pre-built model. Pre-built models are optimized for performance and can be faster than custom models.

Use a GPU: Try using GPUs, which can accelerate the processing of models and reduce the latency of document recognition.

Use batch processing: If you are processing multiple documents, try using batch processing. Batch processing can reduce the overall processing time by processing multiple documents at once.

I hope this helps. Feel free to inquire for additional details or queries.
santoshkc 15,615 Reputation points Microsoft External Staff Moderator

2023-11-24T10:32:31.74+00:00

Hi @AMROUN Lysa ,

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution, please do share that same with the community as it can be helpful to others. Thank you!
AMROUN Lysa 396 Reputation points

2023-11-24T10:41:13.5966667+00:00

Hi @santoshkc

Thank's for your answer , it's ok for me :)

Answer 1

Hi @AMROUN Lysa ,

You're welcome! I'm glad I could help.

If others experiencing the same thing can easily reference this! Since the Microsoft Q&A community has a policy that "The question author cannot accept their own answer. They can only accept answers by others ", I'll repost the solution in case you'd like to accept the answer.
Question: How to reduce document intelligence latency ?
Answer: To reduce the latency of document recognition using Azure Document Intelligence, you can try the following:

Use a smaller model: Smaller models have fewer parameters, which means they can be loaded faster and run faster.
Use a pre-built model: Try using a pre-built model. Pre-built models are optimized for performance and can be faster than custom models.
Use a GPU: Try using GPUs, which can accelerate the processing of models and reduce the latency of document recognition.
Use batch processing: If you are processing multiple documents, try using batch processing. Batch processing can reduce the overall processing time by processing multiple documents at once.

Thank you for your understanding!

If this answers your query, do click Accept Answer and Yes for was this answer helpful.

Vince Kolb 195 Reputation points

2025-01-17T20:23:37.4066667+00:00

we are using document intelligence studio to create our custom DI models.
you mention use GPUs..... where / how can one specify what compute resource to use when the model is being used via API?