Skip to content

Conversation

@kelbrown20
Copy link
Contributor

@kelbrown20 kelbrown20 commented May 13, 2025

This PR adds new AI terms to the style guide glossary

Copy link
Collaborator

@bergerhoffer bergerhoffer left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Here's a first pass at a few bits of feedback!

@bergerhoffer bergerhoffer added the Glossary entry Topics that add or modify glossary entries label May 15, 2025
@kelbrown20 kelbrown20 force-pushed the ai-glossary-updates-v1 branch from 57f011f to 428a59e Compare May 28, 2025 19:33
@bergerhoffer
Copy link
Collaborator

@kelbrown20 Is this ready for a re-review?

@kelbrown20 kelbrown20 force-pushed the ai-glossary-updates-v1 branch from 428a59e to cd48339 Compare June 24, 2025 13:28
@kelbrown20
Copy link
Contributor Author

My apologies for the delay @bergerhoffer!
Yes I believe I fixed everything, but def lmk if there was something I might have missed

Copy link
Contributor

@sbmetz sbmetz left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Some minor wording tweaks

*See also*:

[[inference-serving]]
==== image:images/yes.png[yes] inference serving (verb)
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So is "inferencing" accepted verb usage for serving models? For example, "inferencing the quantized granite model".

Copy link
Contributor Author

@kelbrown20 kelbrown20 Jul 23, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

So this is a great point, early on when researching this, I rarely ever saw usage of the the verb form "inferencing" , but Ive been seeing it more and more and It seems like its a standard. I think part of the confusing part was it seemed like PMs were using "inference serving" primarily as the single word for both inferencing and serving, which still works for certain contexts. So I'll keep this term, update the description to make that more clear, and make "inferencing" an option

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Seems jargony to me, vs "performing/running inference" or "inference serving". (ChatGPT and Gemini support this analysis 😃 )

Copy link
Member

@aireilly aireilly Jul 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Both appear widely used in academic papers.

Copy link
Contributor Author

@kelbrown20 kelbrown20 Jul 24, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

From what I see, I think the its the noun form vs the verb form of inferencing. I saw inferencing used more as a noun when going through those a few of those articles. I do think I need to update these but @aireilly what do you think of the new defs with examples

@kelbrown20 kelbrown20 force-pushed the ai-glossary-updates-v1 branch from cd48339 to 0aa5764 Compare July 24, 2025 14:58
Copy link
Member

@aireilly aireilly left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Added some suggestions


*See also*:

[[inferencing]]
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
[[inferencing]]
[[inference serving]]

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

"Inference serving" seems to be the proper form.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

They seem to both be correct terms, but they have different meanings.

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

OK. So maybe we need "inferencing" and "inference serving"?

Inferencing - the process of running a model
Inference Serving - deploying the inferencing capability on a server

Something like this?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, that looks about right.


[[inference]]
==== image:images/yes.png[yes] inference (noun)
*Description*: The act a model generating outputs from input data. For example, "Inference speeds increased on the new models"
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
*Description*: The act a model generating outputs from input data. For example, "Inference speeds increased on the new models"
*Description*: The process in which a trained model is loaded into memory and generates output based on input data.
For example, "The Llama-3.2-90B-Vision-Instruct-FP8-dynamic model performs inference to identify objects in an image."

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@aireilly This is an improvement in the definition. Thanks. I'd like to rewrite this a bit. In general, we write definitions as complete sentences. How's this?

Suggested change
*Description*: The act a model generating outputs from input data. For example, "Inference speeds increased on the new models"
*Description*: AI inference is the process in which a trained model is loaded into memory and then the makes predictions or performs tasks on new data. For example, "The Llama-3.2-90B-Vision-Instruct-FP8-dynamic model performs inference to identify objects in an image."

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
*Description*: The act a model generating outputs from input data. For example, "Inference speeds increased on the new models"
*Description*: AI inference is the process in which a trained model is loaded into memory and then makes predictions based on input data. For example, "The Llama-3.2-90B-Vision-Instruct-FP8-dynamic model performs inference to identify objects in an image."

What about this?


[[inferencing]]
==== image:images/yes.png[yes] inferencing (noun)
*Description*: A process by which a model processes input data, deduce information, and generates an output. For example, "The inferencing workload is distributed across multiple accelerators."
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
*Description*: A process by which a model processes input data, deduce information, and generates an output. For example, "The inferencing workload is distributed across multiple accelerators."
*Description*: The act of deploying and running a trained model so that it can process input data and generate output.
For example, "Use vLLM to inference serve a trained model."

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@aireilly Both definitions are correct. We just have to decide if we want to put both "inferencing" and "inference serving" in the SSG or choose one over the other.

Copy link
Collaborator

@IngridT1 IngridT1 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I won't be able to merge these without some discussion and changes.


[[inference]]
==== image:images/yes.png[yes] inference (noun)
*Description*: The act a model generating outputs from input data. For example, "Inference speeds increased on the new models"
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@aireilly This is an improvement in the definition. Thanks. I'd like to rewrite this a bit. In general, we write definitions as complete sentences. How's this?

Suggested change
*Description*: The act a model generating outputs from input data. For example, "Inference speeds increased on the new models"
*Description*: AI inference is the process in which a trained model is loaded into memory and then the makes predictions or performs tasks on new data. For example, "The Llama-3.2-90B-Vision-Instruct-FP8-dynamic model performs inference to identify objects in an image."


*See also*:

[[inferencing]]
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

They seem to both be correct terms, but they have different meanings.


[[inferencing]]
==== image:images/yes.png[yes] inferencing (noun)
*Description*: A process by which a model processes input data, deduce information, and generates an output. For example, "The inferencing workload is distributed across multiple accelerators."
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@aireilly Both definitions are correct. We just have to decide if we want to put both "inferencing" and "inference serving" in the SSG or choose one over the other.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Glossary entry Topics that add or modify glossary entries

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants