Skip to content

Conversation

@davidkyle
Copy link
Member

The change from batch to iterative processing in elastic/ml-cpp#2538 had a side effect of changing the response format for Question Answering models where the input is longer than the max sequence length and must be spanned over multiple windows.

Previously the output was a 3d array of doubles with the 2 elements in the first dimension structured as:

[start][span_index][scores] [end][span_index][scores] 

Where start is always 0, end always 1 and span_index is no greater than the number of spans/windows

The new format combines the span index and start/end in the 1st dimension and the 2nd dimension is always of size 1.

[span_1_start][0][scores] [span_1_end][0][scores] [span_2_start][0][scores] [span_2_end][0][scores] ... 

Closes #97917

@davidkyle davidkyle added >bug :ml Machine learning auto-backport-and-merge cloud-deploy Publish cloud docker image for Cloud-First-Testing v8.10.0 v8.9.1 labels Aug 3, 2023
@elasticsearchmachine
Copy link
Collaborator

Pinging @elastic/ml-core (Team:ML)

@elasticsearchmachine elasticsearchmachine added the Team:ML Meta label for the ML team label Aug 3, 2023
@elasticsearchmachine
Copy link
Collaborator

Hi @davidkyle, I've created a changelog YAML for you.

Copy link

@droberts195 droberts195 left a comment

Choose a reason for hiding this comment

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

LGTM

@elasticsearchmachine
Copy link
Collaborator

💚 Backport successful

Status Branch Result
8.9
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

>bug cloud-deploy Publish cloud docker image for Cloud-First-Testing :ml Machine learning Team:ML Meta label for the ML team v8.9.1 v8.10.0

3 participants