Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat: simple mistral lora integration tests #2180

Merged
merged 5 commits into from
Jul 15, 2024
Merged

Conversation

drbh
Copy link
Collaborator

@drbh drbh commented Jul 3, 2024

This PR adds integration tests for mistralai/Mistral-7B-v0.1 with two lora adapters predibase/dbpedia and predibase/customer_support. These test will be improved in the future, and exist to ensure that basic lora functionality works as expected

"predibase/dbpedia",
"predibase/customer_support",
],
cuda_graphs=[0],
Copy link
Collaborator

Choose a reason for hiding this comment

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

Why ?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

currently theres a bug with lora and cuda graphs but wanted to make sure there were basic tests for expected responses. I'll push a update for cuda graphs soon along with some of the other refactoring changes. Happy to keep this open until it's resolved

Copy link
Collaborator

Choose a reason for hiding this comment

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

What is the bug ?

Does load with many different loader breaks things ?

Copy link
Collaborator Author

@drbh drbh Jul 4, 2024

Choose a reason for hiding this comment

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

currently the adapter is ignored when graphs are enabled because adapter_data is passed to model.forward and is not included in cuda_graph_warmup. I need to update the graph warmup to include adapter data and also replace the values before calling replay

@drbh
Copy link
Collaborator Author

drbh commented Jul 8, 2024

this PR has been updated to disable CUDA_GRAPHS if lora adapters are enabled. In the future we'll add support however until support is added we will disable and warn the user

@drbh drbh force-pushed the integration-test-lora branch from f3e6041 to 1ebcc9d Compare July 8, 2024 18:16
@drbh drbh merged commit 5a65066 into main Jul 15, 2024
9 checks passed
@drbh drbh deleted the integration-test-lora branch July 15, 2024 13:16
ErikKaum pushed a commit that referenced this pull request Jul 25, 2024
* feat: simple mistral lora integration tests

* fix: include args in docker launcher

* fix: disable cuda graphs with lora and warn

* fix: adjust docs and precommit issues

* fix: re update docs
ErikKaum pushed a commit that referenced this pull request Jul 26, 2024
* feat: simple mistral lora integration tests

* fix: include args in docker launcher

* fix: disable cuda graphs with lora and warn

* fix: adjust docs and precommit issues

* fix: re update docs
yuanwu2017 pushed a commit to yuanwu2017/tgi-gaudi that referenced this pull request Sep 26, 2024
* feat: simple mistral lora integration tests

* fix: include args in docker launcher

* fix: disable cuda graphs with lora and warn

* fix: adjust docs and precommit issues

* fix: re update docs
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants