Test loading generation config with safetensor weights #31550
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Fix #29522
This PR changes the model used in a generation config loading test to a model that a) is used on other tests b) has a generation config. The purpose of the test is to check that the generation config is loaded correctly with the model (which broke in the past)
The model used for the test is a dummy model, and it has empty weights that break when we load the model with
device_map="auto"
(because we can call.to
with empty weights when the source is ameta
device).