[V1] Use input_ids as input for text-only models #11032
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Outdated suggestions cannot be applied.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.
Currently, the model runner in V1 always uses
input_embeds
rather thaninput_ids
for the compatibility with multi-modal models. However, this excludes the embedding layer from the CUDA graph, and thus causes slight performance regression (and larger performance regression when TP > 1). This PR addresses this by using theinput_ids
for text-only models while keeping usinginput_embeds
for multi-modal models.