Skip to content

Swap Llava export arg order #14238

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.

Sign up for GitHub

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

Merged
merged 4 commits into from
Sep 13, 2025
Merged

Swap Llava export arg order #14238

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
2dd30f5
Swap Llava export arg order
jackzhxng Sep 11, 2025
59cfdb7
Update runner to match
jackzhxng Sep 12, 2025
5c2d7a0
Fix tests
jackzhxng Sep 12, 2025
3a800cb
Fix test
jackzhxng Sep 12, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions examples/models/llava/export_llava.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,7 +77,7 @@ def __init__(self, llava):
super().__init__()
self.text_model = llava.text_model

def forward(self, input_pos, embeddings):
def forward(self, embeddings, input_pos):
return self.text_model(None, {"input_pos": input_pos}, embeddings)

llava_text_model = LlavaTextModel(llava)
Expand All @@ -88,7 +88,7 @@ def forward(self, input_pos, embeddings):
max_seq_len=llava.text_model_args.max_seq_len,
dtype=DType.fp32,
use_kv_cache=True,
example_inputs=(torch.tensor([0], dtype=torch.int64), embeddings),
example_inputs=(embeddings, torch.tensor([0], dtype=torch.int64)),
dynamic_shapes=dynamic_shapes,
)

Expand Down
2 changes: 1 addition & 1 deletion examples/models/llava/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -405,5 +405,5 @@ def _get_image_dynamic_shapes(self):

def _get_prompt_dynamic_shapes(self):
dim = torch.export.Dim("token_dim", min=2, max=self.max_seq_len)
text_model_dynamic_shapes = ({0: 1}, {1: dim})
text_model_dynamic_shapes = ({1: dim}, {0: 1})
Copy link
Contributor

@JacobSzwejbka JacobSzwejbka Sep 12, 2025

Choose a reason for hiding this comment

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

nit: 1 and 0 are the dims of a single tensor. So swapping doesnt do anything

Copy link
Contributor

@JacobSzwejbka JacobSzwejbka Sep 12, 2025

Choose a reason for hiding this comment

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

also do you even have to specify that the batch is static 1?

Copy link
Contributor Author

@jackzhxng jackzhxng Sep 13, 2025

Choose a reason for hiding this comment

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

{0:1} is the dynamic shapes for the first arg, {1:dim} is for the second arg, so I swap them here

return text_model_dynamic_shapes
Loading