Skip to content

Support exaone4 model #1491

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

Open
wants to merge 6 commits into
base: main
Choose a base branch
from
+20 −0
Open

Support exaone4 model #1491

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
b0573c2
support exaone4 model
zhaohb Oct 24, 2025
1936b47
update docs
zhaohb Oct 24, 2025
0f27725
Merge branch 'huggingface:main' into support_exaone4_model
zhaohb Oct 27, 2025
0401fcb
add tests
zhaohb Oct 31, 2025
3be6a32
update
zhaohb Oct 31, 2025
cedcb5d
update
zhaohb Oct 31, 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
1 change: 1 addition & 0 deletions docs/source/openvino/models.mdx
View file
Open in desktop
Copy link
Collaborator

Choose a reason for hiding this comment

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

since we have already supported previous "exaone" version. Please add tests for quantization. It should work as well.

Copy link
Author

Choose a reason for hiding this comment

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

However, when applying quantization, feature-extraction works correctly, but text-classification throws an error:

nncf.errors.InvalidGroupSizeError: Compression of the '__module.model.rotary_emb/aten::matmul/MatMul' layer failed with the following error:
Channel size 1 should be divisible by size of group 128.
Ensure that the group size is divisible by the channel size, or include this node and others with similar issues in the ignored scope:
nncf.compress_weight(
        ...,
        ignored_scope=IgnoredScope(names=["__module.model.rotary_emb/aten::matmul/MatMul"]
        )
)

Copy link
Collaborator

Choose a reason for hiding this comment

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

However, when applying quantization, feature-extraction works correctly, but text-classification throws an error:

nncf.errors.InvalidGroupSizeError: Compression of the '__module.model.rotary_emb/aten::matmul/MatMul' layer failed with the following error:
Channel size 1 should be divisible by size of group 128.
Ensure that the group size is divisible by the channel size, or include this node and others with similar issues in the ignored scope:
nncf.compress_weight(
        ...,
        ignored_scope=IgnoredScope(names=["__module.model.rotary_emb/aten::matmul/MatMul"]
        )
)

What version of nncf you are using?

Copy link
Author

Choose a reason for hiding this comment

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

@nikita-savelyevv I don't think this error is related to the NNCF. It's more likely an issue with my dummy model.
I will double check that. thank you.

Copy link
Collaborator

Choose a reason for hiding this comment

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

@zhaohb It could be, because there was a fix addressing this issue. In any case I would suggest to update to NNCF v 2.18

Copy link
Author

Choose a reason for hiding this comment

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

@nikita-savelyevv
Hi, the issue disappeared after upgrading to version 2.18. It wasn't a problem with the dummy model after all.
Thank you!

Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,7 @@ Here is the list of the supported architectures :
- Encoder Decoder
- ESM
- Exaone
- Exaone4
- Falcon
- Falcon-Mamba
- Flaubert
Expand Down
14 changes: 14 additions & 0 deletions optimum/exporters/openvino/model_configs.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -592,6 +592,20 @@ class GptOssOpenVINOConfig(LlamaOpenVINOConfig):
class ExaoneOpenVINOConfig(LlamaOpenVINOConfig):
pass

@register_in_tasks_manager(
"exaone4",
*[
"feature-extraction",
"feature-extraction-with-past",
"text-generation",
"text-generation-with-past",
"text-classification",
],
library_name="transformers",
)
class Exaone4OpenVINOConfig(LlamaOpenVINOConfig):
MIN_TRANSFORMERS_VERSION = "4.54.0"


@register_in_tasks_manager(
"arcee",
Expand Down
3 changes: 3 additions & 0 deletions tests/openvino/test_decoder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,6 +83,7 @@ class OVModelForCausalLMIntegrationTest(unittest.TestCase):
"phi3",
"gemma2",
"exaone",
"exaone4",
"granite",
"granite-moe",
)
Expand Down Expand Up @@ -142,6 +143,7 @@ class OVModelForCausalLMIntegrationTest(unittest.TestCase):
"arctic",
"chatglm4",
"exaone",
"exaone4",
"decilm",
"minicpm3",
"deepseek",
Expand Down Expand Up @@ -196,6 +198,7 @@ class OVModelForCausalLMIntegrationTest(unittest.TestCase):
"phi3": 2,
"gemma2": 4,
"exaone": 8,
"exaone4": 1,
"granite": 6,
"granite-moe": 6,
"glm": 28,
Expand Down
1 change: 1 addition & 0 deletions tests/openvino/utils_tests.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,6 +78,7 @@
"electra": "optimum-intel-internal-testing/tiny-random-electra",
"esm": "optimum-intel-internal-testing/tiny-random-EsmModel",
"exaone": "optimum-intel-internal-testing/tiny-random-exaone",
"exaone4": "optimum-intel-internal-testing/tiny-random-exaone4",
"gemma": "optimum-intel-internal-testing/tiny-random-GemmaForCausalLM",
"gemma2": "optimum-intel-internal-testing/tiny-random-gemma2",
"got_ocr2": "optimum-intel-internal-testing/tiny-random-got-ocr2-hf",
Expand Down
Loading