Skip to content

Add handling for Qwen3VLMoe on older transformers versions #2040

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
Nov 17, 2025
+23 −13
Merged

Add handling for Qwen3VLMoe on older transformers versions #2040

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
017fec0
Add handling for Qwen3VLMoe on older transformers versions
fynnsu Nov 17, 2025
7a5c90e
Switch to pytest.importorskip
fynnsu Nov 17, 2025
9027587
Use double quotes on type annotation
fynnsu Nov 17, 2025
d1512ad
Merge branch 'main' into fix_qwen3vlmoe
fynnsu Nov 17, 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
18 changes: 12 additions & 6 deletions src/llmcompressor/modeling/qwen3_vl_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,12 +1,18 @@
from __future__ import annotations

from typing import TYPE_CHECKING

import torch
from transformers import Qwen3VLMoeConfig, Qwen3VLMoeTextConfig
from transformers.models.qwen3_vl_moe.modeling_qwen3_vl_moe import (
Qwen3VLMoeTextSparseMoeBlock as OriginalQwen3VLMoeTextSparseMoeBlock,
)

from llmcompressor.modeling.moe_context import MoECalibrationModule
from llmcompressor.utils.dev import skip_weights_initialize

if TYPE_CHECKING:
from transformers import Qwen3VLMoeConfig, Qwen3VLMoeTextConfig
from transformers.models.qwen3_vl_moe.modeling_qwen3_vl_moe import (
Qwen3VLMoeTextSparseMoeBlock,
)


@MoECalibrationModule.register("Qwen3VLMoeTextSparseMoeBlock")
class CalibrateQwen3VLMoeTextSparseMoeBlock(MoECalibrationModule):
Expand All @@ -19,7 +25,7 @@ class CalibrateQwen3VLMoeTextSparseMoeBlock(MoECalibrationModule):

def __init__(
self,
original: OriginalQwen3VLMoeTextSparseMoeBlock,
original: Qwen3VLMoeTextSparseMoeBlock,
config: Qwen3VLMoeConfig,
calibrate_all_experts: bool,
):
Expand Down Expand Up @@ -116,7 +122,7 @@ def __init__(self, config, original):

def replace(
config: Qwen3VLMoeConfig,
original: OriginalQwen3VLMoeTextSparseMoeBlock,
original: Qwen3VLMoeTextSparseMoeBlock,
calibrate_all_experts: bool,
):
return CalibrateQwen3VLMoeTextSparseMoeBlock(
Expand Down
18 changes: 14 additions & 4 deletions tests/llmcompressor/modeling/test_calib_qwen3_vl_moe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,14 +1,24 @@
import pytest
import torch
from transformers import Qwen3VLMoeConfig
from transformers.models.qwen3_vl_moe.modeling_qwen3_vl_moe import (
Qwen3VLMoeTextSparseMoeBlock,
)

from llmcompressor.modeling.qwen3_vl_moe import CalibrateQwen3VLMoeTextSparseMoeBlock
from llmcompressor.utils.helpers import calibration_forward_context
from tests.testing_utils import requires_gpu

try:
from transformers import Qwen3VLMoeConfig
from transformers.models.qwen3_vl_moe.modeling_qwen3_vl_moe import (
Qwen3VLMoeTextSparseMoeBlock,
)
except ImportError:
Qwen3VLMoeConfig = None
Qwen3VLMoeTextSparseMoeBlock = None


@pytest.mark.skipif(
Qwen3VLMoeConfig is None,
reason="Qwen3VLMoe not available in this version of transformers",
)
@requires_gpu
def test_calib_qwen3_vl_moe_module():
config = Qwen3VLMoeConfig()
Expand Down
Loading