Skip to content

Qwen3moe #406

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 10 commits into
base: main
Choose a base branch
from
+508 −0
Open

Qwen3moe #406

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
b10508f
Onboarding qwen3moe
qcdipankar May 15, 2025
a497e33
Dynamic code for Moe Block
qcdipankar May 16, 2025
65b0eab
Fix for Repopointer issue
qcdipankar May 16, 2025
03df777
upgrading transformer to 4.51.3
qcdipankar May 21, 2025
1ba8170
upgrading transformer to 4.51.0
qcdipankar May 21, 2025
1d58304
Merge branch 'main' into qwen3moe
qcdipankar Jun 17, 2025
4b9b65e
Merge branch 'main' into qwen3moe
qcdipankar Jun 23, 2025
a82861b
Cleanup done for qwen3moe
qcdipankar Jul 1, 2025
a1e7af1
Merge branch 'quic:main' into qwen3moe
qcdipankar Jul 1, 2025
b8c50bf
Merge branch 'quic:main' into qwen3moe
qcdipankar Jul 2, 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
25 changes: 25 additions & 0 deletions QEfficient/transformers/models/pytorch_transforms.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -136,6 +136,15 @@
Qwen2Model,
Qwen2RMSNorm,
)
from transformers.models.qwen3_moe.modeling_qwen3_moe import (
Qwen3MoeAttention,
Qwen3MoeDecoderLayer,
Qwen3MoeForCausalLM,
Qwen3MoeModel,
Qwen3MoeRMSNorm,
Qwen3MoeRotaryEmbedding,
Qwen3MoeSparseMoeBlock,
)
from transformers.models.starcoder2.modeling_starcoder2 import (
Starcoder2Attention,
Starcoder2DecoderLayer,
Expand Down Expand Up @@ -303,6 +312,14 @@
QEffQwen2ForCausalLM,
QEffQwen2Model,
)
from QEfficient.transformers.models.qwen3_moe.modeling_qwen3_moe import (
QEffQwen3MoeAttention,
QEffQwen3MoeDecoderLayer,
QEffQwen3MoeForCausalLM,
QEffQwen3MoeModel,
QEffQwen3MoeRotaryEmbedding,
QEffQwen3MoeSparseMoeBlock,
)
from QEfficient.transformers.models.starcoder2.modeling_starcoder2 import (
QEffStarcoder2Attention,
QEFFStarcoder2DecoderLayer,
Expand Down Expand Up @@ -338,6 +355,7 @@ class CustomOpsTransform(ModuleMappingTransform):
MllamaTextRMSNorm: CustomRMSNormAIC,
GraniteRMSNorm: CustomRMSNormAIC,
GraniteMoeRMSNorm: CustomRMSNormAIC,
Qwen3MoeRMSNorm: CustomRMSNormAIC,
Gemma3RMSNorm: QEffGemma3CustomRMSNormAIC,
}

Expand Down Expand Up @@ -388,6 +406,13 @@ class KVCacheTransform(ModuleMappingTransform):
GemmaDecoderLayer: QEffGemmaDecoderLayer,
GemmaModel: QEffGemmaModel,
GemmaForCausalLM: QEffGemmaForCausalLM,
# Qwen3Moe
Copy link
Contributor

Choose a reason for hiding this comment

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

Also updated ReadMe on what all official models will be supported as part of enabling this arch. thanks

Qwen3MoeForCausalLM: QEffQwen3MoeForCausalLM,
Qwen3MoeModel: QEffQwen3MoeModel,
Qwen3MoeDecoderLayer: QEffQwen3MoeDecoderLayer,
Qwen3MoeAttention: QEffQwen3MoeAttention,
Qwen3MoeRotaryEmbedding: QEffQwen3MoeRotaryEmbedding,
Qwen3MoeSparseMoeBlock: QEffQwen3MoeSparseMoeBlock,
# Gemma2
Gemma2Attention: QEffGemma2Attention,
Gemma2DecoderLayer: QEffGemma2DecoderLayer,
Expand Down
6 changes: 6 additions & 0 deletions QEfficient/transformers/models/qwen3_moe/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,6 @@
# -----------------------------------------------------------------------------
#
# Copyright (c) 2025 Qualcomm Innovation Center, Inc. All rights reserved.
# SPDX-License-Identifier: BSD-3-Clause
#
# -----------------------------------------------------------------------------
Loading
Loading