Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
change
sglang
Commits
b57d87c2
Unverified
Commit
b57d87c2
authored
Jun 14, 2025
by
fzyzcjy
Committed by
GitHub
Jun 14, 2025
Browse files
Fix shared experts fusion + weight requant (#7177)
parent
98538822
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
9 additions
and
8 deletions
+9
-8
python/sglang/srt/models/deepseek_v2.py
python/sglang/srt/models/deepseek_v2.py
+9
-8
No files found.
python/sglang/srt/models/deepseek_v2.py
View file @
b57d87c2
...
...
@@ -1960,14 +1960,15 @@ class DeepseekV2ForCausalLM(nn.Module):
)
if
layer_id
in
moe_layers
:
shared_experts
=
layer
.
mlp
.
shared_experts
for
module
in
[
shared_experts
.
gate_up_proj
,
shared_experts
.
down_proj
,
]:
requant_weight_ue8m0_inplace
(
module
.
weight
,
module
.
weight_scale_inv
,
weight_block_size
)
shared_experts
=
getattr
(
layer
.
mlp
,
"shared_experts"
,
None
)
if
shared_experts
is
not
None
:
for
module
in
[
shared_experts
.
gate_up_proj
,
shared_experts
.
down_proj
,
]:
requant_weight_ue8m0_inplace
(
module
.
weight
,
module
.
weight_scale_inv
,
weight_block_size
)
experts
=
layer
.
mlp
.
experts
if
isinstance
(
experts
,
DeepEPMoE
):
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment