Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
Menu
Open sidebar
OpenDAS
dynamo
Commits
640c2d30
Unverified
Commit
640c2d30
authored
Oct 23, 2025
by
Hongkuan Zhou
Committed by
GitHub
Oct 23, 2025
Browse files
fix: sglang dsr1 recipes (#3850)
Signed-off-by:
hongkuanz
<
hongkuanz@nvidia.com
>
parent
8de469cf
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
102 additions
and
70 deletions
+102
-70
recipes/deepseek-r1/sglang-wideep/tep16p-dep16d-disagg.yaml
recipes/deepseek-r1/sglang-wideep/tep16p-dep16d-disagg.yaml
+53
-36
recipes/deepseek-r1/sglang-wideep/tep8p-dep8d-disagg.yaml
recipes/deepseek-r1/sglang-wideep/tep8p-dep8d-disagg.yaml
+49
-34
No files found.
recipes/deepseek-r1/sglang-wideep/tep16p-dep16d-disagg.yaml
View file @
640c2d30
...
@@ -45,29 +45,38 @@ spec:
...
@@ -45,29 +45,38 @@ spec:
path
:
/health
path
:
/health
port
:
9090
port
:
9090
periodSeconds
:
10
periodSeconds
:
10
timeoutSeconds
:
1
80
0
timeoutSeconds
:
10
failureThreshold
:
60
failureThreshold
:
60
0
image
:
my-registry/sglang-wideep-runtime:my-tag
image
:
my-registry/sglang-wideep-runtime:my-tag
workingDir
:
/workspace/components/backends/sglang
workingDir
:
/workspace/components/backends/sglang
command
:
command
:
-
/bin/sh
-
python3
-
-c
-
-m
-
dynamo.sglang
args
:
args
:
-
>-
-
--model-path
exec python3 -m dynamo.sglang
-
deepseek-ai/DeepSeek-R1
--model-path deepseek-ai/DeepSeek-R1
-
--served-model-name
--served-model-name deepseek-ai/DeepSeek-R1
-
deepseek-ai/DeepSeek-R1
--tp 16
-
--tp
--dp 16
-
"
16"
--enable-dp-attention
-
--dp
--ep-size 16
-
"
16"
--trust-remote-code
-
--enable-dp-attention
--skip-tokenizer-init
-
--ep-size
--disaggregation-mode decode
-
"
16"
--disaggregation-transfer-backend nixl
-
--trust-remote-code
--disaggregation-bootstrap-port 30001
-
--skip-tokenizer-init
--mem-fraction-static 0.8
-
--disaggregation-mode
--host 0.0.0.0
-
decode
-
--disaggregation-transfer-backend
-
nixl
-
--disaggregation-bootstrap-port
-
"
30001"
-
--mem-fraction-static
-
"
0.75"
-
--host
-
0.0.0.0
prefill
:
prefill
:
dynamoNamespace
:
sgl-dsr1-16gpu
dynamoNamespace
:
sgl-dsr1-16gpu
componentType
:
worker
componentType
:
worker
...
@@ -89,24 +98,32 @@ spec:
...
@@ -89,24 +98,32 @@ spec:
path
:
/health
path
:
/health
port
:
9090
port
:
9090
periodSeconds
:
10
periodSeconds
:
10
timeoutSeconds
:
1
80
0
timeoutSeconds
:
10
failureThreshold
:
60
failureThreshold
:
60
0
image
:
my-registry/sglang-wideep-runtime:my-tag
image
:
my-registry/sglang-wideep-runtime:my-tag
workingDir
:
/workspace/components/backends/sglang
workingDir
:
/workspace/components/backends/sglang
command
:
command
:
-
/bin/sh
-
python3
-
-c
-
-m
-
dynamo.sglang
args
:
args
:
-
>-
-
--model-path
exec python3 -m dynamo.sglang
-
deepseek-ai/DeepSeek-R1
--model-path deepseek-ai/DeepSeek-R1
-
--served-model-name
--served-model-name deepseek-ai/DeepSeek-R1
-
deepseek-ai/DeepSeek-R1
--tp 16
-
--tp
--ep-size 16
-
"
16"
--trust-remote-code
-
--ep-size
--skip-tokenizer-init
-
"
16"
--disaggregation-mode prefill
-
--trust-remote-code
--disaggregation-transfer-backend nixl
-
--skip-tokenizer-init
--disaggregation-bootstrap-port 30001
-
--disaggregation-mode
--mem-fraction-static 0.8
-
prefill
--host 0.0.0.0
-
--disaggregation-transfer-backend
\ No newline at end of file
-
nixl
-
--disaggregation-bootstrap-port
-
"
30001"
-
--mem-fraction-static
-
"
0.75"
-
--host
-
0.0.0.0
\ No newline at end of file
recipes/deepseek-r1/sglang-wideep/tep8p-dep8d-disagg.yaml
View file @
640c2d30
...
@@ -43,28 +43,36 @@ spec:
...
@@ -43,28 +43,36 @@ spec:
path
:
/health
path
:
/health
port
:
9090
port
:
9090
periodSeconds
:
10
periodSeconds
:
10
timeoutSeconds
:
1
80
0
timeoutSeconds
:
10
failureThreshold
:
60
failureThreshold
:
60
0
image
:
my-registry/sglang-wideep-runtime:my-tag
image
:
my-registry/sglang-wideep-runtime:my-tag
workingDir
:
/workspace/components/backends/sglang
workingDir
:
/workspace/components/backends/sglang
command
:
command
:
-
/bin/sh
-
python3
-
-c
-
-m
-
dynamo.sglang
args
:
args
:
-
>-
-
--model-path
exec python3 -m dynamo.sglang
-
deepseek-ai/DeepSeek-R1
--model-path deepseek-ai/DeepSeek-R1
-
--served-model-name
--served-model-name deepseek-ai/DeepSeek-R1
-
deepseek-ai/DeepSeek-R1
--tp 8
-
--tp
--dp 8
-
"
8"
--enable-dp-attention
-
--dp
--ep-size 8
-
"
8"
--trust-remote-code
-
--enable-dp-attention
--skip-tokenizer-init
-
--ep-size
--disaggregation-mode decode
-
"
8"
--disaggregation-transfer-backend nixl
-
--trust-remote-code
--disaggregation-bootstrap-port 30001
-
--skip-tokenizer-init
--host 0.0.0.0
-
--disaggregation-mode
-
decode
-
--disaggregation-transfer-backend
-
nixl
-
--disaggregation-bootstrap-port
-
"
30001"
-
--host
-
0.0.0.0
prefill
:
prefill
:
dynamoNamespace
:
sgl-dsr1-8gpu
dynamoNamespace
:
sgl-dsr1-8gpu
componentType
:
worker
componentType
:
worker
...
@@ -84,23 +92,30 @@ spec:
...
@@ -84,23 +92,30 @@ spec:
path
:
/health
path
:
/health
port
:
9090
port
:
9090
periodSeconds
:
10
periodSeconds
:
10
timeoutSeconds
:
1
80
0
timeoutSeconds
:
10
failureThreshold
:
60
failureThreshold
:
60
0
image
:
my-registry/sglang-wideep-runtime:my-tag
image
:
my-registry/sglang-wideep-runtime:my-tag
workingDir
:
/workspace/components/backends/sglang
workingDir
:
/workspace/components/backends/sglang
command
:
command
:
-
/bin/sh
-
python3
-
-c
-
-m
-
dynamo.sglang
args
:
args
:
-
>-
-
--model-path
exec python3 -m dynamo.sglang
-
deepseek-ai/DeepSeek-R1
--model-path deepseek-ai/DeepSeek-R1
-
--served-model-name
--served-model-name deepseek-ai/DeepSeek-R1
-
deepseek-ai/DeepSeek-R1
--tp 8
-
--tp
--ep-size 8
-
"
8"
--trust-remote-code
-
--ep-size
--skip-tokenizer-init
-
"
8"
--disaggregation-mode prefill
-
--trust-remote-code
--disaggregation-transfer-backend nixl
-
--skip-tokenizer-init
--disaggregation-bootstrap-port 30001
-
--disaggregation-mode
--host 0.0.0.0
-
prefill
\ No newline at end of file
-
--disaggregation-transfer-backend
-
nixl
-
--disaggregation-bootstrap-port
-
"
30001"
-
--host
-
0.0.0.0
\ No newline at end of file
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