Skip to content

Commit 35c8aed

Browse files
YouNeedCryDearYouNeedCryDear
committed
use -1 for max model len and update smg to 1.4.0
1 parent 9a1bc8b commit 35c8aed

14 files changed

Lines changed: 28 additions & 44 deletions

‎config/runtimes/vllm/Qwen/qwen-35-tp1-fp8-rt.yaml‎

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -61,7 +61,7 @@ spec:
6161
- BM.GPU.H200.8
6262
runner:
6363
name: ome-container
64-
image: fra.ocir.io/idqj093njucb/smg:1.3.3-vllm-v0.18.0
64+
image: fra.ocir.io/idqj093njucb/smg:1.4.0-vllm-v0.18.0
6565
ports:
6666
- containerPort: 8080
6767
name: http1
@@ -80,13 +80,12 @@ spec:
8080
--served-model-name vllm-model \
8181
--limit-mm-per-prompt '{"video": {"count": 1}, "image": {"count": 10}}' \
8282
--generation-config vllm \
83-
--optimization-level 3 \
83+
--max-model-len -1 \
8484
--router-tool-call-parser qwen_coder \
8585
--router-reasoning-parser qwen3_thinking \
8686
--backend vllm \
8787
--router-log-json \
8888
--worker-startup-timeout 1000 \
89-
--router-max-concurrent-requests 512 \
9089
--router-request-id-headers "opc-request-id"
9190
env:
9291
- name: VLLM_LOGGING_LEVEL

‎config/runtimes/vllm/Qwen/qwen-35-tp1-rt.yaml‎

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -59,7 +59,7 @@ spec:
5959
- BM.GPU.H200.8
6060
runner:
6161
name: ome-container
62-
image: fra.ocir.io/idqj093njucb/smg:1.3.3-vllm-v0.18.0
62+
image: fra.ocir.io/idqj093njucb/smg:1.4.0-vllm-v0.18.0
6363
ports:
6464
- containerPort: 8080
6565
name: http1
@@ -78,12 +78,11 @@ spec:
7878
--served-model-name vllm-model \
7979
--limit-mm-per-prompt '{"video": {"count": 1}, "image": {"count": 10}}' \
8080
--generation-config vllm \
81-
--optimization-level 3 \
81+
--max-model-len -1 \
8282
--router-tool-call-parser qwen_coder \
8383
--router-reasoning-parser qwen3_thinking \
8484
--backend vllm \
8585
--router-log-json \
86-
--router-max-concurrent-requests 512 \
8786
--router-request-id-headers "opc-request-id"
8887
env:
8988
- name: VLLM_LOGGING_LEVEL

‎config/runtimes/vllm/Qwen/qwen-35-tp2-fp8-rt.yaml‎

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -61,7 +61,7 @@ spec:
6161
- BM.GPU.H200.8
6262
runner:
6363
name: ome-container
64-
image: fra.ocir.io/idqj093njucb/smg:1.3.3-vllm-v0.18.0
64+
image: fra.ocir.io/idqj093njucb/smg:1.4.0-vllm-v0.18.0
6565
ports:
6666
- containerPort: 8080
6767
name: http1
@@ -80,13 +80,12 @@ spec:
8080
--served-model-name vllm-model \
8181
--limit-mm-per-prompt '{"video": {"count": 1}, "image": {"count": 10}}' \
8282
--generation-config vllm \
83-
--optimization-level 3 \
83+
--max-model-len -1 \
8484
--router-tool-call-parser qwen_coder \
8585
--router-reasoning-parser qwen3_thinking \
8686
--backend vllm \
8787
--router-log-json \
8888
--worker-startup-timeout 1000 \
89-
--router-max-concurrent-requests 512 \
9089
--router-request-id-headers "opc-request-id"
9190
env:
9291
- name: VLLM_LOGGING_LEVEL

‎config/runtimes/vllm/Qwen/qwen-35-tp2-rt.yaml‎

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -59,7 +59,7 @@ spec:
5959
- BM.GPU.H200.8
6060
runner:
6161
name: ome-container
62-
image: fra.ocir.io/idqj093njucb/smg:1.3.3-vllm-v0.18.0
62+
image: fra.ocir.io/idqj093njucb/smg:1.4.0-vllm-v0.18.0
6363
ports:
6464
- containerPort: 8080
6565
name: http1
@@ -78,12 +78,11 @@ spec:
7878
--served-model-name vllm-model \
7979
--limit-mm-per-prompt '{"video": {"count": 1}, "image": {"count": 10}}' \
8080
--generation-config vllm \
81-
--optimization-level 3 \
81+
--max-model-len -1 \
8282
--router-tool-call-parser qwen_coder \
8383
--router-reasoning-parser qwen3_thinking \
8484
--backend vllm \
8585
--router-log-json \
86-
--router-max-concurrent-requests 512 \
8786
--router-request-id-headers "opc-request-id"
8887
env:
8988
- name: VLLM_LOGGING_LEVEL

‎config/runtimes/vllm/Qwen/qwen-35-tp4-rt.yaml‎

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -59,7 +59,7 @@ spec:
5959
- BM.GPU.H200.8
6060
runner:
6161
name: ome-container
62-
image: fra.ocir.io/idqj093njucb/smg:1.3.3-vllm-v0.18.0
62+
image: fra.ocir.io/idqj093njucb/smg:1.4.0-vllm-v0.18.0
6363
ports:
6464
- containerPort: 8080
6565
name: http1
@@ -78,12 +78,11 @@ spec:
7878
--served-model-name vllm-model \
7979
--limit-mm-per-prompt '{"video": {"count": 1}, "image": {"count": 10}}' \
8080
--generation-config vllm \
81-
--optimization-level 3 \
81+
--max-model-len -1 \
8282
--router-tool-call-parser qwen_coder \
8383
--router-reasoning-parser qwen3_thinking \
8484
--backend vllm \
8585
--router-log-json \
86-
--router-max-concurrent-requests 512 \
8786
--router-request-id-headers "opc-request-id"
8887
env:
8988
- name: VLLM_LOGGING_LEVEL

‎config/runtimes/vllm/Qwen/qwen-35-tp8-fp8-rt.yaml‎

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -50,7 +50,7 @@ spec:
5050
- BM.GPU.H200.8
5151
runner:
5252
name: ome-container
53-
image: fra.ocir.io/idqj093njucb/smg:1.3.3-vllm-v0.18.0
53+
image: fra.ocir.io/idqj093njucb/smg:1.4.0-vllm-v0.18.0
5454
ports:
5555
- containerPort: 8080
5656
name: http1
@@ -69,13 +69,12 @@ spec:
6969
--served-model-name vllm-model \
7070
--limit-mm-per-prompt '{"video": {"count": 1}, "image": {"count": 10}}' \
7171
--generation-config vllm \
72-
--optimization-level 3 \
72+
--max-model-len -1 \
7373
--router-tool-call-parser qwen_coder \
7474
--router-reasoning-parser qwen3_thinking \
7575
--backend vllm \
7676
--router-log-json \
7777
--worker-startup-timeout 1800 \
78-
--router-max-concurrent-requests 512 \
7978
--router-request-id-headers "opc-request-id"
8079
env:
8180
- name: VLLM_LOGGING_LEVEL

‎config/runtimes/vllm/Qwen/qwen-35-tp8-rt.yaml‎

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -57,7 +57,7 @@ spec:
5757
- BM.GPU.H200-NC.8
5858
runner:
5959
name: ome-container
60-
image: fra.ocir.io/idqj093njucb/smg:1.3.3-vllm-v0.18.0
60+
image: fra.ocir.io/idqj093njucb/smg:1.4.0-vllm-v0.18.0
6161
ports:
6262
- containerPort: 8080
6363
name: http1
@@ -76,13 +76,12 @@ spec:
7676
--served-model-name vllm-model \
7777
--limit-mm-per-prompt '{"video": {"count": 1}, "image": {"count": 10}}' \
7878
--generation-config vllm \
79-
--optimization-level 3 \
79+
--max-model-len -1 \
8080
--router-tool-call-parser qwen_coder \
8181
--router-reasoning-parser qwen3_thinking \
8282
--backend vllm \
8383
--router-log-json \
8484
--worker-startup-timeout 600 \
85-
--router-max-concurrent-requests 512 \
8685
--router-request-id-headers "opc-request-id"
8786
env:
8887
- name: VLLM_LOGGING_LEVEL

‎config/runtimes/vllm/Qwen/qwen-tp1-fp8-rt.yaml‎

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -182,7 +182,7 @@ spec:
182182
- BM.GPU.H200.8
183183
runner:
184184
name: ome-container
185-
image: fra.ocir.io/idqj093njucb/smg:1.3.3-vllm-v0.18.0
185+
image: fra.ocir.io/idqj093njucb/smg:1.4.0-vllm-v0.18.0
186186
ports:
187187
- containerPort: 8080
188188
name: http1
@@ -201,13 +201,12 @@ spec:
201201
--served-model-name vllm-model \
202202
--limit-mm-per-prompt '{"video": {"count": 1}, "image": {"count": 10}}' \
203203
--generation-config vllm \
204-
--optimization-level 3 \
204+
--max-model-len -1 \
205205
--router-tool-call-parser qwen \
206206
--router-reasoning-parser qwen3 \
207207
--backend vllm \
208208
--router-log-json \
209209
--worker-startup-timeout 1000 \
210-
--router-max-concurrent-requests 512 \
211210
--router-request-id-headers "opc-request-id"
212211
env:
213212
- name: VLLM_LOGGING_LEVEL

‎config/runtimes/vllm/Qwen/qwen-tp1-rt.yaml‎

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -189,7 +189,7 @@ spec:
189189
- BM.GPU.H200.8
190190
runner:
191191
name: ome-container
192-
image: fra.ocir.io/idqj093njucb/smg:1.3.3-vllm-v0.18.0
192+
image: fra.ocir.io/idqj093njucb/smg:1.4.0-vllm-v0.18.0
193193
ports:
194194
- containerPort: 8080
195195
name: http1
@@ -208,12 +208,11 @@ spec:
208208
--served-model-name vllm-model \
209209
--limit-mm-per-prompt '{"video": {"count": 1}, "image": {"count": 10}}' \
210210
--generation-config vllm \
211-
--optimization-level 3 \
211+
--max-model-len -1 \
212212
--router-tool-call-parser qwen \
213213
--router-reasoning-parser qwen3 \
214214
--backend vllm \
215215
--router-log-json \
216-
--router-max-concurrent-requests 512 \
217216
--router-request-id-headers "opc-request-id"
218217
env:
219218
- name: VLLM_LOGGING_LEVEL

‎config/runtimes/vllm/Qwen/qwen-tp2-fp8-rt.yaml‎

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -182,7 +182,7 @@ spec:
182182
- BM.GPU.H200.8
183183
runner:
184184
name: ome-container
185-
image: fra.ocir.io/idqj093njucb/smg:1.3.3-vllm-v0.18.0
185+
image: fra.ocir.io/idqj093njucb/smg:1.4.0-vllm-v0.18.0
186186
ports:
187187
- containerPort: 8080
188188
name: http1
@@ -201,13 +201,12 @@ spec:
201201
--served-model-name vllm-model \
202202
--limit-mm-per-prompt '{"video": {"count": 1}, "image": {"count": 10}}' \
203203
--generation-config vllm \
204-
--optimization-level 3 \
204+
--max-model-len -1 \
205205
--router-tool-call-parser qwen \
206206
--router-reasoning-parser qwen3 \
207207
--backend vllm \
208208
--router-log-json \
209209
--worker-startup-timeout 1000 \
210-
--router-max-concurrent-requests 512 \
211210
--router-request-id-headers "opc-request-id"
212211
env:
213212
- name: VLLM_LOGGING_LEVEL

0 commit comments

Comments
 (0)