main
yaml 319 lines 13.7 KB
Raw
1 # Native Buildkite pipeline — disaggregated PD SLURM tests (MoRIIO).
2 #
3 # ROUTER_TYPE must be set explicitly on every step: run-slurm-disagg-test.sh
4 # defaults to `vllm-router`
5 #
6
7 steps:
8
9 #-------------------------------------- mi350 · disagg 1P1D (TP8, 2 nodes) · SLURM --------------------------------------#
10
11 - label: "DeepSeek-V3-PD-1P1D-TP8-MoRIIO-proxy"
12 agents:
13 queue: amd_mi350_ainic
14 timeout_in_minutes: 180
15 concurrency: 2
16 concurrency_group: "amd-disagg/mi350/2node"
17 commands:
18 - MODEL_NAME=DeepSeek-V3 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
19
20 - label: "DeepSeek-V3-PD-1P1D-TP8-MoRIIO-vllm-router"
21 agents:
22 queue: amd_mi350_ainic
23 timeout_in_minutes: 180
24 concurrency: 2
25 concurrency_group: "amd-disagg/mi350/2node"
26 commands:
27 - MODEL_NAME=DeepSeek-V3 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
28
29 - label: "DeepSeek-R1-MXFP4-PD-1P1D-TP8-MoRIIO-proxy"
30 agents:
31 queue: amd_mi350_ainic
32 timeout_in_minutes: 180
33 concurrency: 2
34 concurrency_group: "amd-disagg/mi350/2node"
35 commands:
36 - MODEL_NAME=DeepSeek-R1-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
37
38 - label: "DeepSeek-R1-MXFP4-PD-1P1D-TP8-MoRIIO-vllm-router"
39 agents:
40 queue: amd_mi350_ainic
41 timeout_in_minutes: 180
42 concurrency: 2
43 concurrency_group: "amd-disagg/mi350/2node"
44 commands:
45 - MODEL_NAME=DeepSeek-R1-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
46
47 - label: "Kimi-K2.5-MXFP4-PD-1P1D-TP8-MoRIIO-proxy"
48 agents:
49 queue: amd_mi350_ainic
50 timeout_in_minutes: 180
51 concurrency: 2
52 concurrency_group: "amd-disagg/mi350/2node"
53 commands:
54 - MODEL_NAME=Kimi-K2.5-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
55
56 - label: "Kimi-K2.5-MXFP4-PD-1P1D-TP8-MoRIIO-vllm-router"
57 agents:
58 queue: amd_mi350_ainic
59 timeout_in_minutes: 180
60 concurrency: 2
61 concurrency_group: "amd-disagg/mi350/2node"
62 commands:
63 - MODEL_NAME=Kimi-K2.5-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
64
65 - label: "Kimi-K2.6-MXFP4-PD-1P1D-TP8-MoRIIO-proxy"
66 agents:
67 queue: amd_mi350_ainic
68 timeout_in_minutes: 180
69 concurrency: 2
70 concurrency_group: "amd-disagg/mi350/2node"
71 commands:
72 - MODEL_NAME=Kimi-K2.6-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
73
74 - label: "Kimi-K2.6-MXFP4-PD-1P1D-TP8-MoRIIO-vllm-router"
75 agents:
76 queue: amd_mi350_ainic
77 timeout_in_minutes: 180
78 concurrency: 2
79 concurrency_group: "amd-disagg/mi350/2node"
80 commands:
81 - MODEL_NAME=Kimi-K2.6-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
82
83 - label: "MiniMax-M3-MXFP8-PD-1P1D-TP8-MoRIIO-proxy"
84 agents:
85 queue: amd_mi350_ainic
86 timeout_in_minutes: 180
87 concurrency: 2
88 concurrency_group: "amd-disagg/mi350/2node"
89 commands:
90 - MODEL_NAME=MiniMax-M3-MXFP8 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
91
92 - label: "MiniMax-M3-MXFP8-PD-1P1D-TP8-MoRIIO-vllm-router"
93 agents:
94 queue: amd_mi350_ainic
95 timeout_in_minutes: 180
96 concurrency: 2
97 concurrency_group: "amd-disagg/mi350/2node"
98 commands:
99 - MODEL_NAME=MiniMax-M3-MXFP8 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
100
101 - label: "GLM-5.2-FP8-PD-1P1D-TP8-MoRIIO-proxy"
102 agents:
103 queue: amd_mi350_ainic
104 timeout_in_minutes: 180
105 concurrency: 2
106 concurrency_group: "amd-disagg/mi350/2node"
107 commands:
108 - MODEL_NAME=GLM-5.2-FP8 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
109
110 - label: "GLM-5.2-FP8-PD-1P1D-TP8-MoRIIO-vllm-router"
111 agents:
112 queue: amd_mi350_ainic
113 timeout_in_minutes: 180
114 concurrency: 2
115 concurrency_group: "amd-disagg/mi350/2node"
116 commands:
117 - MODEL_NAME=GLM-5.2-FP8 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
118
119 #-------------------------------------- mi350 · disagg 2P2D (TP8, 4 nodes) · SLURM --------------------------------------#
120
121 - label: "DeepSeek-V3-PD-2P2D-TP8-MoRIIO-proxy"
122 agents:
123 queue: amd_mi350_ainic
124 timeout_in_minutes: 180
125 concurrency: 2
126 concurrency_group: "amd-disagg/mi350/4node"
127 commands:
128 - MODEL_NAME=DeepSeek-V3 xP=2 yD=2 NODES=4 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
129
130 - label: "DeepSeek-V3-PD-2P2D-TP8-MoRIIO-vllm-router"
131 agents:
132 queue: amd_mi350_ainic
133 timeout_in_minutes: 180
134 concurrency: 2
135 concurrency_group: "amd-disagg/mi350/4node"
136 commands:
137 - MODEL_NAME=DeepSeek-V3 xP=2 yD=2 NODES=4 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
138
139 - label: "DeepSeek-R1-MXFP4-PD-2P2D-TP8-MoRIIO-proxy"
140 agents:
141 queue: amd_mi350_ainic
142 timeout_in_minutes: 180
143 concurrency: 2
144 concurrency_group: "amd-disagg/mi350/4node"
145 commands:
146 - MODEL_NAME=DeepSeek-R1-MXFP4 xP=2 yD=2 NODES=4 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
147
148 - label: "DeepSeek-R1-MXFP4-PD-2P2D-TP8-MoRIIO-vllm-router"
149 agents:
150 queue: amd_mi350_ainic
151 timeout_in_minutes: 180
152 concurrency: 2
153 concurrency_group: "amd-disagg/mi350/4node"
154 commands:
155 - MODEL_NAME=DeepSeek-R1-MXFP4 xP=2 yD=2 NODES=4 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
156
157 - label: "Kimi-K2.5-MXFP4-PD-2P2D-TP8-MoRIIO-proxy"
158 agents:
159 queue: amd_mi350_ainic
160 timeout_in_minutes: 180
161 concurrency: 2
162 concurrency_group: "amd-disagg/mi350/4node"
163 commands:
164 - MODEL_NAME=Kimi-K2.5-MXFP4 xP=2 yD=2 NODES=4 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
165
166 - label: "Kimi-K2.5-MXFP4-PD-2P2D-TP8-MoRIIO-vllm-router"
167 agents:
168 queue: amd_mi350_ainic
169 timeout_in_minutes: 180
170 concurrency: 2
171 concurrency_group: "amd-disagg/mi350/4node"
172 commands:
173 - MODEL_NAME=Kimi-K2.5-MXFP4 xP=2 yD=2 NODES=4 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
174
175 - label: "Kimi-K2.6-MXFP4-PD-2P2D-TP8-MoRIIO-proxy"
176 agents:
177 queue: amd_mi350_ainic
178 timeout_in_minutes: 180
179 concurrency: 2
180 concurrency_group: "amd-disagg/mi350/4node"
181 commands:
182 - MODEL_NAME=Kimi-K2.6-MXFP4 xP=2 yD=2 NODES=4 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
183
184 - label: "Kimi-K2.6-MXFP4-PD-2P2D-TP8-MoRIIO-vllm-router"
185 agents:
186 queue: amd_mi350_ainic
187 timeout_in_minutes: 180
188 concurrency: 2
189 concurrency_group: "amd-disagg/mi350/4node"
190 commands:
191 - MODEL_NAME=Kimi-K2.6-MXFP4 xP=2 yD=2 NODES=4 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
192
193 - label: "MiniMax-M3-MXFP8-PD-2P2D-TP8-MoRIIO-proxy"
194 agents:
195 queue: amd_mi350_ainic
196 timeout_in_minutes: 180
197 concurrency: 2
198 concurrency_group: "amd-disagg/mi350/4node"
199 commands:
200 - MODEL_NAME=MiniMax-M3-MXFP8 xP=2 yD=2 NODES=4 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
201
202 - label: "MiniMax-M3-MXFP8-PD-2P2D-TP8-MoRIIO-vllm-router"
203 agents:
204 queue: amd_mi350_ainic
205 timeout_in_minutes: 180
206 concurrency: 2
207 concurrency_group: "amd-disagg/mi350/4node"
208 commands:
209 - MODEL_NAME=MiniMax-M3-MXFP8 xP=2 yD=2 NODES=4 GPUS_PER_NODE=8 WIDE_EP_MODE=0 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
210
211 #-------------------------------------- mi350 · disagg wide-EP intranode (EP8/DP8, 2 nodes) · SLURM --------------------------------------#
212
213 - label: "DeepSeek-V3-PD-1P1D-EP8/DP8-WideEP-MoRIIO-proxy"
214 agents:
215 queue: amd_mi350_ainic
216 timeout_in_minutes: 180
217 concurrency: 2
218 concurrency_group: "amd-disagg/mi350/2node"
219 commands:
220 - MODEL_NAME=DeepSeek-V3 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
221
222 - label: "DeepSeek-V3-PD-1P1D-EP8/DP8-WideEP-MoRIIO-vllm-router"
223 agents:
224 queue: amd_mi350_ainic
225 timeout_in_minutes: 180
226 concurrency: 2
227 concurrency_group: "amd-disagg/mi350/2node"
228 commands:
229 - MODEL_NAME=DeepSeek-V3 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
230
231 - label: "DeepSeek-R1-MXFP4-PD-1P1D-EP8/DP8-WideEP-MoRIIO-proxy"
232 agents:
233 queue: amd_mi350_ainic
234 timeout_in_minutes: 180
235 concurrency: 2
236 concurrency_group: "amd-disagg/mi350/2node"
237 commands:
238 - MODEL_NAME=DeepSeek-R1-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
239
240 - label: "DeepSeek-R1-MXFP4-PD-1P1D-EP8/DP8-WideEP-MoRIIO-vllm-router"
241 agents:
242 queue: amd_mi350_ainic
243 timeout_in_minutes: 180
244 concurrency: 2
245 concurrency_group: "amd-disagg/mi350/2node"
246 commands:
247 - MODEL_NAME=DeepSeek-R1-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
248
249 - label: "Kimi-K2.5-MXFP4-PD-1P1D-EP8/DP8-WideEP-MoRIIO-proxy"
250 agents:
251 queue: amd_mi350_ainic
252 timeout_in_minutes: 180
253 concurrency: 2
254 concurrency_group: "amd-disagg/mi350/2node"
255 commands:
256 - MODEL_NAME=Kimi-K2.5-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
257
258 - label: "Kimi-K2.5-MXFP4-PD-1P1D-EP8/DP8-WideEP-MoRIIO-vllm-router"
259 agents:
260 queue: amd_mi350_ainic
261 timeout_in_minutes: 180
262 concurrency: 2
263 concurrency_group: "amd-disagg/mi350/2node"
264 commands:
265 - MODEL_NAME=Kimi-K2.5-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
266
267 - label: "Kimi-K2.6-MXFP4-PD-1P1D-EP8/DP8-WideEP-MoRIIO-proxy"
268 agents:
269 queue: amd_mi350_ainic
270 timeout_in_minutes: 180
271 concurrency: 2
272 concurrency_group: "amd-disagg/mi350/2node"
273 commands:
274 - MODEL_NAME=Kimi-K2.6-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
275
276 - label: "Kimi-K2.6-MXFP4-PD-1P1D-EP8/DP8-WideEP-MoRIIO-vllm-router"
277 agents:
278 queue: amd_mi350_ainic
279 timeout_in_minutes: 180
280 concurrency: 2
281 concurrency_group: "amd-disagg/mi350/2node"
282 commands:
283 - MODEL_NAME=Kimi-K2.6-MXFP4 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
284
285 - label: "MiniMax-M3-MXFP8-PD-1P1D-EP8/DP8-WideEP-MoRIIO-proxy"
286 agents:
287 queue: amd_mi350_ainic
288 timeout_in_minutes: 180
289 concurrency: 2
290 concurrency_group: "amd-disagg/mi350/2node"
291 commands:
292 - MODEL_NAME=MiniMax-M3-MXFP8 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
293
294 - label: "MiniMax-M3-MXFP8-PD-1P1D-EP8/DP8-WideEP-MoRIIO-vllm-router"
295 agents:
296 queue: amd_mi350_ainic
297 timeout_in_minutes: 180
298 concurrency: 2
299 concurrency_group: "amd-disagg/mi350/2node"
300 commands:
301 - MODEL_NAME=MiniMax-M3-MXFP8 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
302
303 - label: "GLM-5.2-FP8-PD-1P1D-EP8/DP8-WideEP-MoRIIO-proxy"
304 agents:
305 queue: amd_mi350_ainic
306 timeout_in_minutes: 180
307 concurrency: 2
308 concurrency_group: "amd-disagg/mi350/2node"
309 commands:
310 - MODEL_NAME=GLM-5.2-FP8 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=proxy WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh
311
312 - label: "GLM-5.2-FP8-PD-1P1D-EP8/DP8-WideEP-MoRIIO-vllm-router"
313 agents:
314 queue: amd_mi350_ainic
315 timeout_in_minutes: 180
316 concurrency: 2
317 concurrency_group: "amd-disagg/mi350/2node"
318 commands:
319 - MODEL_NAME=GLM-5.2-FP8 NODES=2 GPUS_PER_NODE=8 WIDE_EP_MODE=1 MORIIO_READ_MODE=0 RUN_AFTER_HEALTH=accuracy ROUTER_TYPE=vllm-router WAIT=1 bash .buildkite/amd-disagg/run-slurm-disagg-test.sh