-
Notifications
You must be signed in to change notification settings - Fork 227
/
deploy.yaml
250 lines (249 loc) · 6.1 KB
/
deploy.yaml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
---
version: "2.0"
services:
rayhead:
image: vllm/vllm-openai:v0.6.2@sha256:730ef3d3c17a217b34cfdbfd99be80b3f459e37ef2fc0c5c43ba70752dad08ae
expose:
- port: 8000
as: 8000
to:
- global: true
- port: 8265
as: 8265
to:
- service: rayworker
- global: true
- port: 6380
as: 6380
to:
- service: rayworker
- global: true
- port: 8078
as: 8078
to:
- service: rayworker
- port: 8079
as: 8079
to:
- service: rayworker
- port: 10002
as: 10002
to:
- service: rayworker
- port: 10003
as: 10003
to:
- service: rayworker
- port: 10004
as: 10004
to:
- service: rayworker
- port: 10005
as: 10005
to:
- service: rayworker
- port: 10006
as: 10006
to:
- service: rayworker
- port: 10007
as: 10007
to:
- service: rayworker
- port: 10008
as: 10008
to:
- service: rayworker
- port: 10009
as: 10009
to:
- service: rayworker
- port: 10010
as: 10010
to:
- service: rayworker
command:
- bash
- "-c"
args:
- >-
ray start --head --port=6380
--dashboard-port=8265 --dashboard-host=0.0.0.0
--object-manager-port=8076 --node-manager-port=8077
--dashboard-agent-grpc-port=8078 --dashboard-agent-listen-port=8079
--min-worker-port=10002 --max-worker-port=10010 --block &
echo "Waiting for Ray to start" &&
sleep 15 &&
vllm serve meta-llama/Meta-Llama-3.1-405B-Instruct --tensor-parallel-size 8 --pipeline-parallel-size 3 --gpu-memory-utilization 0.99 --enable-chunked-prefill=False
env:
- HF_TOKEN=hf_ # Hugging Face API token required for Meta-Llama Models
#- NCCL_DEBUG=INFO # Uncomment to enable NCCL debugging
params:
storage:
shm:
mount: /dev/shm
data:
mount: /root/.cache # Mount the data storage to the cache directory for persistent storage of model files
readOnly: false
rayworker:
image: vllm/vllm-openai:v0.6.2@sha256:730ef3d3c17a217b34cfdbfd99be80b3f459e37ef2fc0c5c43ba70752dad08ae
expose:
- port: 8078
as: 8078
to:
- global: true
- port: 8079
as: 8079
to:
- global: true
command:
- bash
- "-c"
args:
- >-
ray start --address $RAYHEAD_PORT_6380_TCP_ADDR:6380 --min-worker-port=10002 --max-worker-port=10010 --block
env:
- HF_TOKEN=hf_ # Hugging Face API token required for Meta-Llama Models
# - NCCL_DEBUG=INFO
params:
storage:
shm:
mount: /dev/shm
data:
mount: /root/.cache # Mount the data storage to the cache directory for persistent storage of model files
readOnly: false
rayworker2:
image: vllm/vllm-openai:v0.6.2@sha256:730ef3d3c17a217b34cfdbfd99be80b3f459e37ef2fc0c5c43ba70752dad08ae
expose:
- port: 8078
as: 8078
to:
- global: true
- port: 8079
as: 8079
to:
- global: true
command:
- bash
- "-c"
args:
- >-
ray start --address $RAYHEAD_PORT_6380_TCP_ADDR:6380 --min-worker-port=10002 --max-worker-port=10010 --block
env:
- HF_TOKEN=hf_ # Hugging Face API token required for Meta-Llama Models
#- NCCL_DEBUG=INFO
params:
storage:
shm:
mount: /dev/shm
data:
mount: /root/.cache # Mount the data storage to the cache directory for persistent storage of model files
readOnly: false
profiles:
compute:
rayhead:
resources:
cpu:
units: 128
memory:
size: 128Gi
storage:
- size: 100Gi
- name: data
size: 1000Gi
attributes:
persistent: true
class: beta3
- name: shm
size: 10Gi
attributes:
class: ram
persistent: false
gpu:
units: 8
attributes:
vendor:
nvidia:
- model: h100
ram: 80Gi
- model: a100
ram: 80Gi
rayworker:
resources:
cpu:
units: 128
memory:
size: 128Gi
storage:
- size: 100Gi
- name: data
size: 1000Gi
attributes:
persistent: true
class: beta3
- name: shm
size: 10Gi
attributes:
class: ram
persistent: false
gpu:
units: 8
attributes:
vendor:
nvidia:
- model: h100
ram: 80Gi
- model: a100
ram: 80Gi
rayworker2:
resources:
cpu:
units: 128
memory:
size: 128Gi
storage:
- size: 100Gi
- name: data
size: 1000Gi
attributes:
persistent: true
class: beta3
- name: shm
size: 10Gi
attributes:
class: ram
persistent: false
gpu:
units: 8
attributes:
vendor:
nvidia:
- model: h100
ram: 80Gi
- model: a100
ram: 80Gi
placement:
dcloud:
pricing:
rayhead:
denom: uakt
amount: 1000000
rayworker:
denom: uakt
amount: 10000
rayworker2:
denom: uakt
amount: 10000
deployment:
rayhead:
dcloud:
profile: rayhead
count: 1
rayworker:
dcloud:
profile: rayworker
count: 1
rayworker2:
dcloud:
profile: rayworker2
count: 1