-
Notifications
You must be signed in to change notification settings - Fork 12
Expand file tree
/
Copy pathmatrix.yaml
More file actions
225 lines (213 loc) · 5.7 KB
/
matrix.yaml
File metadata and controls
225 lines (213 loc) · 5.7 KB
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
# Runner Packing Rules
#
# - backend: (Required) The accelerated backend to pack for,
# used to select the packing rules and Dockerfile below `pack` directory.
# - services: (Required) The inference service to pack for,
# used to select the Docker build phase described in `pack/${backend}/Dockerfile`.
# - platforms: (Optional) The platforms to build for,
# used to select the Docker Linux build platforms.
# Default to `linux/amd64` and `linux/arm64`.
# - args: (Optional) The build arguments to pass to the Docker build,
# used to override the default build arguments in `pack/${backend}/Dockerfile`.
rules:
#
# Ascend CANN
#
## Ascend CANN 8.5.1, using CANN Kernel for A3.
##
- backend: "cann"
services:
- "vllm"
args:
- "CANN_VERSION=8.5.1"
- "CANN_ARCHS=a3"
- "VLLM_VERSION=0.18.0"
- "CANN_BASE_IMAGE=quay.io/ascend/vllm-ascend:v0.18.0-a3"
## Ascend CANN 8.5.1, using CANN Kernel for 910B.
##
- backend: "cann"
services:
- "vllm"
args:
- "CANN_VERSION=8.5.1"
- "CANN_ARCHS=910b"
- "VLLM_VERSION=0.18.0"
- "CANN_BASE_IMAGE=quay.io/ascend/vllm-ascend:v0.18.0"
## Ascend CANN 8.5.1, using CANN Kernel for 310P.
##
- backend: "cann"
services:
- "vllm"
args:
- "CANN_VERSION=8.5.1"
- "CANN_ARCHS=310p"
- "VLLM_VERSION=0.18.0"
- "CANN_BASE_IMAGE=quay.io/ascend/vllm-ascend:v0.18.0-310p"
## Ascend CANN 8.5.1, SGLang 0.5.12.post1 for A3.
## (no sglang upstream image for 310p variant)
##
- backend: "cann"
services:
- "sglang"
args:
- "CANN_VERSION=8.5.1"
- "CANN_ARCHS=a3"
- "SGLANG_VERSION=0.5.12.post1"
- "SGLANG_BASE_IMAGE=quay.io/ascend/sglang:v0.5.12.post1-cann8.5.0-a3"
## Ascend CANN 8.5.1, SGLang 0.5.12.post1 for 910B.
##
- backend: "cann"
services:
- "sglang"
args:
- "CANN_VERSION=8.5.1"
- "CANN_ARCHS=910b"
- "SGLANG_VERSION=0.5.12.post1"
- "SGLANG_BASE_IMAGE=quay.io/ascend/sglang:v0.5.12.post1-cann8.5.0-910b"
#
# AMD ROCm
#
## AMD ROCm 7.2.1
##
- backend: "rocm"
services:
- "vllm"
platforms:
- "linux/amd64"
args:
- "ROCM_VERSION=7.2.1"
- "VLLM_VERSION=0.21.0"
- "VLLM_BASE_IMAGE=vllm/vllm-openai-rocm:v0.21.0"
## AMD ROCm 7.2.1 - SGLang 0.5.12.post1 (rocm720-mi30x)
##
- backend: "rocm"
services:
- "sglang"
platforms:
- "linux/amd64"
args:
- "ROCM_VERSION=7.2.1"
- "SGLANG_VERSION=0.5.12.post1"
- "SGLANG_BASE_IMAGE=lmsysorg/sglang:v0.5.12.post1-rocm720-mi30x"
#
# NVIDIA CUDA
#
## NVIDIA CUDA 13.0.1
##
- backend: "cuda"
services:
- "vllm"
args:
- "CUDA_VERSION=13.0.1"
- "VLLM_VERSION=0.21.0"
- "VLLM_BASE_IMAGE=vllm/vllm-openai:v0.21.0-ubuntu2404"
## NVIDIA CUDA 13.0.1 - SGLang 0.5.12.post1
##
- backend: "cuda"
services:
- "sglang"
args:
- "CUDA_VERSION=13.0.1"
- "SGLANG_VERSION=0.5.12.post1"
- "SGLANG_BASE_IMAGE=lmsysorg/sglang:v0.5.12.post1-cu130"
## NVIDIA CUDA 12.9.1
##
- backend: "cuda"
services:
- "vllm"
args:
- "CUDA_VERSION=12.9.1"
- "VLLM_VERSION=0.21.0"
- "VLLM_BASE_IMAGE=vllm/vllm-openai:v0.21.0-cu129-ubuntu2404"
## NVIDIA CUDA 12.9.1 - SGLang 0.5.12.post1
##
- backend: "cuda"
services:
- "sglang"
args:
- "CUDA_VERSION=12.9.1"
- "SGLANG_VERSION=0.5.12.post1"
- "SGLANG_BASE_IMAGE=lmsysorg/sglang:v0.5.12.post1-cu129"
- "SGLANG_NVIDIA_HPCX_VERSION=2.24.1_cuda12"
#
# Iluvatar CoreX
#
##
## Use pack flow's arguments to override the default COREX version and version extra,
## which are used in `pack/corex/Dockerfile` to construct the image tags and build arguments.
##
- backend: "corex"
services:
- "vllm"
platforms:
- "linux/amd64"
args:
- "COREX_VERSION="
- "VLLM_VERSION="
#
# Hygon DTK
#
##
## Use pack flow's arguments to override the default DTK version and version extra,
## which are used in `pack/dtk/Dockerfile` to construct the image tags and build arguments.
##
- backend: "dtk"
services:
- "vllm"
platforms:
- "linux/amd64"
args:
- "DTK_VERSION="
- "VLLM_VERSION="
#
# T-Head HGGC
#
##
## Use pack flow's arguments to override the default HGGC version and version extra,
## which are used in `pack/hggc/Dockerfile` to construct the image tags and build arguments.
##
## Get HGGC version via `VERSION=$(cat /usr/local/PPU_SDK/include/hggc.h | grep HGGC_VERSION | grep define | cut -d' ' -f3); echo "$(( ${VERSION} / 1000 )).$(( ${VERSION} % 1000 / 10 )).$(( ${VERSION} % 10 ))"`
- backend: "hggc"
services:
- "vllm"
- "sglang"
platforms:
- "linux/amd64"
args:
- "HGGC_VERSION="
- "HGGC_VERSION_EXTRA="
- "VLLM_VERSION="
#
# MateX MACA
#
##
## Use pack flow's arguments to override the default MACA version and version extra,
## which are used in `pack/maca/Dockerfile` to construct the image tags and build arguments.
##
- backend: "maca"
services:
- "vllm"
- "sglang"
platforms:
- "linux/amd64"
args:
- "MACA_VERSION="
- "MACA_VERSION_EXTRA="
- "VLLM_VERSION="
#
# MThreads MUSA
#
##
## Use pack flow's arguments to override the default MUSA version and version extra,
## which are used in `pack/musa/Dockerfile` to construct the image tags and build arguments.
##
- backend: "musa"
services:
- "vllm"
- "sglang"
platforms:
- "linux/amd64"
args:
- "MUSA_VERSION="
- "MUSA_VERSION_EXTRA="
- "VLLM_VERSION="