A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 10, 20:15 UTC
Add docker buildx bake configuration (#31477)
Signed-off-by: Amr Mahdi <amrmahdi@meta.com>
Amr Mahdi committed
Dec 30, 2025 at 17:08 UTC
e1ee11b2a50b00b8bbcef19d38a13dde115a2260
1 file changed
+76
docker/docker-bake.hcl
new
+76
@@ -0,0 +1,76 @@
1
+# docker-bake.hcl - vLLM Docker build configuration
2
+#
3
+# This file lives in vLLM repo at docker/docker-bake.hcl
4
+#
5
+# Usage:
6
+# cd docker && docker buildx bake # Build default target (openai)
7
+# cd docker && docker buildx bake test # Build test target
8
+# docker buildx bake --print # Show resolved config
9
+#
10
+# Reference: https://docs.docker.com/build/bake/reference/
11
+
12
+# Build configuration
13
+
14
+variable "MAX_JOBS" {
15
+ default = 16
16
+}
17
+
18
+variable "NVCC_THREADS" {
19
+ default = 8
20
+}
21
+
22
+variable "TORCH_CUDA_ARCH_LIST" {
23
+ default = "8.0 8.9 9.0 10.0"
24
+}
25
+
26
+variable "COMMIT" {
27
+ default = ""
28
+}
29
+
30
+# Groups
31
+
32
+group "default" {
33
+ targets = ["openai"]
34
+}
35
+
36
+# Base targets
37
+
38
+target "_common" {
39
+ dockerfile = "docker/Dockerfile"
40
+ context = "."
41
+ args = {
42
+ max_jobs = MAX_JOBS
43
+ nvcc_threads = NVCC_THREADS
44
+ torch_cuda_arch_list = TORCH_CUDA_ARCH_LIST
45
+ }
46
+}
47
+
48
+target "_labels" {
49
+ labels = {
50
+ "org.opencontainers.image.source" = "https://github.com/vllm-project/vllm"
51
+ "org.opencontainers.image.vendor" = "vLLM"
52
+ "org.opencontainers.image.title" = "vLLM"
53
+ "org.opencontainers.image.description" = "vLLM: A high-throughput and memory-efficient inference and serving engine for LLMs"
54
+ "org.opencontainers.image.licenses" = "Apache-2.0"
55
+ "org.opencontainers.image.revision" = COMMIT
56
+ }
57
+ annotations = [
58
+ "index,manifest:org.opencontainers.image.revision=${COMMIT}",
59
+ ]
60
+}
61
+
62
+# Build targets
63
+
64
+target "test" {
65
+ inherits = ["_common", "_labels"]
66
+ target = "test"
67
+ tags = ["vllm:test"]
68
+ output = ["type=docker"]
69
+}
70
+
71
+target "openai" {
72
+ inherits = ["_common", "_labels"]
73
+ target = "vllm-openai"
74
+ tags = ["vllm:openai"]
75
+ output = ["type=docker"]
76
+}