forked from flagos-ai/TransformerEngine-FL
-
Notifications
You must be signed in to change notification settings - Fork 2
Expand file tree
/
Copy pathkunlun.yml
More file actions
72 lines (62 loc) · 1.72 KB
/
Copy pathkunlun.yml
File metadata and controls
72 lines (62 loc) · 1.72 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
# KunlunXin Hardware Configuration for TransformerEngine-FL.
# The CI image points at the Harbor tag validated for KunlunXin QA.
hardware_name: kunlunxin
display_name: 'KunlunXin XPU'
checkout_submodules: recursive
container_pull_policy: always
ci_image: harbor.baai.ac.cn/flagos-dev/transformerengine-fl:b98d43b-kunlun-dev
# Runner labels for the self-hosted KunlunXin node.
runner_labels:
- kl-8g-cicd-te
container_volumes: []
container_options: >-
--privileged
--ipc=host
--shm-size=100g
--ulimit memlock=-1
--ulimit stack=67108864
--ulimit nofile=65535:65535
--user root
--group-add video
--device=/dev/xpu0
--device=/dev/xpu1
--device=/dev/xpu2
--device=/dev/xpu3
--device=/dev/xpu4
--device=/dev/xpu5
--device=/dev/xpu6
--device=/dev/xpu7
--device=/dev/xpuctrl
--device=/dev/knem
setup_script: .github/scripts/setup_kunlun.sh
device_types:
- kunlunxin
coverage:
enabled: true
required: false
python: python3
sources:
- transformer_engine
include:
- transformer_engine/pytorch/*
- transformer_engine/debug/*
- transformer_engine/plugin/*
omit:
- '*/setup.py'
- '*/transformer_engine/plugin/core/_build_config.py'
unit_test_matrix:
- name: pytorch_debug
runner: script
path: tests/plugin/backend/kunlun/run_native_tests.sh
- name: pytorch_unittest
runner: script
path: tests/plugin/backend/kunlun/run_native_tests.sh
- name: pytorch_distributed_unittest
runner: script
path: tests/plugin/backend/kunlun/run_native_tests.sh
- name: pytorch_onnx_unittest
runner: script
path: tests/plugin/backend/kunlun/run_native_tests.sh
integration_test_matrix:
- name: pytorch_mcore_integration
path: tests/integration/kunlun/run_mcore.sh