Skip to content

Commit 4e1acae

Browse files
authored
fix: Use a new cluster for DAG maxtext_muti_tier_checkpointing (GoogleCloudPlatform#1123)
This change fixes DAG `maxtext_muti_tier_checkpointing` by pointing it to a new cluster `auto-v5p-8-bodaborg` due to overlapping DAG's pointing same cluster.
1 parent b02c38e commit 4e1acae

2 files changed

Lines changed: 9 additions & 2 deletions

File tree

dags/common/vm_resource.py

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -255,6 +255,13 @@ class XpkClusters:
255255
project=Project.CLOUD_TPU_MULTIPOD_DEV.value,
256256
zone=Zone.EUROPE_WEST4_B.value,
257257
)
258+
TPU_V5P_8_CLUSTER_V2 = XpkClusterConfig(
259+
name="auto-v5p-8-bodaborg",
260+
device_version=TpuVersion.V5P,
261+
core_count=8,
262+
project=Project.CLOUD_TPU_MULTIPOD_DEV.value,
263+
zone=Zone.EUROPE_WEST4_B.value,
264+
)
258265
TPU_V5P_128_CLUSTER = XpkClusterConfig(
259266
name="v5p-128-bodaborg-europe-west4-b",
260267
device_version=TpuVersion.V5P,

dags/multipod/maxtext_multi_tier_checkpointing.py

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -53,7 +53,7 @@
5353
}
5454
clusters = {
5555
# accelerator: cluster name
56-
"v5p-8": XpkClusters.TPU_V5P_8_CLUSTER,
56+
"v5p-8": XpkClusters.TPU_V5P_8_CLUSTER_V2,
5757
}
5858

5959
for mode, image in docker_images:
@@ -72,4 +72,4 @@
7272
run_model_cmds=command,
7373
docker_image=image.value,
7474
test_owner=test_owner.ABHINAV_S,
75-
).run(ramdisk_directory="local")
75+
).run(ramdisk_directory="local", mtc_enabled=True)

0 commit comments

Comments
 (0)