Add blind Qwen235B fleet job

This commit is contained in:
2026-07-15 23:17:43 +08:00
parent 33c91109d7
commit 91869c531c
2 changed files with 35 additions and 0 deletions

View File

@@ -0,0 +1,25 @@
version = 1
[paths]
state_dir = ".aituner/gpu_fleet/fixed_cohort_state"
artifacts_dir = ".aituner/gpu_fleet/fixed_cohort_artifacts"
[ssh]
connect_timeout_sec = 10
[scheduler]
gpu_free_memory_mb = 1024
gpu_free_utilization_pct = 10
prefer_pack = true
[sync]
mode = "rsync"
local_path = "runs/frontier-multicase-sufficiency-v0/best_effort"
exclude = ["__pycache__/", "*.pyc"]
[[hosts]]
name = "dash0"
ssh_alias = "dash0"
enabled = true
sync_remote_path = "/home/admin/cpfs/wjh/aituner/aituner-frontier-best-effort-20260715/runs/frontier-multicase-sufficiency-v0/best_effort"
fleet_root = "/home/admin/.aituner_gpu_fleet_fixed_cohort"

View File

@@ -0,0 +1,10 @@
# Immutable one-job queue for the blind Qwen235B fixed-cohort rank experiment.
version = 1
[[jobs]]
name = "qwen235b-fixed-cohort-community-v1-blind-20260715"
gpus = 8
gpu_model = "H20"
hosts = ["dash0"]
command = "cd ../../.. && /tmp/wjh-frontier-vllm0102-smoke/.venv/bin/python runs/frontier-multicase-sufficiency-v0/best_effort/fixed_cohort_rank.py run-community --manifest /home/admin/cpfs/wjh/frontier-community-qwen235-smoke-20260715/rankable-fixed-cohort-v1/community-v1-blind/run_manifest.json"
env = { PYTHONPATH = "/home/admin/cpfs/wjh/aituner/aituner-frontier-best-effort-20260715/src" }