28 lines
732 B
TOML
28 lines
732 B
TOML
# This file is an append-only queue source for the monitor.
|
|
# Each job name must stay unique and immutable once appended.
|
|
version = 1
|
|
|
|
[[jobs]]
|
|
name = "smoke-train-h20-1gpu"
|
|
gpus = 1
|
|
gpu_model = "H20"
|
|
hosts = ["dash0", "dash1", "dash2"]
|
|
command = "python train.py --config configs/smoke.toml"
|
|
artifacts = ["outputs/smoke-train-h20-1gpu"]
|
|
env = { WANDB_MODE = "offline" }
|
|
|
|
[[jobs]]
|
|
name = "eval-5090-4gpu"
|
|
gpus = 4
|
|
gpu_model = "5090"
|
|
hosts = ["dash5"]
|
|
command = "python eval.py --config configs/eval.toml"
|
|
artifacts = ["outputs/eval-5090-4gpu", "logs/eval-5090-4gpu.log"]
|
|
|
|
[[jobs]]
|
|
name = "special-dash3-run"
|
|
gpus = 2
|
|
hosts = ["dash3"]
|
|
command = "python benchmark.py --suite long-context"
|
|
artifacts = ["outputs/special-dash3-run"]
|