gahow/aituner
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

experiment: reuse predictors across load contracts

Browse Source
This commit is contained in:
Gahow Wang
2026-07-20 18:51:31 +08:00
parent 809ad9ffef
commit e651ecc923
4 changed files with 22 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
runs/frontier-fidelity-envelope-v1/run_frontier_qwen30_exact_trace_surface.py
View File

@@ -67,6 +67,11 @@ def parse_args() -> argparse.Namespace:
parser.add_argument("--kernel-profile-root", type=Path)
parser.add_argument("--python-deps", type=Path, required=True)
parser.add_argument("--output-root", type=Path, required=True)
parser.add_argument(
"--predictor-cache-root",
type=Path,
help="Reuse a validated predictor cache outside the output root.",
)
parser.add_argument(
"--trace",
action="append",
@@ -558,9 +563,10 @@ def main() -> None:
)
config_results = []
predictor_cache_root = args.predictor_cache_root or args.output_root / "cache"
for config in selected:
loads = []
config_knobs = BASE.knobs(config, paths, args.output_root / "cache")
config_knobs = BASE.knobs(config, paths, predictor_cache_root)
config_knobs["enable_prefix_caching"] = args.prefix_caching
config_knobs["prediction_max_tokens_per_request"] = 40960
config_knobs["decode_cuda_graph_mode"] = args.decode_cuda_graph_mode
@@ -614,7 +620,7 @@ def main() -> None:
command.extend(
[
"--metrics_config_cache_dir",
str(args.output_root / "cache"),
str(predictor_cache_root),
]
)
if args.align_real_graph_runtime:
@@ -829,7 +835,7 @@ def main() -> None:
"arrival": "original_trace_timestamp_and_order",
"input_output": "exact_source_values",
"store_stage_batch_ledger": args.store_stage_batch_ledger,
"predictor_cache_root": str(args.output_root / "cache"),
"predictor_cache_root": str(predictor_cache_root),
"ttft_slo": "1000ms + 1000ms * input_tokens / 8000",
"tpot_slos_ms": TPOT_SLOS_MS,
"primary_tpot_slo_ms": 150.0,

8
runs/frontier-workload-regime-taxonomy-v0/jobs_simulator_per_gpu_v1.toml
View File

@@ -5,7 +5,7 @@ name = "workload-regime-per-gpu-sim-tp1-v1-20260720"
gpus = 1
gpu_model = "H20"
hosts = ["dash1"]
command = "TRACES_ROOT=/home/admin/cpfs/wjh/aituner/workload-regime-20260720/traces-per-gpu-low SIM_GENERATION=per-gpu-v1 bash runs/frontier-workload-regime-taxonomy-v0/run_simulator_group.sh 1 both"
command = "TRACES_ROOT=/home/admin/cpfs/wjh/aituner/workload-regime-20260720/traces-per-gpu-low SIM_GENERATION=per-gpu-v1 PREDICTOR_CACHE_GENERATION=v4 bash runs/frontier-workload-regime-taxonomy-v0/run_simulator_group.sh 1 both"
artifacts = []
[[jobs]]
@@ -13,7 +13,7 @@ name = "workload-regime-per-gpu-sim-tp2-v1-20260720"
gpus = 1
gpu_model = "H20"
hosts = ["dash2"]
command = "TRACES_ROOT=/home/admin/cpfs/wjh/aituner/workload-regime-20260720/traces-per-gpu-low SIM_GENERATION=per-gpu-v1 bash runs/frontier-workload-regime-taxonomy-v0/run_simulator_group.sh 2 both"
command = "TRACES_ROOT=/home/admin/cpfs/wjh/aituner/workload-regime-20260720/traces-per-gpu-low SIM_GENERATION=per-gpu-v1 PREDICTOR_CACHE_GENERATION=v4 bash runs/frontier-workload-regime-taxonomy-v0/run_simulator_group.sh 2 both"
artifacts = []
[[jobs]]
@@ -21,7 +21,7 @@ name = "workload-regime-per-gpu-sim-tp4-noprefix-v1-20260720"
gpus = 1
gpu_model = "H20"
hosts = ["dash3"]
command = "TRACES_ROOT=/home/admin/cpfs/wjh/aituner/workload-regime-20260720/traces-per-gpu-low SIM_GENERATION=per-gpu-v1 bash runs/frontier-workload-regime-taxonomy-v0/run_simulator_group.sh 4 false"
command = "TRACES_ROOT=/home/admin/cpfs/wjh/aituner/workload-regime-20260720/traces-per-gpu-low SIM_GENERATION=per-gpu-v1 PREDICTOR_CACHE_GENERATION=v4 bash runs/frontier-workload-regime-taxonomy-v0/run_simulator_group.sh 4 false"
artifacts = []
[[jobs]]
@@ -29,5 +29,5 @@ name = "workload-regime-per-gpu-sim-tp4-prefix-v1-20260720"
gpus = 1
gpu_model = "H20"
hosts = ["dash4"]
command = "TRACES_ROOT=/home/admin/cpfs/wjh/aituner/workload-regime-20260720/traces-per-gpu-low SIM_GENERATION=per-gpu-v1 bash runs/frontier-workload-regime-taxonomy-v0/run_simulator_group.sh 4 true"
command = "TRACES_ROOT=/home/admin/cpfs/wjh/aituner/workload-regime-20260720/traces-per-gpu-low SIM_GENERATION=per-gpu-v1 PREDICTOR_CACHE_GENERATION=v4 bash runs/frontier-workload-regime-taxonomy-v0/run_simulator_group.sh 4 true"
artifacts = []

9
runs/frontier-workload-regime-taxonomy-v0/run_simulator_group.sh
View File

@@ -25,6 +25,7 @@ ALLREDUCE_CSV="${ALLREDUCE_CSV:-/home/admin/cpfs/wjh/aituner/aituner-graph-piece
CUSTOM_PYTHON_DEPS="${CUSTOM_PYTHON_DEPS:-${EXPERIMENT_ROOT}/python-deps-v2}"
BASE_PYTHON_DEPS="${BASE_PYTHON_DEPS:-/home/admin/cpfs/wjh/venvs/qwen36-vllm-0.20.2/lib/python3.12/site-packages}"
SIM_GENERATION="${SIM_GENERATION:-v4}"
PREDICTOR_CACHE_GENERATION="${PREDICTOR_CACHE_GENERATION:-${SIM_GENERATION}}"
RUNNER="${CHECKOUT}/runs/frontier-fidelity-envelope-v1/run_frontier_qwen30_exact_trace_surface.py"
for path in "${TRACES_ROOT}/manifest.json" "${FRONTIER_SOURCE}" \
@@ -35,10 +36,15 @@ for path in "${TRACES_ROOT}/manifest.json" "${FRONTIER_SOURCE}" \
done
run_mode() {
local prefix_mode="$1" output_root prefix_flag family
local prefix_mode="$1" output_root predictor_cache_root prefix_flag family
local -a families traces configs command
output_root="${EXPERIMENT_ROOT}/sim-${SIM_GENERATION}/tp${TP}-prefix-${prefix_mode}"
predictor_cache_root="${EXPERIMENT_ROOT}/sim-${PREDICTOR_CACHE_GENERATION}/tp${TP}-prefix-${prefix_mode}/cache"
mkdir -p "${output_root}/provenance"
[[ -d "${predictor_cache_root}" ]] || {
echo "ERROR: missing predictor cache ${predictor_cache_root}" >&2
exit 1
}
if [[ "${prefix_mode}" == "true" ]]; then
prefix_flag="--prefix-caching"
@@ -99,6 +105,7 @@ run_mode() {
--kernel-profile-root "${KERNEL_PROFILE_ROOT}" \
--python-deps "${CUSTOM_PYTHON_DEPS}" \
--output-root "${output_root}" \
--predictor-cache-root "${predictor_cache_root}" \
"${command[@]}" "${configs[@]}" \
--rate-contract trace-window "${prefix_flag}" \
--cc-backend vidur --allreduce-csv "${ALLREDUCE_CSV}" \