Gas_benchmarks: launch benchmark for `Ticket_costs.Constants.cost_collect_tickets_step`
The MR !4491 (merged) adds a new benchmark COLLECT_TICKETS_STEP_alpha, defined in Ticket_benchmarks. This issue is to launch that benchmark and if necessary correct the value of Ticket_costs.Constants.cost_collect_tickets_step
.
For example, the script
BENCHMARK="COLLECT_TICKETS_STEP_alpha"
MODEL="size_collect_tickets_step_model"
BENCH_NUM=1000
NSAMPLES=500
SEED=1234
WORKLOAD_FILE="${BENCHMARK}.workload"
CSV_FILE="${BENCHMARK}.csv"
# Run benchmark
./tezos-snoop benchmark \
"${BENCHMARK}" \
and save to "${BENCHMARK}.workload" \
--bench-num "${BENCH_NUM}" \
--nsamples "${NSAMPLES}" \
--seed "${SEED}"
# Infer parameters
./tezos-snoop \
infer parameters for model "${MODEL}" \
on data "${WORKLOAD_FILE}" using lasso \
--lasso-positive \
--dump-csv "${CSV_FILE}" \
--plot
will run the benchmark. I don't know what values of BENCH_NUM
and NSAMPLES
we typically use so those may need to be changed.
Edited by Thomas Athorne