diff --git a/edml/controllers/strategy_optimization.py b/edml/controllers/strategy_optimization.py
index eddbf5742a526285cf6a3793b208900beda9f3ec..cd7bae8ac0db1c8eb4bcf818713ed8b2f1ed0e47 100644
--- a/edml/controllers/strategy_optimization.py
+++ b/edml/controllers/strategy_optimization.py
@@ -123,13 +123,15 @@ class ServerChoiceOptimizer:
self.global_params.train_global_time is not None
and self.global_params.last_server_device_id is not None
):
- return (
+ latency = (
self.global_params.train_global_time
- self._round_runtime_with_server_no_latency(
self.global_params.last_server_device_id
)
)
- return 0 # latency not known
+ if latency > 0:
+ return latency
+ return 0 # latency not known or runtime was overestimated previously
def _round_runtime_with_server_no_latency(self, server_device_id):
"""
@@ -356,20 +358,7 @@ class EnergySimulator:
device_params_list, global_params
)
- def simulate_greedy_selection(self):
- """
- Simulates the greedy server choice algorithm.
- Returns:
- num_rounds: number of rounds until the first device runs out of battery
- server_selection_schedule: list of server device ids for each round
- device_batteries: list of battery levels for each device after the last successful round
- """
-
- def __get_device_with_max_battery__(device_battery_list):
- return max(
- range(len(device_battery_list)), key=device_battery_list.__getitem__
- )
-
+ def _simulate_selection(self, selection_callback=None):
def __all_devices_alive__(device_battery_list):
return all(battery > 0 for battery in device_battery_list)
@@ -394,7 +383,9 @@ class EnergySimulator:
server_selection_schedule = []
num_rounds = 0
while all_devices_alive:
- server_device_idx = __get_device_with_max_battery__(device_batteries)
+ server_device_idx = selection_callback(
+ device_battery_list=device_batteries, num_rounds=num_rounds
+ )
new_batteries = device_batteries.copy()
for idx, device in enumerate(self.device_params_list):
new_batteries[idx] = new_batteries[idx] - energy[idx][server_device_idx]
@@ -409,6 +400,36 @@ class EnergySimulator:
break
return num_rounds, server_selection_schedule, device_batteries
+ def simulate_greedy_selection(self):
+ """
+ Simulates the greedy server choice algorithm.
+ Returns:
+ num_rounds: number of rounds until the first device runs out of battery
+ server_selection_schedule: list of server device ids for each round
+ device_batteries: list of battery levels for each device after the last successful round
+ """
+
+ def __get_device_with_max_battery__(device_battery_list, **kwargs):
+ return max(
+ range(len(device_battery_list)), key=device_battery_list.__getitem__
+ )
+
+ return self._simulate_selection(__get_device_with_max_battery__)
+
+ def simulate_sequential_selection(self):
+ """
+ Simulates the sequential server choice algorithm.
+ Returns:
+ num_rounds: number of rounds until the first device runs out of battery
+ server_selection_schedule: list of server device ids for each round
+ device_batteries: list of battery levels for each device after the last successful round
+ """
+
+ def __sequential_selection__(device_battery_list, num_rounds):
+ return num_rounds % len(device_battery_list)
+
+ return self._simulate_selection(__sequential_selection__)
+
def simulate_smart_selection(self):
"""
Simulates the smart server choice algorithm.
@@ -574,47 +595,52 @@ def run_grid_search(
for partition in partitions:
for cost_sec in cost_per_sec:
for cost_sent in cost_per_byte_sent:
- for cost_received in cost_per_byte_received:
- for cost_flop in cost_per_flop:
- global_params.cost_per_sec = cost_sec
- global_params.cost_per_byte_sent = cost_sent
- global_params.cost_per_byte_received = cost_received
- global_params.cost_per_flop = cost_flop
- for idx, device in enumerate(device_params_list):
- device.current_battery = battery[idx]
- device.comp_latency_factor = latency[idx]
- device.train_samples = (
- partition[idx] * total_train_samples
- )
- device.validation_samples = (
- partition[idx] * total_val_samples
- )
- energy_simulator = EnergySimulator(
- device_params_list, global_params
- )
- num_rounds_smart, _, _ = (
- energy_simulator.simulate_smart_selection()
- )
- num_rounds_greedy, _, _ = (
- energy_simulator.simulate_greedy_selection()
- )
- num_rounds_fl, _ = (
- energy_simulator.simulate_federated_learning()
+ # for cost_received in cost_per_byte_received:
+ cost_received = cost_sent
+ for cost_flop in cost_per_flop:
+ global_params.cost_per_sec = cost_sec
+ global_params.cost_per_byte_sent = cost_sent
+ global_params.cost_per_byte_received = cost_received
+ global_params.cost_per_flop = cost_flop
+ for idx, device in enumerate(device_params_list):
+ device.current_battery = battery[idx]
+ device.comp_latency_factor = latency[idx]
+ device.train_samples = (
+ partition[idx] * total_train_samples
)
- results.append(
- {
- "battery": battery,
- "latency": latency,
- "partition": partition,
- "cost_per_sec": cost_sec,
- "cost_per_byte_sent": cost_sent,
- "cost_per_byte_received": cost_received,
- "cost_per_flop": cost_flop,
- "num_rounds_smart": num_rounds_smart,
- "num_rounds_greedy": num_rounds_greedy,
- "num_rounds_fl": num_rounds_fl,
- }
+ device.validation_samples = (
+ partition[idx] * total_val_samples
)
+ energy_simulator = EnergySimulator(
+ device_params_list, global_params
+ )
+ num_rounds_smart, _, _ = (
+ energy_simulator.simulate_smart_selection()
+ )
+ num_rounds_greedy, _, _ = (
+ energy_simulator.simulate_greedy_selection()
+ )
+ num_rounds_seq, _, _ = (
+ energy_simulator.simulate_sequential_selection()
+ )
+ num_rounds_fl, _ = (
+ energy_simulator.simulate_federated_learning()
+ )
+ results.append(
+ {
+ "battery": battery,
+ "latency": latency,
+ "partition": partition,
+ "cost_per_sec": cost_sec,
+ "cost_per_byte_sent": cost_sent,
+ "cost_per_byte_received": cost_received,
+ "cost_per_flop": cost_flop,
+ "num_rounds_smart": num_rounds_smart,
+ "num_rounds_seq": num_rounds_seq,
+ "num_rounds_greedy": num_rounds_greedy,
+ "num_rounds_fl": num_rounds_fl,
+ }
+ )
return results