
############################## Experiment details ##############################

dataset: speech_commands
dataset_dir: ../datasets/speech_commands/
output_dir: output/debug/signavg-speech_commands-2022-07-10_19:43:40/
model: mel
num_rounds: 100
eval_every: 10
ServerType: <class 'rayleaf.entities.server.Server'>
client_types: [(<class 'rayleaf.entities.client.Client'>, 200)]
clients_per_round: 40
client_lr: 0.01
batch_size: 64
seed: 0
use_val_set: False
num_epochs: 5
gpus_per_client_cluster: 0.75
num_client_clusters: 8
save_model: False

############################## Simulation ##############################

Spawning 8 ClientClusters using cuda device (this may take a while)
200 total clients: 200 Clients

>>> Training Accuracy: Round 0 <<<
+-----------+-----------+-------------------+-------------------+
|   average |    median |   10th percentile |   90th percentile |
|-----------+-----------+-------------------+-------------------|
| 0.0282357 | 0.0226575 |        0.00980392 |         0.0520833 |
+-----------+-----------+-------------------+-------------------+
>>> Test Accuracy: Round 0 <<<
+-----------+-----------+-------------------+-------------------+
|   average |    median |   10th percentile |   90th percentile |
|-----------+-----------+-------------------+-------------------|
| 0.0296071 | 0.0357599 |                 0 |         0.0833333 |
+-----------+-----------+-------------------+-------------------+
--- Round 1 of 100: Training 40 clients: 40 Clients ---
--- Round 2 of 100: Training 40 clients: 40 Clients ---
--- Round 3 of 100: Training 40 clients: 40 Clients ---
