T.E.D.D. 1140 S

T.E.D.D. 1104 S: 26M Parameters. 260MB

Hyper Parameters

control_mode: keyboard
cnn_model_name: efficientnet_v2_s
pretrained_cnn: true
encoder_type: transformer
embedded_size: 384
nhead: 8
num_layers_encoder: 2
learning_rate: 1e-05
optimizer_name: adafactor
scheduler_name: cosine
warmup_factor: 0.05
max_epochs: 20
batch_size: 32
mask_prob: 0.2
dropout_cnn_out: 0.3
dropout_encoder: 0.1
dropout_encoder_features: 0.3
positional_embeddings_dropout: 0.1
weight_decay: 1e-3

Accuracy in the test datasets:

	Time	Weather	Micro-Acc K@1	Micro-Acc k@3	Macro-Acc K@1
City	🌞	☀️	51.0	83.0	46.3
City	🌞	☔	49.0	82.5	45.2
City	🌛	☀️	56.3	86.6	49.0
City	🌛	☔	49.4	81.4	42.5
Highway	🌞	☀️	70.3	100	68.5
Highway	🌞	☔	71.2	100	37.6
Highway	🌛	☀️	80.9	100	49.1
Highway	🌛	☔	69.3	100	61.1

The release includes the best epoch in the development set and the last epoch.