https://github.com/nixtla logo
#neural-forecast
Title
# neural-forecast
o

Oleksii Kaplenko

07/11/2023, 6:06 AM
Hi, I'm slightly confused about tensorboard charts. The model I trained:
models = [NHITS(h=horizon,
input_size=6*horizon,
loss=MSE(),
valid_loss=MSE(),
exclude_insample_y=False,
max_steps=100000,
val_check_steps=10,
batch_size=32,
valid_batch_size=32,
n_freq_downsample=[8, 4, 2, 1, 1],
n_blocks=5*[1],
mlp_units=5 * [[64, 64]],
dropout_prob_theta=0.2,
hist_exog_list=hist_exog_list,
learning_rate=1e-4,
step_size=step_size,
scaler_type='identity'
)
]
3 Views