inference.py

from iterativedecomposition import Model, OverfitResonanceModel, n_frames, samplerate, n_samples
import torch

from util import playable
from util.playable import listen_to_sound

hidden_channels = 512
device = 'cpu'

resonance_model = OverfitResonanceModel(
    n_noise_filters=32,
    noise_expressivity=8,
    noise_filter_samples=128,
    noise_deformations=16,
    instr_expressivity=8,
    n_events=1,
    n_resonances=4096,
    n_envelopes=256,
    n_decays=32,
    n_deformations=32,
    n_samples=n_samples,
    n_frames=n_frames,
    samplerate=samplerate,
    hidden_channels=hidden_channels,
    wavetable_device=device,
    fine_positioning=True
).to(device)

model = Model(
    resonance_model=resonance_model,
    in_channels=1024,
    hidden_channels=hidden_channels).to(device)

model.load_state_dict(torch.load(
    'iterativedecomposition2.dat',
    map_location=lambda storage, loc: storage))

rnd = model.random_sequence(device=device)
print(rnd.shape)
rnd = playable(rnd, samplerate=22050, normalize=True)
# listen_to_sound(rnd)