| from src.model.base_model import FingeringPredictor, FingeringPredictorBaseline |
| import torch |
| from typing import Any, List |
| from src.config import NUM_FRETS, NUM_STRINGS, PITCH_CLASSES |
| from torch import nn |
|
|
|
|
| class MultiLayerFingeringPredictor(FingeringPredictor): |
| def __init__(self, hidden_layers_sizes: List[int], |
| with_mute: bool = False, learning_rate: float = 0.001, |
| *args: Any, **kwargs: Any) -> None: |
| super().__init__(with_mute=with_mute, learning_rate=learning_rate, **kwargs) |
| |
| input_layer = nn.Linear((self.num_frets*NUM_STRINGS) + 2*PITCH_CLASSES, |
| out_features=hidden_layers_sizes[0]) |
| output_layer = nn.Linear(hidden_layers_sizes[-1], out_features=NUM_STRINGS*self.num_frets) |
| hidden_layers = [] |
| if len(hidden_layers_sizes) > 1: |
| for i, size in enumerate(hidden_layers_sizes[1:]): |
| layer = nn.Linear(in_features=hidden_layers_sizes[i-1], out_features=size) |
| hidden_layers.append(layer) |
| self.model = nn.Sequential(input_layer, |
| *hidden_layers, |
| output_layer) |
| print(self.model) |
|
|
|
|
| class MultilayerBaseline(FingeringPredictorBaseline): |
| def __init__(self, hidden_layers_sizes: List[int], |
| with_mute: bool = False, learning_rate: float = 0.001, |
| *args: Any, **kwargs: Any) -> None: |
| super().__init__(with_mute=with_mute, learning_rate=learning_rate, **kwargs) |
| |
| input_layer = nn.Linear(2*PITCH_CLASSES, |
| out_features=hidden_layers_sizes[0]) |
| output_layer = nn.Linear(hidden_layers_sizes[-1], out_features=NUM_STRINGS*self.num_frets) |
| hidden_layers = [] |
| if len(hidden_layers_sizes) > 1: |
| for i, size in enumerate(hidden_layers_sizes[1:]): |
| layer = nn.Linear(in_features=hidden_layers_sizes[i-1], out_features=size) |
| hidden_layers.append(layer) |
| self.model = nn.Sequential(input_layer, |
| *hidden_layers, |
| output_layer) |
| print(self.model) |
|
|
|
|
