Delete emotion_model.py
Browse files- emotion_model.py +0 -52
emotion_model.py
DELETED
|
@@ -1,52 +0,0 @@
|
|
| 1 |
-
import tensorflow as tf
|
| 2 |
-
import torch
|
| 3 |
-
import numpy as np
|
| 4 |
-
import os
|
| 5 |
-
|
| 6 |
-
# Load the Keras model
|
| 7 |
-
keras_model = tf.keras.models.load_model('wav2vec_model.h5')
|
| 8 |
-
|
| 9 |
-
# Create a PyTorch model with the same architecture
|
| 10 |
-
class EmotionClassifier(torch.nn.Module):
|
| 11 |
-
def __init__(self, input_shape, num_classes):
|
| 12 |
-
super().__init__()
|
| 13 |
-
# Adjust this architecture to match your Keras model
|
| 14 |
-
self.flatten = torch.nn.Flatten()
|
| 15 |
-
self.layers = torch.nn.Sequential(
|
| 16 |
-
torch.nn.Linear(input_shape, 128),
|
| 17 |
-
torch.nn.ReLU(),
|
| 18 |
-
torch.nn.Dropout(0.3),
|
| 19 |
-
torch.nn.Linear(128, 64),
|
| 20 |
-
torch.nn.ReLU(),
|
| 21 |
-
torch.nn.Dropout(0.3),
|
| 22 |
-
torch.nn.Linear(64, num_classes)
|
| 23 |
-
)
|
| 24 |
-
|
| 25 |
-
def forward(self, x):
|
| 26 |
-
x = self.flatten(x)
|
| 27 |
-
return self.layers(x)
|
| 28 |
-
|
| 29 |
-
# Create PyTorch model
|
| 30 |
-
# Adjust these parameters based on your Keras model
|
| 31 |
-
input_shape = 13 * 128 # n_mfcc * max_length
|
| 32 |
-
num_classes = 7 # Number of emotions
|
| 33 |
-
pytorch_model = EmotionClassifier(input_shape, num_classes)
|
| 34 |
-
|
| 35 |
-
# Copy weights from Keras to PyTorch
|
| 36 |
-
# This would need to be adjusted based on your exact architecture
|
| 37 |
-
for i, layer in enumerate(keras_model.layers):
|
| 38 |
-
if isinstance(layer, tf.keras.layers.Dense):
|
| 39 |
-
# Get Keras weights and bias
|
| 40 |
-
keras_weights = layer.get_weights()[0]
|
| 41 |
-
keras_bias = layer.get_weights()[1]
|
| 42 |
-
|
| 43 |
-
# Find the corresponding PyTorch layer
|
| 44 |
-
# This is simplified; you'd need to match layers properly
|
| 45 |
-
pytorch_layer = pytorch_model.layers[i * 2]
|
| 46 |
-
|
| 47 |
-
# Copy weights and bias
|
| 48 |
-
pytorch_layer.weight.data = torch.tensor(keras_weights.T, dtype=torch.float32)
|
| 49 |
-
pytorch_layer.bias.data = torch.tensor(keras_bias, dtype=torch.float32)
|
| 50 |
-
|
| 51 |
-
# Save the PyTorch model
|
| 52 |
-
torch.save(pytorch_model.state_dict(), 'emotion_model.pt')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|