Spaces:
Running
Running
Commit
·
92da7ef
1
Parent(s):
6e0def0
CPU
Browse files- app.py +4 -5
- config.toml +2 -2
- data_loading/feature_extractor.py +2 -2
- data_loading/pretrained_extractors.py +2 -2
app.py
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
-
import spaces
|
2 |
import gradio as gr
|
3 |
import torchaudio
|
4 |
import pandas as pd
|
@@ -16,8 +16,8 @@ import torch
|
|
16 |
from models.models import BiFormer
|
17 |
|
18 |
|
19 |
-
DEVICE = torch.device('cuda')
|
20 |
-
|
21 |
|
22 |
# Configure logging
|
23 |
logging.basicConfig(level=logging.INFO)
|
@@ -150,7 +150,7 @@ def get_top_emotion(probabilities):
|
|
150 |
return f"{LABEL_TO_EMOTION[max_idx]} ({max(probabilities)*100:.1f}%)"
|
151 |
|
152 |
|
153 |
-
@spaces.GPU
|
154 |
def process_audio(audio_path):
|
155 |
"""Main processing pipeline."""
|
156 |
try:
|
@@ -312,6 +312,5 @@ def create_demo():
|
|
312 |
|
313 |
|
314 |
if __name__ == "__main__":
|
315 |
-
print('START DEMO')
|
316 |
demo = create_demo()
|
317 |
demo.launch()
|
|
|
1 |
+
# import spaces
|
2 |
import gradio as gr
|
3 |
import torchaudio
|
4 |
import pandas as pd
|
|
|
16 |
from models.models import BiFormer
|
17 |
|
18 |
|
19 |
+
# DEVICE = torch.device('cuda')
|
20 |
+
DEVICE = torch.device('cpu')
|
21 |
|
22 |
# Configure logging
|
23 |
logging.basicConfig(level=logging.INFO)
|
|
|
150 |
return f"{LABEL_TO_EMOTION[max_idx]} ({max(probabilities)*100:.1f}%)"
|
151 |
|
152 |
|
153 |
+
# @spaces.GPU
|
154 |
def process_audio(audio_path):
|
155 |
"""Main processing pipeline."""
|
156 |
try:
|
|
|
312 |
|
313 |
|
314 |
if __name__ == "__main__":
|
|
|
315 |
demo = create_demo()
|
316 |
demo.launch()
|
config.toml
CHANGED
@@ -54,7 +54,7 @@ text_column = "text"
|
|
54 |
whisper_model = "base"
|
55 |
|
56 |
# Указываем, где запускать Whisper: "cuda" (GPU) или "cpu"
|
57 |
-
whisper_device = "
|
58 |
|
59 |
# Если для dev/test в CSV нет текста, нужно ли всё же вызывать Whisper?
|
60 |
use_whisper_for_nontrain_if_no_text = true
|
@@ -120,7 +120,7 @@ audio_embedding_dim = 256 # размерность аудио-эмбеддин
|
|
120 |
text_embedding_dim = 1024 # размерность текст-эмбеддинга
|
121 |
emb_normalize = false # нормализовать ли вектор L2-нормой
|
122 |
max_tokens = 95 # ограничение на длину текста (токенов) при токенизации
|
123 |
-
device = "
|
124 |
|
125 |
# audio_pooling = "mean" # "mean", "cls", "max", "min", "last", "attention"
|
126 |
# text_pooling = "cls" # "mean", "cls", "max", "min", "last", "sum", "attention"
|
|
|
54 |
whisper_model = "base"
|
55 |
|
56 |
# Указываем, где запускать Whisper: "cuda" (GPU) или "cpu"
|
57 |
+
whisper_device = "cpu"
|
58 |
|
59 |
# Если для dev/test в CSV нет текста, нужно ли всё же вызывать Whisper?
|
60 |
use_whisper_for_nontrain_if_no_text = true
|
|
|
120 |
text_embedding_dim = 1024 # размерность текст-эмбеддинга
|
121 |
emb_normalize = false # нормализовать ли вектор L2-нормой
|
122 |
max_tokens = 95 # ограничение на длину текста (токенов) при токенизации
|
123 |
+
device = "cpu" # "cuda" или "cpu", куда грузить модель
|
124 |
|
125 |
# audio_pooling = "mean" # "mean", "cls", "max", "min", "last", "attention"
|
126 |
# text_pooling = "cls" # "mean", "cls", "max", "min", "last", "sum", "attention"
|
data_loading/feature_extractor.py
CHANGED
@@ -14,8 +14,8 @@ from transformers import (
|
|
14 |
from data_loading.pretrained_extractors import EmotionModel, get_model_mamba, Mamba
|
15 |
|
16 |
|
17 |
-
DEVICE = torch.device('cuda')
|
18 |
-
|
19 |
|
20 |
|
21 |
class PretrainedAudioEmbeddingExtractor:
|
|
|
14 |
from data_loading.pretrained_extractors import EmotionModel, get_model_mamba, Mamba
|
15 |
|
16 |
|
17 |
+
# DEVICE = torch.device('cuda')
|
18 |
+
DEVICE = torch.device('cpu')
|
19 |
|
20 |
|
21 |
class PretrainedAudioEmbeddingExtractor:
|
data_loading/pretrained_extractors.py
CHANGED
@@ -14,8 +14,8 @@ from einops import rearrange, einsum
|
|
14 |
from torch import Tensor
|
15 |
from einops import rearrange
|
16 |
|
17 |
-
DEVICE = torch.device('cuda')
|
18 |
-
|
19 |
|
20 |
## Audio models
|
21 |
|
|
|
14 |
from torch import Tensor
|
15 |
from einops import rearrange
|
16 |
|
17 |
+
# DEVICE = torch.device('cuda')
|
18 |
+
DEVICE = torch.device('cpu')
|
19 |
|
20 |
## Audio models
|
21 |
|