Update app.py
Browse files
app.py
CHANGED
@@ -2,7 +2,7 @@ import gradio as gr
|
|
2 |
from transformers import WhisperProcessor, WhisperForConditionalGeneration
|
3 |
from datasets import load_dataset
|
4 |
import torch
|
5 |
-
import librosa
|
6 |
|
7 |
# 加载 Whisper 模型和 processor
|
8 |
model_name = "openai/whisper-large-v3-turbo"
|
@@ -15,8 +15,8 @@ ds = load_dataset("CoIR-Retrieval/CodeSearchNet-php-queries-corpus")
|
|
15 |
|
16 |
def transcribe(audio_path):
|
17 |
# 加载音频文件并转换为信号
|
18 |
-
audio, sr = librosa.load(audio_path, sr=16000)
|
19 |
-
input_values = processor(
|
20 |
|
21 |
# 模型推理
|
22 |
with torch.no_grad():
|
@@ -32,7 +32,7 @@ def transcribe(audio_path):
|
|
32 |
# Gradio 界面
|
33 |
iface = gr.Interface(
|
34 |
fn=transcribe,
|
35 |
-
inputs=gr.Audio( type="filepath"),
|
36 |
outputs="text",
|
37 |
title="Whisper Transcription for Developers",
|
38 |
description="使用 Whisper 和 bigcode 数据集转录开发者相关术语。"
|
|
|
2 |
from transformers import WhisperProcessor, WhisperForConditionalGeneration
|
3 |
from datasets import load_dataset
|
4 |
import torch
|
5 |
+
# import librosa
|
6 |
|
7 |
# 加载 Whisper 模型和 processor
|
8 |
model_name = "openai/whisper-large-v3-turbo"
|
|
|
15 |
|
16 |
def transcribe(audio_path):
|
17 |
# 加载音频文件并转换为信号
|
18 |
+
# audio, sr = librosa.load(audio_path, sr=16000)
|
19 |
+
input_values = processor(audio_path, return_tensors="pt", sampling_rate=16000).input_values
|
20 |
|
21 |
# 模型推理
|
22 |
with torch.no_grad():
|
|
|
32 |
# Gradio 界面
|
33 |
iface = gr.Interface(
|
34 |
fn=transcribe,
|
35 |
+
inputs=gr.Audio(sources="microphone", type="filepath"),
|
36 |
outputs="text",
|
37 |
title="Whisper Transcription for Developers",
|
38 |
description="使用 Whisper 和 bigcode 数据集转录开发者相关术语。"
|