Gijs Wijngaard
commited on
Commit
·
252b17b
1
Parent(s):
1311855
init
Browse files
app.py
CHANGED
|
@@ -3,7 +3,7 @@ import gradio as gr
|
|
| 3 |
import torch
|
| 4 |
import torchaudio
|
| 5 |
from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq
|
| 6 |
-
|
| 7 |
|
| 8 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 9 |
model_name = "ibm-granite/granite-speech-3.3-8b"
|
|
@@ -23,7 +23,7 @@ def _load_audio_mono_16k(file_path: str) -> torch.Tensor:
|
|
| 23 |
wav = torchaudio.functional.resample(wav, sr, 16000)
|
| 24 |
return wav
|
| 25 |
|
| 26 |
-
|
| 27 |
def process_audio(audio_path: str, instruction: str, max_tokens: int = 200) -> str:
|
| 28 |
if not audio_path:
|
| 29 |
return "Please upload an audio file."
|
|
|
|
| 3 |
import torch
|
| 4 |
import torchaudio
|
| 5 |
from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq
|
| 6 |
+
import spaces
|
| 7 |
|
| 8 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 9 |
model_name = "ibm-granite/granite-speech-3.3-8b"
|
|
|
|
| 23 |
wav = torchaudio.functional.resample(wav, sr, 16000)
|
| 24 |
return wav
|
| 25 |
|
| 26 |
+
@spaces.GPU
|
| 27 |
def process_audio(audio_path: str, instruction: str, max_tokens: int = 200) -> str:
|
| 28 |
if not audio_path:
|
| 29 |
return "Please upload an audio file."
|