Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,39 +1,35 @@
|
|
1 |
|
2 |
|
3 |
-
import whisper
|
4 |
import gradio as gr
|
|
|
5 |
from accelerate import init_empty_weights, load_checkpoint_and_dispatch
|
6 |
-
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
|
7 |
-
|
8 |
-
# Initialize the device map for ZeRO
|
9 |
-
from accelerate.utils import set_module_tensor_to_device
|
10 |
import torch
|
11 |
|
12 |
-
|
|
|
13 |
print(f"Using ZeRO-powered device map: {device_map}")
|
14 |
|
15 |
-
# Load the model using ZeRO
|
16 |
-
model_name = "
|
17 |
|
18 |
-
|
19 |
with init_empty_weights():
|
20 |
-
whisper_model = whisper.load_model(model_name)
|
21 |
|
22 |
-
#
|
23 |
-
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
24 |
-
|
25 |
-
# Load model with Accelerate/ZeRO
|
26 |
whisper_model = load_checkpoint_and_dispatch(
|
27 |
whisper_model,
|
28 |
device_map=device_map,
|
29 |
-
dtype=torch.float16 #
|
30 |
)
|
31 |
|
|
|
|
|
32 |
# Define the transcription function
|
33 |
def transcribe(audio):
|
34 |
# Perform transcription using the Whisper model
|
35 |
result = whisper_model.transcribe(audio)
|
36 |
-
return result[
|
37 |
|
38 |
# Create the Gradio interface
|
39 |
demo = gr.Interface(
|
@@ -41,7 +37,7 @@ demo = gr.Interface(
|
|
41 |
inputs=gr.Audio(source="microphone", type="filepath", label="Speak into the microphone"), # Input audio
|
42 |
outputs=gr.Textbox(label="Transcription"), # Output transcription
|
43 |
title="Whisper Speech-to-Text with ZeRO", # Title of the interface
|
44 |
-
description="Record audio using your microphone and get a transcription using the Whisper model optimized
|
45 |
)
|
46 |
|
47 |
# Launch the Gradio interface
|
|
|
1 |
|
2 |
|
|
|
3 |
import gradio as gr
|
4 |
+
import whisper
|
5 |
from accelerate import init_empty_weights, load_checkpoint_and_dispatch
|
|
|
|
|
|
|
|
|
6 |
import torch
|
7 |
|
8 |
+
# Check if GPU is available and set up device map
|
9 |
+
device_map = "auto" # Automatically balance layers across available devices
|
10 |
print(f"Using ZeRO-powered device map: {device_map}")
|
11 |
|
12 |
+
# Load the Whisper model using Accelerate with ZeRO
|
13 |
+
model_name = "tiny" # Change to "base", "small", etc., as needed
|
14 |
|
15 |
+
print(f"Loading the Whisper model: {model_name} with ZeRO optimization...")
|
16 |
with init_empty_weights():
|
17 |
+
whisper_model = whisper.load_model(model_name) # Load model structure without weights
|
18 |
|
19 |
+
# Dispatch the model across devices using ZeRO
|
|
|
|
|
|
|
20 |
whisper_model = load_checkpoint_and_dispatch(
|
21 |
whisper_model,
|
22 |
device_map=device_map,
|
23 |
+
dtype=torch.float16 # Use mixed precision for efficiency
|
24 |
)
|
25 |
|
26 |
+
print("Model successfully loaded with ZeRO optimization!")
|
27 |
+
|
28 |
# Define the transcription function
|
29 |
def transcribe(audio):
|
30 |
# Perform transcription using the Whisper model
|
31 |
result = whisper_model.transcribe(audio)
|
32 |
+
return result["text"]
|
33 |
|
34 |
# Create the Gradio interface
|
35 |
demo = gr.Interface(
|
|
|
37 |
inputs=gr.Audio(source="microphone", type="filepath", label="Speak into the microphone"), # Input audio
|
38 |
outputs=gr.Textbox(label="Transcription"), # Output transcription
|
39 |
title="Whisper Speech-to-Text with ZeRO", # Title of the interface
|
40 |
+
description="Record audio using your microphone and get a transcription using the Whisper model optimized with ZeRO."
|
41 |
)
|
42 |
|
43 |
# Launch the Gradio interface
|