Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,29 +1,98 @@
|
|
1 |
|
2 |
-
|
3 |
-
|
4 |
'''
|
5 |
This script calls the ada model from openai api to predict the next few words.
|
6 |
'''
|
7 |
import os
|
8 |
-
import openai
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
|
|
|
|
|
2 |
'''
|
3 |
This script calls the ada model from openai api to predict the next few words.
|
4 |
'''
|
5 |
import os
|
6 |
+
#import openai
|
7 |
+
import os
|
8 |
+
from pprint import pprint
|
9 |
+
os.system("pip install git+https://github.com/openai/whisper.git")
|
10 |
+
import gradio as gr
|
11 |
+
import whisper
|
12 |
+
from transformers import pipeline
|
13 |
+
import torch
|
14 |
+
from transformers import AutoModelForCausalLM
|
15 |
+
from transformers import AutoTokenizer
|
16 |
+
import time
|
17 |
+
# import streaming.py
|
18 |
+
# from next_word_prediction import GPT2
|
19 |
+
|
20 |
+
|
21 |
+
|
22 |
+
|
23 |
+
#gpt2 = AutoModelForCausalLM.from_pretrained("gpt2", return_dict_in_generate=True)
|
24 |
+
#tokenizer = AutoTokenizer.from_pretrained("gpt2")
|
25 |
+
|
26 |
+
### /code snippet
|
27 |
+
|
28 |
+
|
29 |
+
# get gpt2 model
|
30 |
+
generator = pipeline('text-generation', model='gpt2')
|
31 |
+
|
32 |
+
# whisper model specification
|
33 |
+
model = whisper.load_model("tiny")
|
34 |
+
|
35 |
+
|
36 |
+
|
37 |
+
def inference(audio, state=""):
|
38 |
+
|
39 |
+
#time.sleep(2)
|
40 |
+
#text = p(audio)["text"]
|
41 |
+
#state += text + " "
|
42 |
+
# load audio data
|
43 |
+
audio = whisper.load_audio(audio)
|
44 |
+
# ensure sample is in correct format for inference
|
45 |
+
audio = whisper.pad_or_trim(audio)
|
46 |
+
|
47 |
+
# generate a log-mel spetrogram of the audio data
|
48 |
+
mel = whisper.log_mel_spectrogram(audio).to(model.device)
|
49 |
+
|
50 |
+
_, probs = model.detect_language(mel)
|
51 |
+
|
52 |
+
# decode audio data
|
53 |
+
options = whisper.DecodingOptions(fp16 = False)
|
54 |
+
# transcribe speech to text
|
55 |
+
result = whisper.decode(model, mel, options)
|
56 |
+
|
57 |
+
PROMPT = """The following is a transcript of a conversation. Predict a few nouns, verbs, or adjectives that may be used next. Predict the next few words as a list of options.
|
58 |
+
A few examples are provided below and then the current transcript is provided.
|
59 |
+
Examples:
|
60 |
+
Transcript: Tomorrow night we're going out to
|
61 |
+
Next: The Movies, A Restaurant, A Baseball Game, The Theater, A Party
|
62 |
+
Transcript: I would like to order a cheeseburger with a side of
|
63 |
+
Next: Fries, Milkshake, Apples, Salad, Katsup
|
64 |
+
"""
|
65 |
+
text = PROMPT + result.text
|
66 |
+
|
67 |
+
openai.api_key = os.environ["Openai_APIkey"]
|
68 |
+
|
69 |
+
response = openai.Completion.create(
|
70 |
+
model="text-ada-001",
|
71 |
+
prompt=text,
|
72 |
+
temperature=1,
|
73 |
+
max_tokens=4,
|
74 |
+
n=4)
|
75 |
+
for i in range(4):
|
76 |
+
print(response['choices'][i]['text'])
|
77 |
+
|
78 |
+
|
79 |
+
# result.text
|
80 |
+
#return getText, gr.update(visible=True), gr.update(visible=True), gr.update(visible=True)
|
81 |
+
return result.text, state, response
|
82 |
+
|
83 |
+
|
84 |
+
|
85 |
+
# get audio from microphone
|
86 |
+
|
87 |
+
gr.Interface(
|
88 |
+
fn=inference,
|
89 |
+
inputs=[
|
90 |
+
gr.inputs.Audio(source="microphone", type="filepath"),
|
91 |
+
"state"
|
92 |
+
],
|
93 |
+
outputs=[
|
94 |
+
"textbox",
|
95 |
+
"state",
|
96 |
+
"textbox"
|
97 |
+
],
|
98 |
+
live=True).launch()
|