ysharma HF staff commited on
Commit
6c3ad82
1 Parent(s): 006d225
Files changed (1) hide show
  1. app.py +39 -19
app.py CHANGED
@@ -4,32 +4,52 @@ import gradio as gr
4
  import whisper
5
 
6
  model = whisper.load_model("base")
 
 
 
 
 
7
 
8
  def fun(audio) : #, state=''):
9
- text = model.transcribe(audio)["text"]
10
- #state += text + " "
11
- return text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
12
 
13
  def fun1(audio, state=''):
14
  text = model.transcribe(audio)["text"]
15
  state += text + " "
16
  return state, state
17
 
18
- # Set the starting state to an empty string
19
-
20
- #gr.Interface(
21
- # fn=transcribe,
22
- # inputs=[
23
- # gr.Audio(source="microphone", type="filepath", streaming=True),
24
- # "state"
25
- # ],
26
- # outputs=[
27
- # "textbox",
28
- # "state"
29
- # ],
30
- # live=True).launch()
31
-
32
-
33
  gr.Interface(
34
  title = 'Testing Whisper',
35
  fn=fun,
@@ -38,6 +58,6 @@ gr.Interface(
38
  # "state"
39
  ],
40
  outputs=[
41
- "textbox", # "state"
42
  ],
43
  live=True).launch()
 
4
  import whisper
5
 
6
  model = whisper.load_model("base")
7
+ ##Bloom
8
+ API_URL = "https://api-inference.huggingface.co/models/bigscience/bloom"
9
+ HF_TOKEN = os.environ["HF_TOKEN"]
10
+ headers = {"Authorization": f"Bearer {HF_TOKEN}"}
11
+
12
 
13
  def fun(audio) : #, state=''):
14
+ text1 = model.transcribe(audio)["text"]
15
+ text2 = lang_model_response(text)
16
+ return text1, text2
17
+
18
+ def lang_model_response(prompt):
19
+ print(f"*****Inside meme_generate - Prompt is :{prompt}")
20
+ if len(prompt) == 0:
21
+ prompt = """Can you help me please?"""
22
+
23
+ json_ = {"inputs": prompt,
24
+ "parameters":
25
+ {
26
+ "top_p": top_p, #0.90 default
27
+ "max_new_tokens": 64,
28
+ "temperature": temp, #1.1 default
29
+ "return_full_text": True,
30
+ "do_sample": True,
31
+ },
32
+ "options":
33
+ {"use_cache": True,
34
+ "wait_for_model": True,
35
+ },}
36
+ response = requests.post(API_URL, headers=headers, json=json_)
37
+ print(f"Response is : {response}")
38
+ output = response.json()
39
+ print(f"output is : {output}")
40
+ output_tmp = output[0]['generated_text']
41
+ print(f"output_tmp is: {output_tmp}")
42
+ solution = output_tmp[0] #output_tmp.split("\nQ:")[0]
43
+ print(f"Final response after splits is: {solution}")
44
+
45
+ #meme_image, new_prompt = write_on_image(solution)
46
+ return solution
47
 
48
  def fun1(audio, state=''):
49
  text = model.transcribe(audio)["text"]
50
  state += text + " "
51
  return state, state
52
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
53
  gr.Interface(
54
  title = 'Testing Whisper',
55
  fn=fun,
 
58
  # "state"
59
  ],
60
  outputs=[
61
+ "textbox", "textbox"
62
  ],
63
  live=True).launch()