Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -8,6 +8,22 @@ import edge_tts
|
|
8 |
import asyncio
|
9 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
DESCRIPTION = """
|
12 |
# QwQ Edge 💬
|
13 |
"""
|
@@ -26,25 +42,21 @@ h1 {
|
|
26 |
}
|
27 |
'''
|
28 |
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
torch_dtype=torch.bfloat16,
|
41 |
-
)
|
42 |
-
model.eval()
|
43 |
|
44 |
|
45 |
-
async def text_to_speech(text: str, output_file="output.mp3"):
|
46 |
"""Convert text to speech using Edge TTS and save as MP3"""
|
47 |
-
voice = "en-US-GuyNeural" # Change this to your preferred voice
|
48 |
communicate = edge_tts.Communicate(text, voice)
|
49 |
await communicate.save(output_file)
|
50 |
return output_file
|
@@ -62,7 +74,24 @@ def generate(
|
|
62 |
):
|
63 |
"""Generates chatbot response and handles TTS requests"""
|
64 |
is_tts = message.strip().lower().startswith("@tts")
|
65 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
|
67 |
conversation = [*chat_history, {"role": "user", "content": message}]
|
68 |
|
@@ -95,7 +124,8 @@ def generate(
|
|
95 |
final_response = "".join(outputs)
|
96 |
|
97 |
if is_tts:
|
98 |
-
|
|
|
99 |
yield gr.Audio(output_file, autoplay=True) # Return playable audio
|
100 |
else:
|
101 |
yield final_response # Return text response
|
@@ -112,12 +142,12 @@ demo = gr.ChatInterface(
|
|
112 |
],
|
113 |
stop_btn=None,
|
114 |
examples=[
|
115 |
-
["@
|
116 |
-
["A train travels 60 kilometers per hour. If it travels for 5 hours, how far will it travel in total?"],
|
117 |
["Write a Python function to check if a number is prime."],
|
118 |
-
["@
|
119 |
["Rewrite the following sentence in passive voice: 'The dog chased the cat.'"],
|
120 |
-
["@
|
121 |
],
|
122 |
cache_examples=False,
|
123 |
type="messages",
|
|
|
8 |
import asyncio
|
9 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
10 |
|
11 |
+
MAX_MAX_NEW_TOKENS = 2048
|
12 |
+
DEFAULT_MAX_NEW_TOKENS = 1024
|
13 |
+
MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
|
14 |
+
|
15 |
+
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
16 |
+
|
17 |
+
model_id = "prithivMLmods/FastThink-0.5B-Tiny"
|
18 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
19 |
+
model = AutoModelForCausalLM.from_pretrained(
|
20 |
+
model_id,
|
21 |
+
device_map="auto",
|
22 |
+
torch_dtype=torch.bfloat16,
|
23 |
+
)
|
24 |
+
model.eval()
|
25 |
+
|
26 |
+
|
27 |
DESCRIPTION = """
|
28 |
# QwQ Edge 💬
|
29 |
"""
|
|
|
42 |
}
|
43 |
'''
|
44 |
|
45 |
+
# List of voices
|
46 |
+
voices = [
|
47 |
+
"en-US-JennyNeural", # @tts1
|
48 |
+
"en-US-GuyNeural", # @tts2
|
49 |
+
"en-US-AriaNeural", # @tts3
|
50 |
+
"en-US-DavisNeural", # @tts4
|
51 |
+
"en-US-JaneNeural", # @tts5
|
52 |
+
"en-US-JasonNeural", # @tts6
|
53 |
+
"en-US-NancyNeural", # @tts7
|
54 |
+
"en-US-TonyNeural", # @tts8
|
55 |
+
]
|
|
|
|
|
|
|
56 |
|
57 |
|
58 |
+
async def text_to_speech(text: str, voice: str, output_file="output.mp3"):
|
59 |
"""Convert text to speech using Edge TTS and save as MP3"""
|
|
|
60 |
communicate = edge_tts.Communicate(text, voice)
|
61 |
await communicate.save(output_file)
|
62 |
return output_file
|
|
|
74 |
):
|
75 |
"""Generates chatbot response and handles TTS requests"""
|
76 |
is_tts = message.strip().lower().startswith("@tts")
|
77 |
+
tts_index = None
|
78 |
+
|
79 |
+
if is_tts:
|
80 |
+
# Extract the number after @tts
|
81 |
+
tts_part = message.strip().lower().split()[0] # Get the @ttsX part
|
82 |
+
if len(tts_part) > 4: # Check if it's @ttsX (e.g., @tts1, @tts2, etc.)
|
83 |
+
try:
|
84 |
+
tts_index = int(tts_part[4:]) - 1 # Convert to 0-based index
|
85 |
+
if tts_index < 0 or tts_index >= len(voices):
|
86 |
+
gr.Warning(f"Invalid TTS voice index. Using default voice.")
|
87 |
+
tts_index = 0
|
88 |
+
except ValueError:
|
89 |
+
gr.Warning(f"Invalid TTS voice index. Using default voice.")
|
90 |
+
tts_index = 0
|
91 |
+
else:
|
92 |
+
tts_index = 0 # Default to the first voice if no number is provided
|
93 |
+
|
94 |
+
message = message.replace(tts_part, "").strip() # Remove @ttsX from the message
|
95 |
|
96 |
conversation = [*chat_history, {"role": "user", "content": message}]
|
97 |
|
|
|
124 |
final_response = "".join(outputs)
|
125 |
|
126 |
if is_tts:
|
127 |
+
voice = voices[tts_index] # Select the voice based on the index
|
128 |
+
output_file = asyncio.run(text_to_speech(final_response, voice))
|
129 |
yield gr.Audio(output_file, autoplay=True) # Return playable audio
|
130 |
else:
|
131 |
yield final_response # Return text response
|
|
|
142 |
],
|
143 |
stop_btn=None,
|
144 |
examples=[
|
145 |
+
["@tts1 Who is Nikola Tesla, and why did he die?"],
|
146 |
+
["@tts2 A train travels 60 kilometers per hour. If it travels for 5 hours, how far will it travel in total?"],
|
147 |
["Write a Python function to check if a number is prime."],
|
148 |
+
["@tts3 What causes rainbows to form?"],
|
149 |
["Rewrite the following sentence in passive voice: 'The dog chased the cat.'"],
|
150 |
+
["@tts4 What is the capital of France?"],
|
151 |
],
|
152 |
cache_examples=False,
|
153 |
type="messages",
|