Update app.py
Browse files
app.py
CHANGED
@@ -17,23 +17,12 @@ tts = TTS("tts_models/multilingual/multi-dataset/xtts_v1", gpu=True)
|
|
17 |
|
18 |
title = "Speak with Llama2 70B"
|
19 |
|
20 |
-
DESCRIPTION = """# Speak with Llama2 70B
|
21 |
-
|
22 |
-
This Space demonstrates how to speak to a chatbot, based solely on open-source models.
|
23 |
-
It relies on 3 models:
|
24 |
-
1. [Whisper-large-v2](https://huggingface.co/spaces/sanchit-gandhi/whisper-large-v2) as an ASR model, to transcribe recorded audio to text. It is called through a [gradio client](https://www.gradio.app/docs/client).
|
25 |
-
2. [Llama-2-70b-chat-hf](https://huggingface.co/meta-llama/Llama-2-70b-chat-hf) as the chat model, the actual chat model. It is also called through a [gradio client](https://www.gradio.app/docs/client).
|
26 |
-
3. [Coqui's XTTS](https://huggingface.co/spaces/coqui/xtts) as a TTS model, to generate the chatbot answers. This time, the model is hosted locally.
|
27 |
-
|
28 |
-
|
29 |
-
Note:
|
30 |
-
- As a derivate work of [Llama-2-70b-chat](https://huggingface.co/meta-llama/Llama-2-70b-chat-hf) by Meta,
|
31 |
-
this demo is governed by the original [license](https://huggingface.co/spaces/ysharma/Explore_llamav2_with_TGI/blob/main/LICENSE.txt) and [acceptable use policy](https://huggingface.co/spaces/ysharma/Explore_llamav2_with_TGI/blob/main/USE_POLICY.md).
|
32 |
-
- By using this demo you agree to the terms of the Coqui Public Model License at https://coqui.ai/cpml
|
33 |
-
"""
|
34 |
css = """.toast-wrap { display: none !important } """
|
35 |
|
36 |
|
|
|
|
|
37 |
system_message = "\nYou are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.\n\nIf a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information."
|
38 |
temperature = 0.9
|
39 |
top_p = 0.6
|
@@ -67,85 +56,67 @@ def transcribe(wav_path):
|
|
67 |
|
68 |
|
69 |
def add_text(history, text, agree):
|
70 |
-
if
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
else:
|
75 |
-
gr.Warning("Please accept the Terms & Condition!")
|
76 |
-
return None, gr.update(value="", interactive=True)
|
77 |
|
78 |
def add_file(history, file, agree):
|
79 |
-
if
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
return history
|
87 |
-
else:
|
88 |
-
gr.Warning("Please accept the Terms & Condition!")
|
89 |
-
return None
|
90 |
|
91 |
|
92 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
93 |
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
system_prompt = system_message
|
99 |
-
|
100 |
-
history[-1][1] = ""
|
101 |
-
for character in text_client.submit(
|
102 |
-
history,
|
103 |
-
system_prompt,
|
104 |
-
temperature,
|
105 |
-
4096,
|
106 |
-
temperature,
|
107 |
-
repetition_penalty,
|
108 |
-
api_name="/chat"
|
109 |
-
):
|
110 |
-
history[-1][1] = character
|
111 |
-
yield history
|
112 |
-
else:
|
113 |
-
gr.Warning("Please accept the Terms & Condition!")
|
114 |
-
return None
|
115 |
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
text_to_generate = text_to_generate.replace("\n", " ").strip()
|
120 |
-
text_to_generate = nltk.sent_tokenize(text_to_generate)
|
121 |
-
|
122 |
-
filename = f"{uuid.uuid4()}.wav"
|
123 |
-
sampling_rate = tts.synthesizer.tts_config.audio["sample_rate"]
|
124 |
-
silence = [0] * int(0.25 * sampling_rate)
|
125 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
126 |
|
127 |
-
|
128 |
-
|
129 |
-
wav = tts.tts(text=sentence,
|
130 |
-
speaker_wav="examples/female.wav",
|
131 |
-
decoder_iterations=20,
|
132 |
-
speed=1.2,
|
133 |
-
language="en")
|
134 |
-
|
135 |
-
yield (sampling_rate, np.array(wav)) #np.array(wav + silence))
|
136 |
-
else:
|
137 |
-
gr.Warning("Please accept the Terms & Condition!")
|
138 |
-
return None
|
139 |
|
140 |
|
141 |
with gr.Blocks(title=title) as demo:
|
142 |
gr.Markdown(DESCRIPTION)
|
143 |
|
144 |
-
agree = gr.Checkbox(
|
145 |
-
label="Agree",
|
146 |
-
value=False,
|
147 |
-
info="I agree to the terms of the Coqui Public Model License at https://coqui.ai/cpml",
|
148 |
-
)
|
149 |
|
150 |
chatbot = gr.Chatbot(
|
151 |
[],
|
@@ -169,19 +140,27 @@ with gr.Blocks(title=title) as demo:
|
|
169 |
clear_btn = gr.ClearButton([chatbot, audio])
|
170 |
|
171 |
|
172 |
-
txt_msg = txt.submit(add_text, [chatbot, txt
|
173 |
-
bot,
|
174 |
-
).then(generate_speech,
|
175 |
|
176 |
txt_msg.then(lambda: gr.update(interactive=True), None, [txt], queue=False)
|
177 |
|
178 |
-
file_msg = btn.stop_recording(add_file, [chatbot, btn
|
179 |
-
bot,
|
180 |
-
).then(generate_speech,
|
181 |
|
182 |
|
183 |
-
gr.Markdown("""
|
184 |
-
|
185 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
186 |
demo.queue()
|
187 |
demo.launch(debug=True)
|
|
|
17 |
|
18 |
title = "Speak with Llama2 70B"
|
19 |
|
20 |
+
DESCRIPTION = """# Speak with Llama2 70B"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
21 |
css = """.toast-wrap { display: none !important } """
|
22 |
|
23 |
|
24 |
+
os.environ["GRADIO_TEMP_DIR"] = "/home/yoach/spaces/tmp"
|
25 |
+
|
26 |
system_message = "\nYou are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe. Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.\n\nIf a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information."
|
27 |
temperature = 0.9
|
28 |
top_p = 0.6
|
|
|
56 |
|
57 |
|
58 |
def add_text(history, text, agree):
|
59 |
+
history = [] if history is None else history
|
60 |
+
history = history + [(text, None)]
|
61 |
+
return history, gr.update(value="", interactive=False)
|
62 |
+
|
|
|
|
|
|
|
63 |
|
64 |
def add_file(history, file, agree):
|
65 |
+
history = [] if history is None else history
|
66 |
+
text = transcribe(
|
67 |
+
file
|
68 |
+
)
|
69 |
+
|
70 |
+
history = history + [(text, None)]
|
71 |
+
return history
|
|
|
|
|
|
|
|
|
72 |
|
73 |
|
74 |
+
|
75 |
+
def bot(history, agree, system_prompt=""):
|
76 |
+
history = [] if history is None else history
|
77 |
+
|
78 |
+
if system_prompt == "":
|
79 |
+
system_prompt = system_message
|
80 |
+
|
81 |
+
history[-1][1] = ""
|
82 |
+
for character in text_client.submit(
|
83 |
+
history,
|
84 |
+
system_prompt,
|
85 |
+
temperature,
|
86 |
+
4096,
|
87 |
+
temperature,
|
88 |
+
repetition_penalty,
|
89 |
+
api_name="/chat"
|
90 |
+
):
|
91 |
+
history[-1][1] = character
|
92 |
+
yield history
|
93 |
+
|
94 |
|
95 |
+
def generate_speech(history):
|
96 |
+
text_to_generate = history[-1][1]
|
97 |
+
text_to_generate = text_to_generate.replace("\n", " ").strip()
|
98 |
+
text_to_generate = nltk.sent_tokenize(text_to_generate)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
|
100 |
+
filename = f"{uuid.uuid4()}.wav"
|
101 |
+
sampling_rate = tts.synthesizer.tts_config.audio["sample_rate"]
|
102 |
+
silence = [0] * int(0.25 * sampling_rate)
|
|
|
|
|
|
|
|
|
|
|
|
|
103 |
|
104 |
+
|
105 |
+
for sentence in text_to_generate:
|
106 |
+
# generate speech by cloning a voice using default settings
|
107 |
+
wav = tts.tts(text=sentence,
|
108 |
+
speaker_wav="examples/female.wav",
|
109 |
+
decoder_iterations=20,
|
110 |
+
speed=1.2,
|
111 |
+
language="en")
|
112 |
|
113 |
+
yield (sampling_rate, np.array(wav)) #np.array(wav + silence))
|
114 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
115 |
|
116 |
|
117 |
with gr.Blocks(title=title) as demo:
|
118 |
gr.Markdown(DESCRIPTION)
|
119 |
|
|
|
|
|
|
|
|
|
|
|
120 |
|
121 |
chatbot = gr.Chatbot(
|
122 |
[],
|
|
|
140 |
clear_btn = gr.ClearButton([chatbot, audio])
|
141 |
|
142 |
|
143 |
+
txt_msg = txt.submit(add_text, [chatbot, txt], [chatbot, txt], queue=False).then(
|
144 |
+
bot, chatbot, chatbot
|
145 |
+
).then(generate_speech, chatbot, audio)
|
146 |
|
147 |
txt_msg.then(lambda: gr.update(interactive=True), None, [txt], queue=False)
|
148 |
|
149 |
+
file_msg = btn.stop_recording(add_file, [chatbot, btn], [chatbot], queue=False).then(
|
150 |
+
bot, chatbot, chatbot
|
151 |
+
).then(generate_speech, chatbot, audio)
|
152 |
|
153 |
|
154 |
+
gr.Markdown("""
|
155 |
+
This Space demonstrates how to speak to a chatbot, based solely on open-source models.
|
156 |
+
It relies on 3 models:
|
157 |
+
1. [Whisper-large-v2](https://huggingface.co/spaces/sanchit-gandhi/whisper-large-v2) as an ASR model, to transcribe recorded audio to text. It is called through a [gradio client](https://www.gradio.app/docs/client).
|
158 |
+
2. [Llama-2-70b-chat-hf](https://huggingface.co/meta-llama/Llama-2-70b-chat-hf) as the chat model, the actual chat model. It is also called through a [gradio client](https://www.gradio.app/docs/client).
|
159 |
+
3. [Coqui's XTTS](https://huggingface.co/spaces/coqui/xtts) as a TTS model, to generate the chatbot answers. This time, the model is hosted locally.
|
160 |
+
|
161 |
+
Note:
|
162 |
+
- As a derivate work of [Llama-2-70b-chat](https://huggingface.co/meta-llama/Llama-2-70b-chat-hf) by Meta,
|
163 |
+
this demo is governed by the original [license](https://huggingface.co/spaces/ysharma/Explore_llamav2_with_TGI/blob/main/LICENSE.txt) and [acceptable use policy](https://huggingface.co/spaces/ysharma/Explore_llamav2_with_TGI/blob/main/USE_POLICY.md).
|
164 |
+
- By using this demo you agree to the terms of the Coqui Public Model License at https://coqui.ai/cpml""")
|
165 |
demo.queue()
|
166 |
demo.launch(debug=True)
|