Commit
•
e1419c7
1
Parent(s):
173e7dd
feat/openchat (#5)
Browse files- openchat support (f15edb74a5535fb3057b0bc3a05fce71ee2781b3)
app.py
CHANGED
@@ -4,6 +4,13 @@ import gradio as gr
|
|
4 |
|
5 |
from text_generation import Client, InferenceAPIClient
|
6 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
|
8 |
def get_client(model: str):
|
9 |
if model == "Rallio67/joi2_20B_instruct_alpha":
|
@@ -14,26 +21,30 @@ def get_client(model: str):
|
|
14 |
|
15 |
|
16 |
def get_usernames(model: str):
|
|
|
|
|
|
|
|
|
17 |
if model == "Rallio67/joi2_20B_instruct_alpha":
|
18 |
-
return "User: ", "Joi: "
|
19 |
if model == "togethercomputer/GPT-NeoXT-Chat-Base-20B":
|
20 |
-
return "<human>: ", "<bot>: "
|
21 |
-
return "User: ", "Assistant: "
|
22 |
|
23 |
|
24 |
def predict(
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
):
|
35 |
client = get_client(model)
|
36 |
-
user_name, assistant_name = get_usernames(model)
|
37 |
|
38 |
history.append(inputs)
|
39 |
|
@@ -43,19 +54,20 @@ def predict(
|
|
43 |
|
44 |
if not user_data.startswith(user_name):
|
45 |
user_data = user_name + user_data
|
46 |
-
if not model_data.startswith(
|
47 |
-
model_data =
|
48 |
|
49 |
-
past.append(user_data + model_data +
|
50 |
|
51 |
if not inputs.startswith(user_name):
|
52 |
inputs = user_name + inputs
|
53 |
|
54 |
-
total_inputs = "".join(past) + inputs +
|
55 |
|
56 |
partial_words = ""
|
57 |
|
58 |
-
for i, response in enumerate(
|
|
|
59 |
total_inputs,
|
60 |
top_p=top_p if top_p < 1.0 else None,
|
61 |
top_k=top_k,
|
@@ -65,7 +77,8 @@ def predict(
|
|
65 |
temperature=temperature,
|
66 |
max_new_tokens=500,
|
67 |
stop_sequences=[user_name.rstrip(), assistant_name.rstrip()],
|
68 |
-
|
|
|
69 |
if response.token.special:
|
70 |
continue
|
71 |
|
@@ -81,7 +94,8 @@ def predict(
|
|
81 |
history[-1] = partial_words
|
82 |
|
83 |
chat = [
|
84 |
-
(history[i].strip(), history[i + 1].strip())
|
|
|
85 |
]
|
86 |
yield chat, history
|
87 |
|
@@ -90,6 +104,26 @@ def reset_textbox():
|
|
90 |
return gr.update(value="")
|
91 |
|
92 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
93 |
title = """<h1 align="center">🔥Large Language Model API 🚀Streaming🚀</h1>"""
|
94 |
description = """Language models can be conditioned to act like dialogue agents through a conversational prompt that typically takes the form:
|
95 |
|
@@ -104,17 +138,21 @@ Assistant: <utterance>
|
|
104 |
In this app, you can explore the outputs of multiple LLMs when prompted in this way.
|
105 |
"""
|
106 |
|
|
|
|
|
|
|
|
|
107 |
with gr.Blocks(
|
108 |
-
|
109 |
#chatbot {height: 520px; overflow: auto;}"""
|
110 |
) as demo:
|
111 |
gr.HTML(title)
|
112 |
with gr.Column(elem_id="col_container"):
|
113 |
model = gr.Radio(
|
114 |
-
value="
|
115 |
choices=[
|
|
|
116 |
"Rallio67/joi2_20B_instruct_alpha",
|
117 |
-
# "togethercomputer/GPT-NeoXT-Chat-Base-20B",
|
118 |
"google/flan-t5-xxl",
|
119 |
"google/flan-ul2",
|
120 |
"bigscience/bloom",
|
@@ -124,10 +162,12 @@ with gr.Blocks(
|
|
124 |
label="Model",
|
125 |
interactive=True,
|
126 |
)
|
|
|
127 |
chatbot = gr.Chatbot(elem_id="chatbot")
|
128 |
inputs = gr.Textbox(
|
129 |
placeholder="Hi there!", label="Type an input and press Enter"
|
130 |
)
|
|
|
131 |
state = gr.State([])
|
132 |
b1 = gr.Button()
|
133 |
|
@@ -135,7 +175,7 @@ with gr.Blocks(
|
|
135 |
top_p = gr.Slider(
|
136 |
minimum=-0,
|
137 |
maximum=1.0,
|
138 |
-
value=0.
|
139 |
step=0.05,
|
140 |
interactive=True,
|
141 |
label="Top-p (nucleus sampling)",
|
@@ -143,7 +183,7 @@ with gr.Blocks(
|
|
143 |
temperature = gr.Slider(
|
144 |
minimum=-0,
|
145 |
maximum=5.0,
|
146 |
-
value=0.
|
147 |
step=0.1,
|
148 |
interactive=True,
|
149 |
label="Temperature",
|
@@ -151,7 +191,7 @@ with gr.Blocks(
|
|
151 |
top_k = gr.Slider(
|
152 |
minimum=1,
|
153 |
maximum=50,
|
154 |
-
value=
|
155 |
step=1,
|
156 |
interactive=True,
|
157 |
label="Top-k",
|
@@ -159,12 +199,20 @@ with gr.Blocks(
|
|
159 |
repetition_penalty = gr.Slider(
|
160 |
minimum=0.1,
|
161 |
maximum=3.0,
|
162 |
-
value=1.
|
163 |
step=0.01,
|
164 |
interactive=True,
|
165 |
label="Repetition Penalty",
|
166 |
)
|
167 |
-
watermark = gr.Checkbox(value=
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
168 |
|
169 |
inputs.submit(
|
170 |
predict,
|
|
|
4 |
|
5 |
from text_generation import Client, InferenceAPIClient
|
6 |
|
7 |
+
openchat_preprompt = (
|
8 |
+
"\n<human>: Hi!\n<bot>: My name is Bot, model version is 0.15, part of an open-source kit for "
|
9 |
+
"fine-tuning new bots! I was created by Together, LAION, and Ontocord.ai and the open-source "
|
10 |
+
"community. I am not human, not evil and not alive, and thus have no thoughts and feelings, "
|
11 |
+
"but I am programmed to be helpful, polite, honest, and friendly.\n"
|
12 |
+
)
|
13 |
+
|
14 |
|
15 |
def get_client(model: str):
|
16 |
if model == "Rallio67/joi2_20B_instruct_alpha":
|
|
|
21 |
|
22 |
|
23 |
def get_usernames(model: str):
|
24 |
+
"""
|
25 |
+
Returns:
|
26 |
+
(str, str, str, str): pre-prompt, username, bot name, separator
|
27 |
+
"""
|
28 |
if model == "Rallio67/joi2_20B_instruct_alpha":
|
29 |
+
return "", "User: ", "Joi: ", "\n\n"
|
30 |
if model == "togethercomputer/GPT-NeoXT-Chat-Base-20B":
|
31 |
+
return openchat_preprompt, "<human>: ", "<bot>: ", "\n"
|
32 |
+
return "", "User: ", "Assistant: ", "\n"
|
33 |
|
34 |
|
35 |
def predict(
|
36 |
+
model: str,
|
37 |
+
inputs: str,
|
38 |
+
top_p: float,
|
39 |
+
temperature: float,
|
40 |
+
top_k: int,
|
41 |
+
repetition_penalty: float,
|
42 |
+
watermark: bool,
|
43 |
+
chatbot,
|
44 |
+
history,
|
45 |
):
|
46 |
client = get_client(model)
|
47 |
+
preprompt, user_name, assistant_name, sep = get_usernames(model)
|
48 |
|
49 |
history.append(inputs)
|
50 |
|
|
|
54 |
|
55 |
if not user_data.startswith(user_name):
|
56 |
user_data = user_name + user_data
|
57 |
+
if not model_data.startswith(sep + assistant_name):
|
58 |
+
model_data = sep + assistant_name + model_data
|
59 |
|
60 |
+
past.append(user_data + model_data.rstrip() + sep)
|
61 |
|
62 |
if not inputs.startswith(user_name):
|
63 |
inputs = user_name + inputs
|
64 |
|
65 |
+
total_inputs = preprompt + "".join(past) + inputs + sep + assistant_name.rstrip()
|
66 |
|
67 |
partial_words = ""
|
68 |
|
69 |
+
for i, response in enumerate(
|
70 |
+
client.generate_stream(
|
71 |
total_inputs,
|
72 |
top_p=top_p if top_p < 1.0 else None,
|
73 |
top_k=top_k,
|
|
|
77 |
temperature=temperature,
|
78 |
max_new_tokens=500,
|
79 |
stop_sequences=[user_name.rstrip(), assistant_name.rstrip()],
|
80 |
+
)
|
81 |
+
):
|
82 |
if response.token.special:
|
83 |
continue
|
84 |
|
|
|
94 |
history[-1] = partial_words
|
95 |
|
96 |
chat = [
|
97 |
+
(history[i].strip(), history[i + 1].strip())
|
98 |
+
for i in range(0, len(history) - 1, 2)
|
99 |
]
|
100 |
yield chat, history
|
101 |
|
|
|
104 |
return gr.update(value="")
|
105 |
|
106 |
|
107 |
+
def radio_on_change(
|
108 |
+
value: str, disclaimer, top_p, top_k, temperature, repetition_penalty, watermark
|
109 |
+
):
|
110 |
+
if value == "togethercomputer/GPT-NeoXT-Chat-Base-20B":
|
111 |
+
top_p = top_p.update(value=0.25)
|
112 |
+
top_k = top_k.update(value=50)
|
113 |
+
temperature = temperature.update(value=0.6)
|
114 |
+
repetition_penalty = repetition_penalty.update(value=1.01)
|
115 |
+
watermark = watermark.update(False)
|
116 |
+
disclaimer = disclaimer.update(visible=True)
|
117 |
+
else:
|
118 |
+
top_p = top_p.update(value=0.95)
|
119 |
+
top_k = top_k.update(value=4)
|
120 |
+
temperature = temperature.update(value=0.5)
|
121 |
+
repetition_penalty = repetition_penalty.update(value=1.03)
|
122 |
+
watermark = watermark.update(True)
|
123 |
+
disclaimer = disclaimer.update(visible=False)
|
124 |
+
return disclaimer, top_p, top_k, temperature, repetition_penalty, watermark
|
125 |
+
|
126 |
+
|
127 |
title = """<h1 align="center">🔥Large Language Model API 🚀Streaming🚀</h1>"""
|
128 |
description = """Language models can be conditioned to act like dialogue agents through a conversational prompt that typically takes the form:
|
129 |
|
|
|
138 |
In this app, you can explore the outputs of multiple LLMs when prompted in this way.
|
139 |
"""
|
140 |
|
141 |
+
openchat_disclaimer = """
|
142 |
+
<div align="center">Checkout the official <a href=https://huggingface.co/spaces/togethercomputer/OpenChatKit>OpenChatKit feedback app</a> for the full experience.</div>
|
143 |
+
"""
|
144 |
+
|
145 |
with gr.Blocks(
|
146 |
+
css="""#col_container {margin-left: auto; margin-right: auto;}
|
147 |
#chatbot {height: 520px; overflow: auto;}"""
|
148 |
) as demo:
|
149 |
gr.HTML(title)
|
150 |
with gr.Column(elem_id="col_container"):
|
151 |
model = gr.Radio(
|
152 |
+
value="togethercomputer/GPT-NeoXT-Chat-Base-20B",
|
153 |
choices=[
|
154 |
+
"togethercomputer/GPT-NeoXT-Chat-Base-20B",
|
155 |
"Rallio67/joi2_20B_instruct_alpha",
|
|
|
156 |
"google/flan-t5-xxl",
|
157 |
"google/flan-ul2",
|
158 |
"bigscience/bloom",
|
|
|
162 |
label="Model",
|
163 |
interactive=True,
|
164 |
)
|
165 |
+
|
166 |
chatbot = gr.Chatbot(elem_id="chatbot")
|
167 |
inputs = gr.Textbox(
|
168 |
placeholder="Hi there!", label="Type an input and press Enter"
|
169 |
)
|
170 |
+
disclaimer = gr.Markdown(openchat_disclaimer)
|
171 |
state = gr.State([])
|
172 |
b1 = gr.Button()
|
173 |
|
|
|
175 |
top_p = gr.Slider(
|
176 |
minimum=-0,
|
177 |
maximum=1.0,
|
178 |
+
value=0.25,
|
179 |
step=0.05,
|
180 |
interactive=True,
|
181 |
label="Top-p (nucleus sampling)",
|
|
|
183 |
temperature = gr.Slider(
|
184 |
minimum=-0,
|
185 |
maximum=5.0,
|
186 |
+
value=0.6,
|
187 |
step=0.1,
|
188 |
interactive=True,
|
189 |
label="Temperature",
|
|
|
191 |
top_k = gr.Slider(
|
192 |
minimum=1,
|
193 |
maximum=50,
|
194 |
+
value=50,
|
195 |
step=1,
|
196 |
interactive=True,
|
197 |
label="Top-k",
|
|
|
199 |
repetition_penalty = gr.Slider(
|
200 |
minimum=0.1,
|
201 |
maximum=3.0,
|
202 |
+
value=1.01,
|
203 |
step=0.01,
|
204 |
interactive=True,
|
205 |
label="Repetition Penalty",
|
206 |
)
|
207 |
+
watermark = gr.Checkbox(value=False, label="Text watermarking")
|
208 |
+
|
209 |
+
model.change(
|
210 |
+
lambda value: radio_on_change(
|
211 |
+
value, disclaimer, top_p, top_k, temperature, repetition_penalty, watermark
|
212 |
+
),
|
213 |
+
inputs=model,
|
214 |
+
outputs=[disclaimer, top_p, top_k, temperature, repetition_penalty, watermark],
|
215 |
+
)
|
216 |
|
217 |
inputs.submit(
|
218 |
predict,
|