Spaces:
Paused
Paused
added examples, duplicate button, more
Browse filesregenerate and delete last turn buttons added
app.py
CHANGED
@@ -86,7 +86,7 @@ def parse_text(text):
|
|
86 |
return text
|
87 |
|
88 |
|
89 |
-
def predict(input, chatbot, max_length, top_p, temperature, history, past_key_values):
|
90 |
chatbot.append((parse_text(input), ""))
|
91 |
for response, history, past_key_values in model.stream_chat(tokenizer, input, history, past_key_values=past_key_values,
|
92 |
return_past_key_values=True,
|
@@ -130,8 +130,52 @@ def reset_state():
|
|
130 |
return [], [], None
|
131 |
|
132 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
133 |
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
134 |
gr.HTML("""<h1 align="center">ChatGLM2-6B-int4</h1>""")
|
|
|
|
|
135 |
with gr.Accordion("Info", open=False):
|
136 |
_ = """
|
137 |
A query takes from 30 seconds to a few tens of seconds, dependent on the number of words/characters
|
@@ -156,8 +200,11 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
156 |
with gr.Column(scale=12):
|
157 |
user_input = gr.Textbox(show_label=False, placeholder="Input...", lines=10).style(
|
158 |
container=False)
|
|
|
159 |
with gr.Column(min_width=32, scale=1):
|
160 |
submitBtn = gr.Button("Submit", variant="primary")
|
|
|
|
|
161 |
with gr.Column(scale=1):
|
162 |
emptyBtn = gr.Button("Clear History")
|
163 |
max_length = gr.Slider(0, 32768, value=8192/2, step=1.0, label="Maximum length", interactive=True)
|
@@ -167,14 +214,22 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
167 |
history = gr.State([])
|
168 |
past_key_values = gr.State(None)
|
169 |
|
170 |
-
user_input.submit(predict, [user_input, chatbot, max_length, top_p, temperature, history, past_key_values],
|
171 |
[chatbot, history, past_key_values], show_progress=True)
|
172 |
-
submitBtn.click(predict, [user_input, chatbot, max_length, top_p, temperature, history, past_key_values],
|
173 |
[chatbot, history, past_key_values], show_progress=True, api_name="predict")
|
174 |
submitBtn.click(reset_user_input, [], [user_input])
|
175 |
|
176 |
emptyBtn.click(reset_state, outputs=[chatbot, history, past_key_values], show_progress=True)
|
177 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
178 |
with gr.Accordion("For Translation API", open=False):
|
179 |
input_text = gr.Text()
|
180 |
tr_btn = gr.Button("Go", variant="primary")
|
@@ -182,6 +237,19 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
182 |
tr_btn.click(trans_api, [input_text, max_length, top_p, temperature], out_text, show_progress=True, api_name="tr")
|
183 |
input_text.submit(trans_api, [input_text, max_length, top_p, temperature], out_text, show_progress=True, api_name="tr")
|
184 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
185 |
# demo.queue().launch(share=False, inbrowser=True)
|
186 |
# demo.queue().launch(share=True, inbrowser=True, debug=True)
|
187 |
|
|
|
86 |
return text
|
87 |
|
88 |
|
89 |
+
def predict(RETRY_FLAG, input, chatbot, max_length, top_p, temperature, history, past_key_values):
|
90 |
chatbot.append((parse_text(input), ""))
|
91 |
for response, history, past_key_values in model.stream_chat(tokenizer, input, history, past_key_values=past_key_values,
|
92 |
return_past_key_values=True,
|
|
|
130 |
return [], [], None
|
131 |
|
132 |
|
133 |
+
# Delete last turn
|
134 |
+
def delete_last_turn(chat, history):
|
135 |
+
if chat and history:
|
136 |
+
chat.pop(-1)
|
137 |
+
history.pop(-1)
|
138 |
+
history.pop(-1)
|
139 |
+
return chat, history
|
140 |
+
|
141 |
+
|
142 |
+
# Regenerate response
|
143 |
+
def retry_last_answer(
|
144 |
+
user_input,
|
145 |
+
chatbot,
|
146 |
+
max_length,
|
147 |
+
top_p,
|
148 |
+
temperature,
|
149 |
+
history,
|
150 |
+
past_key_values
|
151 |
+
):
|
152 |
+
|
153 |
+
if chat and history:
|
154 |
+
# Removing the previous conversation from chat
|
155 |
+
chat.pop(-1)
|
156 |
+
# Removing bot response from the history
|
157 |
+
history.pop(-1)
|
158 |
+
# Setting up a flag to capture a retry
|
159 |
+
RETRY_FLAG = True
|
160 |
+
# Getting last message from user
|
161 |
+
user_message = history[-1]
|
162 |
+
|
163 |
+
yield from predict(
|
164 |
+
RETRY_FLAG,
|
165 |
+
user_input,
|
166 |
+
chatbot,
|
167 |
+
max_length,
|
168 |
+
top_p,
|
169 |
+
temperature,
|
170 |
+
history,
|
171 |
+
past_key_values
|
172 |
+
)
|
173 |
+
|
174 |
+
|
175 |
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
176 |
gr.HTML("""<h1 align="center">ChatGLM2-6B-int4</h1>""")
|
177 |
+
gr.HTML("""<center><a href="https://huggingface.co/spaces/ysharma/chatglm2-6b-4bit?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>To avoid the queue and for faster inference Duplicate this Space and upgrade to GPU</center>""")
|
178 |
+
|
179 |
with gr.Accordion("Info", open=False):
|
180 |
_ = """
|
181 |
A query takes from 30 seconds to a few tens of seconds, dependent on the number of words/characters
|
|
|
200 |
with gr.Column(scale=12):
|
201 |
user_input = gr.Textbox(show_label=False, placeholder="Input...", lines=10).style(
|
202 |
container=False)
|
203 |
+
RETRY_FLAG = gr.Checkbox(value=False, visible=False)
|
204 |
with gr.Column(min_width=32, scale=1):
|
205 |
submitBtn = gr.Button("Submit", variant="primary")
|
206 |
+
deleteBtn = gr.Button("Delete last turn", variant="secondary")
|
207 |
+
retryBtn = gr.Button("Regenerate", variant="secondary")
|
208 |
with gr.Column(scale=1):
|
209 |
emptyBtn = gr.Button("Clear History")
|
210 |
max_length = gr.Slider(0, 32768, value=8192/2, step=1.0, label="Maximum length", interactive=True)
|
|
|
214 |
history = gr.State([])
|
215 |
past_key_values = gr.State(None)
|
216 |
|
217 |
+
user_input.submit(predict, [RETRY_FLAG, user_input, chatbot, max_length, top_p, temperature, history, past_key_values],
|
218 |
[chatbot, history, past_key_values], show_progress=True)
|
219 |
+
submitBtn.click(predict, [RETRY_FLAG, user_input, chatbot, max_length, top_p, temperature, history, past_key_values],
|
220 |
[chatbot, history, past_key_values], show_progress=True, api_name="predict")
|
221 |
submitBtn.click(reset_user_input, [], [user_input])
|
222 |
|
223 |
emptyBtn.click(reset_state, outputs=[chatbot, history, past_key_values], show_progress=True)
|
224 |
|
225 |
+
regenerate_button.click(
|
226 |
+
retry_last_answer,
|
227 |
+
inputs = [user_input, chatbot, max_length, top_p, temperature, history, past_key_values],
|
228 |
+
#outputs = [chatbot, history, last_user_message, user_message]
|
229 |
+
outputs=[chatbot, history, past_key_values]
|
230 |
+
)
|
231 |
+
delete_turn_button.click(delete_last_turn, [chatbot, history], [chatbot, history])
|
232 |
+
|
233 |
with gr.Accordion("For Translation API", open=False):
|
234 |
input_text = gr.Text()
|
235 |
tr_btn = gr.Button("Go", variant="primary")
|
|
|
237 |
tr_btn.click(trans_api, [input_text, max_length, top_p, temperature], out_text, show_progress=True, api_name="tr")
|
238 |
input_text.submit(trans_api, [input_text, max_length, top_p, temperature], out_text, show_progress=True, api_name="tr")
|
239 |
|
240 |
+
with gr.Accordion("Example inputs", open=True):
|
241 |
+
examples = gr.Examples(
|
242 |
+
examples=[["Explain the plot of Cinderella in a sentence."],
|
243 |
+
["How long does it take to become proficient in French, and what are the best methods for retaining information?"],
|
244 |
+
["What are some common mistakes to avoid when writing code?"],
|
245 |
+
["Build a prompt to generate a beautiful portrait of a horse"],
|
246 |
+
["Suggest four metaphors to describe the benefits of AI"],
|
247 |
+
["Write a pop song about leaving home for the sandy beaches."],
|
248 |
+
["Write a summary demonstrating my ability to do beat-boxing"]],
|
249 |
+
inputs = [user_input],
|
250 |
+
|
251 |
+
)
|
252 |
+
|
253 |
# demo.queue().launch(share=False, inbrowser=True)
|
254 |
# demo.queue().launch(share=True, inbrowser=True, debug=True)
|
255 |
|