Spaces:
Running
Running
Pclanglais
commited on
Commit
•
d6b6a6e
1
Parent(s):
50af2bb
Update app.py
Browse files
app.py
CHANGED
@@ -100,7 +100,19 @@ class StopOnTokens(StoppingCriteria):
|
|
100 |
return False
|
101 |
|
102 |
|
103 |
-
def predict(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
104 |
|
105 |
print(history_transformer_format)
|
106 |
stop = StopOnTokens()
|
@@ -129,8 +141,6 @@ def predict(history_transformer_format):
|
|
129 |
|
130 |
messages = system_prompt + messages
|
131 |
|
132 |
-
print(messages)
|
133 |
-
|
134 |
model_inputs = tokenizer([messages], return_tensors="pt").to("cuda")
|
135 |
streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
|
136 |
generate_kwargs = dict(
|
@@ -145,27 +155,12 @@ def predict(history_transformer_format):
|
|
145 |
t = Thread(target=model.generate, kwargs=generate_kwargs)
|
146 |
t.start()
|
147 |
|
148 |
-
|
149 |
for new_token in streamer:
|
150 |
if new_token != '<':
|
151 |
-
|
152 |
-
yield
|
153 |
-
|
154 |
-
def user(message, history):
|
155 |
-
global source_text
|
156 |
-
global assess_rag
|
157 |
-
#For now, we only query the vector database once, at the start.
|
158 |
-
if len(history) == 0:
|
159 |
-
assess_rag = classification_chatrag(message)
|
160 |
-
if assess_rag:
|
161 |
-
source_text = vector_search(message)
|
162 |
-
else:
|
163 |
-
source_text = "Albert-Tchap n'utilise pas de sources comme votre requête n'a pas l'air d'en recueillir."
|
164 |
-
|
165 |
-
history_transformer_format = history + [[message, ""]]
|
166 |
-
|
167 |
-
print(history_transformer_format)
|
168 |
-
return "", history_transformer_format
|
169 |
|
170 |
# Define the Gradio interface
|
171 |
title = "Tchap"
|
@@ -176,17 +171,9 @@ examples = [
|
|
176 |
0.7 # temperature
|
177 |
]
|
178 |
]
|
179 |
-
|
180 |
with gr.Blocks() as demo:
|
181 |
-
|
182 |
-
msg = gr.Textbox()
|
183 |
-
clear = gr.Button("Clear")
|
184 |
-
|
185 |
-
msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
|
186 |
-
predict, chatbot, chatbot
|
187 |
-
)
|
188 |
-
clear.click(lambda: None, None, chatbot, queue=False)
|
189 |
-
|
190 |
|
191 |
-
|
192 |
-
demo.launch()
|
|
|
100 |
return False
|
101 |
|
102 |
|
103 |
+
def predict(message, history):
|
104 |
+
|
105 |
+
global source_text
|
106 |
+
global assess_rag
|
107 |
+
#For now, we only query the vector database once, at the start.
|
108 |
+
if len(history) == 0:
|
109 |
+
assess_rag = classification_chatrag(message)
|
110 |
+
if assess_rag:
|
111 |
+
source_text = vector_search(message)
|
112 |
+
else:
|
113 |
+
source_text = "Albert-Tchap n'utilise pas de sources comme votre requête n'a pas l'air d'en recueillir."
|
114 |
+
|
115 |
+
history_transformer_format = history + [[message, ""]]
|
116 |
|
117 |
print(history_transformer_format)
|
118 |
stop = StopOnTokens()
|
|
|
141 |
|
142 |
messages = system_prompt + messages
|
143 |
|
|
|
|
|
144 |
model_inputs = tokenizer([messages], return_tensors="pt").to("cuda")
|
145 |
streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)
|
146 |
generate_kwargs = dict(
|
|
|
155 |
t = Thread(target=model.generate, kwargs=generate_kwargs)
|
156 |
t.start()
|
157 |
|
158 |
+
partial_message = ""
|
159 |
for new_token in streamer:
|
160 |
if new_token != '<':
|
161 |
+
partial_message += new_token
|
162 |
+
yield partial_message
|
163 |
+
return messages
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
164 |
|
165 |
# Define the Gradio interface
|
166 |
title = "Tchap"
|
|
|
171 |
0.7 # temperature
|
172 |
]
|
173 |
]
|
174 |
+
demo = gr.Blocks()
|
175 |
with gr.Blocks() as demo:
|
176 |
+
gr.ChatInterface(predict)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
177 |
|
178 |
+
if __name__ == "__main__":
|
179 |
+
demo.queue().launch()
|