Spaces:
Sleeping
Sleeping
Add stream kwarg; add dict history support
Browse files
app.py
CHANGED
|
@@ -14,31 +14,43 @@ def respond(
|
|
| 14 |
max_tokens,
|
| 15 |
temperature,
|
| 16 |
top_p,
|
|
|
|
| 17 |
):
|
| 18 |
-
messages = [
|
| 19 |
-
|
|
|
|
| 20 |
for val in history:
|
| 21 |
-
if val
|
| 22 |
-
messages.append(
|
| 23 |
-
|
| 24 |
-
messages.append(
|
| 25 |
-
|
|
|
|
|
|
|
|
|
|
| 26 |
messages.append({"role": "user", "content": message})
|
| 27 |
|
| 28 |
response = ""
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
|
| 43 |
"""
|
| 44 |
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
|
|
@@ -56,6 +68,7 @@ demo = gr.ChatInterface(
|
|
| 56 |
step=0.05,
|
| 57 |
label="Top-p (nucleus sampling)",
|
| 58 |
),
|
|
|
|
| 59 |
],
|
| 60 |
)
|
| 61 |
|
|
|
|
| 14 |
max_tokens,
|
| 15 |
temperature,
|
| 16 |
top_p,
|
| 17 |
+
stream: bool = True,
|
| 18 |
):
|
| 19 |
+
messages = []
|
| 20 |
+
if system_message:
|
| 21 |
+
messages.append({"role": "system", "content": system_message})
|
| 22 |
for val in history:
|
| 23 |
+
if isinstance(val, dict) and "role" in val and "content" in val:
|
| 24 |
+
messages.append(val)
|
| 25 |
+
elif isinstance(val, (tuple, list)):
|
| 26 |
+
messages.append(
|
| 27 |
+
{"role": "user", "content": val[0]}
|
| 28 |
+
if val[0] else
|
| 29 |
+
{"role": "assistant", "content": val[1]}
|
| 30 |
+
)
|
| 31 |
messages.append({"role": "user", "content": message})
|
| 32 |
|
| 33 |
response = ""
|
| 34 |
+
if stream:
|
| 35 |
+
for message in client.chat_completion(
|
| 36 |
+
messages,
|
| 37 |
+
max_tokens=max_tokens,
|
| 38 |
+
stream=True,
|
| 39 |
+
temperature=temperature,
|
| 40 |
+
top_p=top_p,
|
| 41 |
+
):
|
| 42 |
+
token = message.choices[0].delta.content
|
| 43 |
+
|
| 44 |
+
response += token
|
| 45 |
+
yield response
|
| 46 |
+
else:
|
| 47 |
+
completion = client.chat_completion(
|
| 48 |
+
messages,
|
| 49 |
+
max_tokens=max_tokens,
|
| 50 |
+
temperature=temperature,
|
| 51 |
+
top_p=top_p,
|
| 52 |
+
)
|
| 53 |
+
return completion.choices[0].message
|
| 54 |
|
| 55 |
"""
|
| 56 |
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
|
|
|
|
| 68 |
step=0.05,
|
| 69 |
label="Top-p (nucleus sampling)",
|
| 70 |
),
|
| 71 |
+
gr.Checkbox(value=True, label="Streaming", info="Streaming response vs full completion"),
|
| 72 |
],
|
| 73 |
)
|
| 74 |
|