Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -48,14 +48,23 @@ tok = Tokenizer.from_file(tokenizer_file)
|
|
48 |
# ----------------------------
|
49 |
# π¬ Gradio Chat Wrapper
|
50 |
# ----------------------------
|
51 |
-
def beeper_reply(message, history, temperature, top_k, top_p):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
52 |
# Build conversation context
|
53 |
prompt_parts = []
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
|
|
59 |
|
60 |
# Add current message
|
61 |
prompt_parts.append(f"User: {message}")
|
@@ -70,9 +79,9 @@ def beeper_reply(message, history, temperature, top_k, top_p):
|
|
70 |
cfg=config,
|
71 |
prompt=prompt,
|
72 |
max_new_tokens=128,
|
73 |
-
temperature=temperature,
|
74 |
top_k=int(top_k),
|
75 |
-
top_p=top_p,
|
76 |
repetition_penalty=config["repetition_penalty"],
|
77 |
presence_penalty=config["presence_penalty"],
|
78 |
frequency_penalty=config["frequency_penalty"],
|
@@ -96,7 +105,7 @@ demo = gr.ChatInterface(
|
|
96 |
gr.Slider(1, 100, value=40, step=1, label="Top-k"),
|
97 |
gr.Slider(0.1, 1.0, value=0.9, step=0.05, label="Top-p"),
|
98 |
],
|
99 |
-
chatbot=gr.Chatbot(label="Chat with Beeper π€"),
|
100 |
title="Beeper - A Rose-based Tiny Language Model",
|
101 |
description="Hello! I'm Beeper, a small language model trained with love and care. Please be patient with me - I'm still learning! π",
|
102 |
examples=[
|
@@ -105,6 +114,7 @@ demo = gr.ChatInterface(
|
|
105 |
["What do you like to do for fun?"],
|
106 |
],
|
107 |
theme=gr.themes.Soft(),
|
|
|
108 |
)
|
109 |
|
110 |
if __name__ == "__main__":
|
|
|
48 |
# ----------------------------
|
49 |
# π¬ Gradio Chat Wrapper
|
50 |
# ----------------------------
|
51 |
+
def beeper_reply(message, history, temperature=None, top_k=None, top_p=None):
|
52 |
+
# Use defaults if not provided (for examples caching)
|
53 |
+
if temperature is None:
|
54 |
+
temperature = 0.9
|
55 |
+
if top_k is None:
|
56 |
+
top_k = 40
|
57 |
+
if top_p is None:
|
58 |
+
top_p = 0.9
|
59 |
+
|
60 |
# Build conversation context
|
61 |
prompt_parts = []
|
62 |
+
if history:
|
63 |
+
for h in history:
|
64 |
+
if h[0]: # User message exists
|
65 |
+
prompt_parts.append(f"User: {h[0]}")
|
66 |
+
if h[1]: # Assistant response exists
|
67 |
+
prompt_parts.append(f"Beeper: {h[1]}")
|
68 |
|
69 |
# Add current message
|
70 |
prompt_parts.append(f"User: {message}")
|
|
|
79 |
cfg=config,
|
80 |
prompt=prompt,
|
81 |
max_new_tokens=128,
|
82 |
+
temperature=float(temperature),
|
83 |
top_k=int(top_k),
|
84 |
+
top_p=float(top_p),
|
85 |
repetition_penalty=config["repetition_penalty"],
|
86 |
presence_penalty=config["presence_penalty"],
|
87 |
frequency_penalty=config["frequency_penalty"],
|
|
|
105 |
gr.Slider(1, 100, value=40, step=1, label="Top-k"),
|
106 |
gr.Slider(0.1, 1.0, value=0.9, step=0.05, label="Top-p"),
|
107 |
],
|
108 |
+
chatbot=gr.Chatbot(label="Chat with Beeper π€", type="messages"),
|
109 |
title="Beeper - A Rose-based Tiny Language Model",
|
110 |
description="Hello! I'm Beeper, a small language model trained with love and care. Please be patient with me - I'm still learning! π",
|
111 |
examples=[
|
|
|
114 |
["What do you like to do for fun?"],
|
115 |
],
|
116 |
theme=gr.themes.Soft(),
|
117 |
+
cache_examples=False, # Disable caching to avoid the startup issue
|
118 |
)
|
119 |
|
120 |
if __name__ == "__main__":
|