List of models ? - What wind.surf will do ?
Browse files
app.py
CHANGED
@@ -23,20 +23,27 @@ def load_model(model_name):
|
|
23 |
current_model = Llama.from_pretrained(
|
24 |
repo_id=model_info["repo_id"],
|
25 |
filename=model_info["filename"],
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
return current_model
|
32 |
|
33 |
# Initialize with first model
|
34 |
current_model = load_model(list(MODELS.keys())[0])
|
35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
36 |
def respond(
|
37 |
message,
|
38 |
history: list[tuple[str, str]],
|
39 |
model_name,
|
|
|
40 |
system_message,
|
41 |
max_tokens,
|
42 |
temperature,
|
@@ -47,7 +54,7 @@ def respond(
|
|
47 |
# Load new model if changed
|
48 |
if current_model is None or model_name != current_model.model_path:
|
49 |
current_model = load_model(model_name)
|
50 |
-
|
51 |
messages = [{"role": "system", "content": system_message}]
|
52 |
|
53 |
for val in history:
|
@@ -79,13 +86,21 @@ For information on how to customize the ChatInterface, peruse the gradio docs: h
|
|
79 |
demo = gr.ChatInterface(
|
80 |
respond,
|
81 |
title="GGUF is popular format on PC in LM Studio or on Tablet/Mobile in PocketPal APPs",
|
82 |
-
|
83 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
84 |
additional_inputs=[
|
85 |
gr.Dropdown(
|
86 |
choices=list(MODELS.keys()),
|
87 |
value=list(MODELS.keys())[0],
|
88 |
-
label="Select Model"
|
|
|
|
|
89 |
),
|
90 |
gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
|
91 |
gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
|
@@ -146,4 +161,9 @@ demo = gr.ChatInterface(
|
|
146 |
|
147 |
|
148 |
if __name__ == "__main__":
|
149 |
-
|
|
|
|
|
|
|
|
|
|
|
|
23 |
current_model = Llama.from_pretrained(
|
24 |
repo_id=model_info["repo_id"],
|
25 |
filename=model_info["filename"],
|
26 |
+
verbose=True,
|
27 |
+
n_ctx=32768,
|
28 |
+
n_threads=2,
|
29 |
+
chat_format="chatml"
|
30 |
+
)
|
31 |
return current_model
|
32 |
|
33 |
# Initialize with first model
|
34 |
current_model = load_model(list(MODELS.keys())[0])
|
35 |
|
36 |
+
def get_description(model_name):
|
37 |
+
return f"🤖 Current Model: {model_name}\n\nTry models locclay in: 🖥️ [LM Studio AI for PC](https://lmstudio.ai) | 📱 PocketPal AI ([Android](https://play.google.com/store/apps/details?id=com.pocketpalai) & [iOS](https://play.google.com/store/apps/details?id=com.pocketpalai)) on Tablet or Mobile"
|
38 |
+
|
39 |
+
def update_description(model_name):
|
40 |
+
return get_description(model_name)
|
41 |
+
|
42 |
def respond(
|
43 |
message,
|
44 |
history: list[tuple[str, str]],
|
45 |
model_name,
|
46 |
+
description_text,
|
47 |
system_message,
|
48 |
max_tokens,
|
49 |
temperature,
|
|
|
54 |
# Load new model if changed
|
55 |
if current_model is None or model_name != current_model.model_path:
|
56 |
current_model = load_model(model_name)
|
57 |
+
|
58 |
messages = [{"role": "system", "content": system_message}]
|
59 |
|
60 |
for val in history:
|
|
|
86 |
demo = gr.ChatInterface(
|
87 |
respond,
|
88 |
title="GGUF is popular format on PC in LM Studio or on Tablet/Mobile in PocketPal APPs",
|
89 |
+
|
90 |
+
gr.Textbox(
|
91 |
+
value=get_description(list(MODELS.keys())[0]),
|
92 |
+
label="About",
|
93 |
+
interactive=False,
|
94 |
+
lines=4
|
95 |
+
),
|
96 |
+
|
97 |
additional_inputs=[
|
98 |
gr.Dropdown(
|
99 |
choices=list(MODELS.keys()),
|
100 |
value=list(MODELS.keys())[0],
|
101 |
+
label="Select Model",
|
102 |
+
interactive=False,
|
103 |
+
allow_custom_value=False
|
104 |
),
|
105 |
gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
|
106 |
gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
|
|
|
161 |
|
162 |
|
163 |
if __name__ == "__main__":
|
164 |
+
with gr.Blocks() as blocks:
|
165 |
+
chatbot = demo.render()
|
166 |
+
model_dropdown = [comp for comp in chatbot.children if isinstance(comp, gr.Dropdown)][0]
|
167 |
+
description_box = [comp for comp in chatbot.children if isinstance(comp, gr.Textbox)][0]
|
168 |
+
model_dropdown.change(fn=update_description, inputs=[model_dropdown], outputs=[description_box])
|
169 |
+
blocks.launch()
|