Spaces:
Runtime error
Runtime error
update app.py
Browse filesThis setup can be extremely powerful and potentially dangerous depending on its use. With high flexibility in text generation and minimal constraints, it can generate misleading, inappropriate, or harmful content if not used responsibly. The AI's capabilities to create highly varied and convincing outputs mean that misuse could lead to spreading misinformation, generating offensive material, or other unethical uses. Always exercise caution and ethical consideration when deploying such powerful tools. 🚨
app.py
CHANGED
@@ -1,22 +1,17 @@
|
|
1 |
from huggingface_hub import InferenceClient
|
2 |
import gradio as gr
|
3 |
|
4 |
-
client = InferenceClient(
|
5 |
-
"mistralai/Mistral-7B-Instruct-v0.3"
|
6 |
-
)
|
7 |
-
|
8 |
|
9 |
def format_prompt(message, history):
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
|
17 |
-
def generate(
|
18 |
-
prompt, history, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,
|
19 |
-
):
|
20 |
temperature = float(temperature)
|
21 |
if temperature < 1e-2:
|
22 |
temperature = 1e-2
|
@@ -41,55 +36,51 @@ def generate(
|
|
41 |
yield output
|
42 |
return output
|
43 |
|
44 |
-
|
45 |
additional_inputs=[
|
46 |
gr.Slider(
|
47 |
label="Temperature",
|
48 |
-
value=0
|
49 |
minimum=0.0,
|
50 |
maximum=1.0,
|
51 |
-
step=0.
|
52 |
interactive=True,
|
53 |
info="Higher values produce more diverse outputs",
|
54 |
),
|
55 |
gr.Slider(
|
56 |
label="Max new tokens",
|
57 |
-
value=
|
58 |
minimum=0,
|
59 |
-
maximum=
|
60 |
-
step=
|
61 |
interactive=True,
|
62 |
info="The maximum numbers of new tokens",
|
63 |
),
|
64 |
gr.Slider(
|
65 |
label="Top-p (nucleus sampling)",
|
66 |
-
value=0
|
67 |
minimum=0.0,
|
68 |
-
maximum=1,
|
69 |
-
step=0.
|
70 |
interactive=True,
|
71 |
info="Higher values sample more low-probability tokens",
|
72 |
),
|
73 |
gr.Slider(
|
74 |
label="Repetition penalty",
|
75 |
-
value=1.
|
76 |
minimum=1.0,
|
77 |
maximum=2.0,
|
78 |
-
step=0.
|
79 |
interactive=True,
|
80 |
info="Penalize repeated tokens",
|
81 |
)
|
82 |
]
|
83 |
|
84 |
-
|
85 |
gr.ChatInterface(
|
86 |
fn=generate,
|
87 |
chatbot=gr.Chatbot(show_label=False, show_share_button=False, show_copy_button=True, likeable=True, layout="panel"),
|
88 |
additional_inputs=additional_inputs,
|
89 |
-
title="
|
90 |
).launch(show_api=False)
|
91 |
|
92 |
-
|
93 |
gr.load("models/ehristoforu/dalle-3-xl-v2").launch()
|
94 |
-
|
95 |
-
gr.load("models/microsoft/Phi-3-mini-4k-instruct").launch()
|
|
|
1 |
from huggingface_hub import InferenceClient
|
2 |
import gradio as gr
|
3 |
|
4 |
+
client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.3")
|
|
|
|
|
|
|
5 |
|
6 |
def format_prompt(message, history):
|
7 |
+
prompt = "<s>"
|
8 |
+
for user_prompt, bot_response in history:
|
9 |
+
prompt += f"[INST] {user_prompt} [/INST]"
|
10 |
+
prompt += f" {bot_response}</s> "
|
11 |
+
prompt += f"[INST] {message} [/INST]"
|
12 |
+
return prompt
|
13 |
|
14 |
+
def generate(prompt, history, temperature=1.0, max_new_tokens=1048, top_p=1.0, repetition_penalty=1.0):
|
|
|
|
|
15 |
temperature = float(temperature)
|
16 |
if temperature < 1e-2:
|
17 |
temperature = 1e-2
|
|
|
36 |
yield output
|
37 |
return output
|
38 |
|
|
|
39 |
additional_inputs=[
|
40 |
gr.Slider(
|
41 |
label="Temperature",
|
42 |
+
value=1.0,
|
43 |
minimum=0.0,
|
44 |
maximum=1.0,
|
45 |
+
step=0.01,
|
46 |
interactive=True,
|
47 |
info="Higher values produce more diverse outputs",
|
48 |
),
|
49 |
gr.Slider(
|
50 |
label="Max new tokens",
|
51 |
+
value=1048,
|
52 |
minimum=0,
|
53 |
+
maximum=2048,
|
54 |
+
step=128,
|
55 |
interactive=True,
|
56 |
info="The maximum numbers of new tokens",
|
57 |
),
|
58 |
gr.Slider(
|
59 |
label="Top-p (nucleus sampling)",
|
60 |
+
value=1.0,
|
61 |
minimum=0.0,
|
62 |
+
maximum=1.0,
|
63 |
+
step=0.01,
|
64 |
interactive=True,
|
65 |
info="Higher values sample more low-probability tokens",
|
66 |
),
|
67 |
gr.Slider(
|
68 |
label="Repetition penalty",
|
69 |
+
value=1.0,
|
70 |
minimum=1.0,
|
71 |
maximum=2.0,
|
72 |
+
step=0.01,
|
73 |
interactive=True,
|
74 |
info="Penalize repeated tokens",
|
75 |
)
|
76 |
]
|
77 |
|
|
|
78 |
gr.ChatInterface(
|
79 |
fn=generate,
|
80 |
chatbot=gr.Chatbot(show_label=False, show_share_button=False, show_copy_button=True, likeable=True, layout="panel"),
|
81 |
additional_inputs=additional_inputs,
|
82 |
+
title="Mistral 7B v0.3"
|
83 |
).launch(show_api=False)
|
84 |
|
|
|
85 |
gr.load("models/ehristoforu/dalle-3-xl-v2").launch()
|
86 |
+
gr.load("models/microsoft/Phi-3-mini-4k-instruct").launch()
|
|