Commit
•
a3ea183
1
Parent(s):
eeb9742
Update app.py
Browse files
app.py
CHANGED
@@ -1,26 +1,16 @@
|
|
1 |
-
from huggingface_hub import InferenceClient
|
2 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
|
4 |
-
|
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} "
|
11 |
-
prompt += f"[INST] {message} [/INST]"
|
12 |
-
return prompt
|
13 |
-
|
14 |
-
def generate(
|
15 |
-
prompt, history, temperature=0.3, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,
|
16 |
-
):
|
17 |
-
temperature = float(temperature)
|
18 |
-
if temperature < 1e-2:
|
19 |
-
temperature = 1e-2
|
20 |
top_p = float(top_p)
|
21 |
|
22 |
generate_kwargs = dict(
|
23 |
-
temperature=
|
24 |
max_new_tokens=max_new_tokens,
|
25 |
top_p=top_p,
|
26 |
repetition_penalty=repetition_penalty,
|
@@ -28,7 +18,7 @@ def generate(
|
|
28 |
seed=42,
|
29 |
)
|
30 |
|
31 |
-
formatted_prompt =
|
32 |
|
33 |
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
|
34 |
output = ""
|
@@ -38,11 +28,12 @@ def generate(
|
|
38 |
yield output
|
39 |
return output
|
40 |
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
|
|
47 |
|
48 |
-
|
|
|
|
|
1 |
import gradio as gr
|
2 |
+
from huggingface_hub import InferenceClient
|
3 |
+
|
4 |
+
client = InferenceClient("mistralai/Mistral-7B-v0.1")
|
5 |
+
|
6 |
+
# Fixed temperature value
|
7 |
+
fixed_temperature = 0.9 # You can adjust this value as needed
|
8 |
|
9 |
+
def generate(prompt, max_new_tokens=6056, top_p=0.95, repetition_penalty=1.0):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
top_p = float(top_p)
|
11 |
|
12 |
generate_kwargs = dict(
|
13 |
+
temperature=fixed_temperature,
|
14 |
max_new_tokens=max_new_tokens,
|
15 |
top_p=top_p,
|
16 |
repetition_penalty=repetition_penalty,
|
|
|
18 |
seed=42,
|
19 |
)
|
20 |
|
21 |
+
formatted_prompt = f"<s>[INST] {prompt} [/INST]"
|
22 |
|
23 |
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
|
24 |
output = ""
|
|
|
28 |
yield output
|
29 |
return output
|
30 |
|
31 |
+
iface = gr.Interface(
|
32 |
+
fn=generate,
|
33 |
+
inputs="text",
|
34 |
+
outputs="text",
|
35 |
+
title="Mistralai-Mistral-7B-Instruct Chat",
|
36 |
+
live=False # Set live to False to add a "Submit" button
|
37 |
+
)
|
38 |
|
39 |
+
iface.launch()
|