Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,80 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
2 |
+
import torch
|
3 |
+
import gradio as gr
|
4 |
+
|
5 |
+
model_id = "witfoo/witq-1.0"
|
6 |
+
dtype = torch.float16 # float16 for Tesla T4, V100, bfloat16 for Ampere+
|
7 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id)
|
8 |
+
model = AutoModelForCausalLM.from_pretrained(
|
9 |
+
model_id,
|
10 |
+
torch_dtype=dtype,
|
11 |
+
device_map="auto",
|
12 |
+
)
|
13 |
+
|
14 |
+
preamble = "Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request."
|
15 |
+
|
16 |
+
|
17 |
+
|
18 |
+
def input_tokens(instruction, prompt):
|
19 |
+
messages = [
|
20 |
+
{"role": "system", "content": preamble + " " + instruction},
|
21 |
+
{"role": "user", "content": prompt},
|
22 |
+
]
|
23 |
+
inputs = tokenizer.apply_chat_template(
|
24 |
+
messages,
|
25 |
+
add_generation_prompt=True,
|
26 |
+
return_tensors="pt"
|
27 |
+
).to(model.device)
|
28 |
+
return inputs
|
29 |
+
|
30 |
+
|
31 |
+
|
32 |
+
def generate_response(instruction, input_text):
|
33 |
+
input_ids = input_tokens(instruction, input_text)
|
34 |
+
terminators = [
|
35 |
+
tokenizer.eos_token_id,
|
36 |
+
tokenizer.convert_tokens_to_ids("<|eot_id|>")
|
37 |
+
]
|
38 |
+
outputs = model.generate(
|
39 |
+
input_ids,
|
40 |
+
max_new_tokens=256,
|
41 |
+
eos_token_id=terminators,
|
42 |
+
do_sample=True,
|
43 |
+
temperature=0.6,
|
44 |
+
top_p=0.9,
|
45 |
+
)
|
46 |
+
|
47 |
+
# Extract the response portion
|
48 |
+
response = outputs[0][input_ids.shape[-1]:]
|
49 |
+
result = tokenizer.decode(response, skip_special_tokens=True)
|
50 |
+
return result
|
51 |
+
|
52 |
+
def chatbot(instructions, input_text):
|
53 |
+
response = generate_response(instructions, input_text)
|
54 |
+
return response
|
55 |
+
|
56 |
+
trained_instructions = [
|
57 |
+
"Answer this question",
|
58 |
+
"Create a JSON artifact from the message",
|
59 |
+
"Identify this syslog message",
|
60 |
+
"Explain this syslog message",
|
61 |
+
]
|
62 |
+
|
63 |
+
iface = gr.Interface(
|
64 |
+
fn=chatbot,
|
65 |
+
inputs=[
|
66 |
+
gr.Dropdown(choices=trained_instructions, label="Instruction"),
|
67 |
+
gr.Textbox(lines=2, placeholder="Enter your input here...", label="Input Text")
|
68 |
+
],
|
69 |
+
outputs=gr.Textbox(label="Response"),
|
70 |
+
title="WitQ Chatbot"
|
71 |
+
)
|
72 |
+
|
73 |
+
|
74 |
+
app = gr.Blocks()
|
75 |
+
|
76 |
+
with app:
|
77 |
+
iface.render()
|
78 |
+
|
79 |
+
|
80 |
+
app.launch()
|