Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from transformers import pipeline
|
2 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
3 |
+
import gradio as gr
|
4 |
+
from nltk.tokenize import sent_tokenize
|
5 |
+
import torch
|
6 |
+
|
7 |
+
model = "janny127/autotrain-5e45b-p5z66"
|
8 |
+
tokenizer = AutoTokenizer.from_pretrained(model)
|
9 |
+
|
10 |
+
pipeline = pipeline(
|
11 |
+
"text-generation",
|
12 |
+
model=model,
|
13 |
+
torch_dtype=torch.float16,
|
14 |
+
device_map="auto",
|
15 |
+
)
|
16 |
+
|
17 |
+
def predict(prompt, history):
|
18 |
+
# Prompt
|
19 |
+
formatted_prompt = (
|
20 |
+
f"### Human: {prompt}### Assistant:"
|
21 |
+
)
|
22 |
+
|
23 |
+
# Generate the Texts
|
24 |
+
sequences = pipeline(
|
25 |
+
formatted_prompt,
|
26 |
+
do_sample=True,
|
27 |
+
top_k=50,
|
28 |
+
top_p = 0.7,
|
29 |
+
num_return_sequences=1,
|
30 |
+
repetition_penalty=1.1,
|
31 |
+
max_new_tokens=500,
|
32 |
+
)
|
33 |
+
generated_text = sequences[0]['generated_text']
|
34 |
+
|
35 |
+
final_result = generated_text.split("### Assistant:")[1]
|
36 |
+
if " Human: " in final_result:
|
37 |
+
final_result = final_result.split(" Human: ")[0]
|
38 |
+
if " #" in final_result:
|
39 |
+
final_result = final_result.split(" #")[0]
|
40 |
+
|
41 |
+
# return generated_text.strip()
|
42 |
+
return final_result.strip()
|
43 |
+
|
44 |
+
gr.ChatInterface(predict,
|
45 |
+
title="Tinyllama_chatBot",
|
46 |
+
description="Ask Tiny llama any questions",
|
47 |
+
examples=['How to cook a fish?', 'Who is the president of US now?']
|
48 |
+
).launch() # Launching the web interface.
|
49 |
+
|
50 |
+
|