Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
!pip install git+https://github.com/huggingface/transformers
|
2 |
+
|
3 |
+
import gradio as gr
|
4 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
5 |
+
from threading import Thread
|
6 |
+
|
7 |
+
tok = AutoTokenizer.from_pretrained("distilgpt2")
|
8 |
+
model = AutoModelForCausalLM.from_pretrained("distilgpt2")
|
9 |
+
|
10 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
11 |
+
n_gpu = 0 if torch.cuda.is_available()==False else torch.cuda.device_count()
|
12 |
+
model.to(device)
|
13 |
+
|
14 |
+
early_stop_pattern = tok.eos_token
|
15 |
+
print(f'Early stop pattern = \"{early_stop_pattern}\"')
|
16 |
+
|
17 |
+
def generate(text = ""):
|
18 |
+
streamer = TextIteratorStreamer(tok)
|
19 |
+
if len(text) == 0:
|
20 |
+
text = " "
|
21 |
+
inputs = tok([text], return_tensors="pt")
|
22 |
+
generation_kwargs = dict(inputs, streamer=streamer, repetition_penalty=2.0, do_sample=True, top_k=40, top_p=0.97, max_new_tokens=128)
|
23 |
+
thread = Thread(target=model.generate, kwargs=generation_kwargs)
|
24 |
+
thread.start()
|
25 |
+
generated_text = ""
|
26 |
+
for new_text in streamer:
|
27 |
+
yield generated_text + new_text
|
28 |
+
#print(new_text, end ="")
|
29 |
+
generated_text += new_text
|
30 |
+
if early_stop_pattern in generated_text:
|
31 |
+
generated_text = generated_text[: generated_text.find(early_stop_pattern) if early_stop_pattern else None]
|
32 |
+
streamer.end()
|
33 |
+
#print("\n--\n")
|
34 |
+
yield generated_text
|
35 |
+
return
|
36 |
+
|
37 |
+
demo = gr.Interface(
|
38 |
+
title="TextIteratorStreamer + Gradio demo",
|
39 |
+
fn=generate,
|
40 |
+
inputs=gr.inputs.Textbox(lines=5, label="Input Text"),
|
41 |
+
outputs=gr.outputs.Textbox(label="Generated Text"),
|
42 |
+
)
|
43 |
+
|
44 |
+
demo.queue()
|
45 |
+
demo.launch()
|