Spaces:
Sleeping
Sleeping
Upload app.py
Browse files
app.py
ADDED
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
|
3 |
+
num_sequences=4
|
4 |
+
|
5 |
+
demo_mode = False
|
6 |
+
if not demo_mode:
|
7 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
8 |
+
|
9 |
+
model = AutoModelForCausalLM.from_pretrained("d2weber/german-gpt2-finetuned-coldmirror-hpodcast1")
|
10 |
+
tokenizer = AutoTokenizer.from_pretrained("dbmdz/german-gpt2", use_fast=True)
|
11 |
+
lm = pipeline("text-generation", model=model, tokenizer=tokenizer)
|
12 |
+
|
13 |
+
def generate(*args, **kwargs):
|
14 |
+
return [o["generated_text"] for o in lm(*args, **kwargs, pad_token_id=tokenizer.eos_token_id)]
|
15 |
+
|
16 |
+
with gr.Blocks() as app:
|
17 |
+
prompt = gr.TextArea(value="Hallo und herzlich willkommen", label="Input")
|
18 |
+
sequences = []
|
19 |
+
for _ in range(num_sequences):
|
20 |
+
seq = gr.Textbox("", visible=False)
|
21 |
+
box = gr.CheckboxGroup(choices=[], label="", interactive=True)
|
22 |
+
sequences.append(seq)
|
23 |
+
|
24 |
+
@seq.change(inputs=seq, outputs=box)
|
25 |
+
def split(seq):
|
26 |
+
return gr.CheckboxGroup(seq.split(), value=[])
|
27 |
+
|
28 |
+
@box.select(inputs=[prompt, seq], outputs=prompt)
|
29 |
+
def handle(prompt, sequence, selected: gr.SelectData):
|
30 |
+
to_append = " ".join(sequence.split()[:selected.index+1])
|
31 |
+
delimiter = " " if to_append[:1].isalnum() else ""
|
32 |
+
return prompt.rstrip() + delimiter + to_append
|
33 |
+
|
34 |
+
max_new_tokens = gr.Slider(1, 100, value=18, step=1, label="How long should the generated sequences be:")
|
35 |
+
|
36 |
+
gr.Examples([
|
37 |
+
["Hallo und herzlich willkommen"],
|
38 |
+
], prompt)
|
39 |
+
|
40 |
+
@prompt.change(inputs=[prompt, max_new_tokens], outputs=sequences)
|
41 |
+
def handle(prompt, max_new_tokens):
|
42 |
+
prompt = prompt.rstrip()
|
43 |
+
texts = ["some new words"]*num_sequences if demo_mode else generate(
|
44 |
+
prompt,
|
45 |
+
return_full_text=False,
|
46 |
+
num_return_sequences=num_sequences,
|
47 |
+
max_new_tokens=int(max_new_tokens),
|
48 |
+
)
|
49 |
+
return texts
|
50 |
+
|
51 |
+
app.launch()
|