Spaces:
Sleeping
Sleeping
eaglelandsonce
commited on
Commit
•
38926c5
1
Parent(s):
367af4e
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
3 |
+
|
4 |
+
@st.cache(allow_output_mutation=True)
|
5 |
+
def load_model():
|
6 |
+
|
7 |
+
checkpoint = "CohereForAI/aya-101"
|
8 |
+
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
|
9 |
+
model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint)
|
10 |
+
|
11 |
+
return tokenizer, model
|
12 |
+
|
13 |
+
tokenizer, model = load_model()
|
14 |
+
|
15 |
+
st.title("Neural Machine Translation")
|
16 |
+
|
17 |
+
text = st.text_input("Enter text to translate:")
|
18 |
+
src_language = "auto"
|
19 |
+
tgt_language = st.selectbox("Translate to:", ["English", "Hindi", "Turkish"])
|
20 |
+
|
21 |
+
if st.button("Translate"):
|
22 |
+
|
23 |
+
text_input_ids = tokenizer.encode(text, return_tensors="pt")
|
24 |
+
|
25 |
+
if tgt_language == "English":
|
26 |
+
tgt_language = "en"
|
27 |
+
elif tgt_language == "Hindi":
|
28 |
+
tgt_language = "hi"
|
29 |
+
elif tgt_language == "Turkish":
|
30 |
+
tgt_language = "tr"
|
31 |
+
|
32 |
+
output_ids = model.generate(text_input_ids,
|
33 |
+
forced_bos_token_id=tokenizer.lang_code_to_id[tgt_language])
|
34 |
+
translated_text = tokenizer.decode(output_ids[0], skip_special_tokens=True)
|
35 |
+
|
36 |
+
st.write(f"Translated text in {tgt_language}:")
|
37 |
+
st.write(translated_text)
|