Spaces:
Runtime error
Runtime error
Simon Salmon
commited on
Commit
·
449de4f
1
Parent(s):
07bf672
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,12 +1,16 @@
|
|
| 1 |
import torch
|
| 2 |
from transformers import T5ForConditionalGeneration,T5Tokenizer, AutoTokenizer, AutoModelForSeq2SeqLM
|
| 3 |
import streamlit as st
|
|
|
|
| 4 |
model_name = st.text_input("Pick a Model", "seduerr/t5-pawraphrase")
|
| 5 |
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
|
| 6 |
tokenizer = AutoTokenizer.from_pretrained("t5-base")
|
| 7 |
|
| 8 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 9 |
model = model.to(device)
|
|
|
|
|
|
|
|
|
|
| 10 |
def translate_to_english(model, tokenizer, text):
|
| 11 |
translated_text = []
|
| 12 |
text = "paraphrase: " + text + " </s>"
|
|
@@ -16,10 +20,11 @@ def translate_to_english(model, tokenizer, text):
|
|
| 16 |
input_ids=input_ids, attention_mask=attention_masks,
|
| 17 |
do_sample=True,
|
| 18 |
max_length=256,
|
|
|
|
| 19 |
top_k=120,
|
| 20 |
top_p=0.98,
|
| 21 |
early_stopping=True,
|
| 22 |
-
num_return_sequences=
|
| 23 |
)
|
| 24 |
for beam_output in beam_outputs:
|
| 25 |
sent = tokenizer.decode(beam_output, skip_special_tokens=True,clean_up_tokenization_spaces=True)
|
|
@@ -27,11 +32,10 @@ def translate_to_english(model, tokenizer, text):
|
|
| 27 |
translated_text.append(sent)
|
| 28 |
return translated_text
|
| 29 |
|
| 30 |
-
st.title("Auto Translate (To English)")
|
| 31 |
text = st.text_input("Okay")
|
| 32 |
st.text("What you wrote: ")
|
| 33 |
st.write(text)
|
| 34 |
-
st.text("
|
| 35 |
if text:
|
| 36 |
translated_text = translate_to_english(model, tokenizer, text)
|
| 37 |
st.write(translated_text if translated_text else "No translation found")
|
|
|
|
| 1 |
import torch
|
| 2 |
from transformers import T5ForConditionalGeneration,T5Tokenizer, AutoTokenizer, AutoModelForSeq2SeqLM
|
| 3 |
import streamlit as st
|
| 4 |
+
st.title("Paraphrase")
|
| 5 |
model_name = st.text_input("Pick a Model", "seduerr/t5-pawraphrase")
|
| 6 |
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
|
| 7 |
tokenizer = AutoTokenizer.from_pretrained("t5-base")
|
| 8 |
|
| 9 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
| 10 |
model = model.to(device)
|
| 11 |
+
temp = st.sidebar.slider("Temperature", 0.7, 1.5)
|
| 12 |
+
number_of_outputs = st.sidebar.slider("Number of Outputs", 1, 10)
|
| 13 |
+
|
| 14 |
def translate_to_english(model, tokenizer, text):
|
| 15 |
translated_text = []
|
| 16 |
text = "paraphrase: " + text + " </s>"
|
|
|
|
| 20 |
input_ids=input_ids, attention_mask=attention_masks,
|
| 21 |
do_sample=True,
|
| 22 |
max_length=256,
|
| 23 |
+
temperature = temp,
|
| 24 |
top_k=120,
|
| 25 |
top_p=0.98,
|
| 26 |
early_stopping=True,
|
| 27 |
+
num_return_sequences=number_of_outputs,
|
| 28 |
)
|
| 29 |
for beam_output in beam_outputs:
|
| 30 |
sent = tokenizer.decode(beam_output, skip_special_tokens=True,clean_up_tokenization_spaces=True)
|
|
|
|
| 32 |
translated_text.append(sent)
|
| 33 |
return translated_text
|
| 34 |
|
|
|
|
| 35 |
text = st.text_input("Okay")
|
| 36 |
st.text("What you wrote: ")
|
| 37 |
st.write(text)
|
| 38 |
+
st.text("Output: ")
|
| 39 |
if text:
|
| 40 |
translated_text = translate_to_english(model, tokenizer, text)
|
| 41 |
st.write(translated_text if translated_text else "No translation found")
|