Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -2,49 +2,36 @@ import streamlit as st
|
|
2 |
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
3 |
import torch
|
4 |
|
5 |
-
st.title("
|
6 |
|
7 |
-
#
|
8 |
model_list = [
|
9 |
-
("
|
10 |
-
("
|
11 |
-
("distilbert-base-uncased", "DistilBERT Base")
|
12 |
]
|
13 |
|
14 |
@st.cache_resource
|
15 |
-
def
|
16 |
-
|
|
|
|
|
17 |
|
18 |
-
|
19 |
-
def load_tokenizer(model_name):
|
20 |
-
return AutoTokenizer.from_pretrained(model_name)
|
21 |
-
|
22 |
-
# Load models + tokenizers
|
23 |
-
models = [(load_model(name), load_tokenizer(name), label) for name, label in model_list]
|
24 |
|
25 |
-
#
|
26 |
-
text_input = st.text_area("Enter text
|
27 |
|
28 |
-
|
29 |
-
selected_model_label = st.selectbox("Select a model:", [label for _, _, label in models])
|
30 |
-
|
31 |
-
# Find selected model
|
32 |
-
for model, tokenizer, label in models:
|
33 |
-
if label == selected_model_label:
|
34 |
-
selected_model = model
|
35 |
-
selected_tokenizer = tokenizer
|
36 |
-
break
|
37 |
|
38 |
if st.button("Classify"):
|
39 |
-
if text_input.strip()
|
40 |
st.warning("Please enter some text!")
|
41 |
else:
|
42 |
-
|
|
|
43 |
with torch.no_grad():
|
44 |
-
|
45 |
-
logits = outputs.logits
|
46 |
probs = torch.softmax(logits, dim=-1).squeeze().tolist()
|
47 |
-
|
48 |
-
st.write("### Classification probabilities:")
|
49 |
for i, prob in enumerate(probs):
|
50 |
st.write(f"Class {i}: {prob:.4f}")
|
|
|
2 |
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
3 |
import torch
|
4 |
|
5 |
+
st.title("Text Sentiment Classifier")
|
6 |
|
7 |
+
# Valid fine-tuned models
|
8 |
model_list = [
|
9 |
+
("distilbert-base-uncased-finetuned-sst-2-english", "DistilBERT (SST-2)"),
|
10 |
+
("textattack/roberta-base-imdb", "RoBERTa (IMDB Sentiment)")
|
|
|
11 |
]
|
12 |
|
13 |
@st.cache_resource
|
14 |
+
def load_model_and_tokenizer(model_name):
|
15 |
+
model = AutoModelForSequenceClassification.from_pretrained(model_name)
|
16 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
17 |
+
return model, tokenizer
|
18 |
|
19 |
+
models = {label: load_model_and_tokenizer(name) for name, label in model_list}
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
+
# UI
|
22 |
+
text_input = st.text_area("Enter text:")
|
23 |
|
24 |
+
model_choice = st.selectbox("Choose model:", list(models.keys()))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
if st.button("Classify"):
|
27 |
+
if not text_input.strip():
|
28 |
st.warning("Please enter some text!")
|
29 |
else:
|
30 |
+
model, tokenizer = models[model_choice]
|
31 |
+
inputs = tokenizer(text_input, return_tensors="pt", truncation=True)
|
32 |
with torch.no_grad():
|
33 |
+
logits = model(**inputs).logits
|
|
|
34 |
probs = torch.softmax(logits, dim=-1).squeeze().tolist()
|
35 |
+
st.write("### Results:")
|
|
|
36 |
for i, prob in enumerate(probs):
|
37 |
st.write(f"Class {i}: {prob:.4f}")
|