Spaces:
Running
Running
Fix issues
Browse files
app.py
CHANGED
@@ -1,11 +1,10 @@
|
|
1 |
import os
|
2 |
-
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
|
3 |
|
4 |
from PIL import Image
|
5 |
from huggingface_hub import hf_hub_download
|
6 |
|
7 |
-
|
8 |
-
unicorn_image_path = "unicorn.png"
|
9 |
|
10 |
import gradio as gr
|
11 |
from transformers import (
|
@@ -22,7 +21,6 @@ from tensorflow.keras.models import load_model
|
|
22 |
from tensorflow.keras.preprocessing.sequence import pad_sequences
|
23 |
import re
|
24 |
|
25 |
-
# Load GRU, LSTM, and BiLSTM models and tokenizers
|
26 |
gru_repo_id = "arjahojnik/GRU-sentiment-model"
|
27 |
gru_model_path = hf_hub_download(repo_id=gru_repo_id, filename="best_GRU_tuning_model.h5")
|
28 |
gru_model = load_model(gru_model_path)
|
@@ -44,13 +42,11 @@ bilstm_tokenizer_path = hf_hub_download(repo_id=bilstm_repo_id, filename="my_tok
|
|
44 |
with open(bilstm_tokenizer_path, "rb") as f:
|
45 |
bilstm_tokenizer = pickle.load(f)
|
46 |
|
47 |
-
# Preprocessing function for text
|
48 |
def preprocess_text(text):
|
49 |
text = text.lower()
|
50 |
text = re.sub(r"[^a-zA-Z\s]", "", text).strip()
|
51 |
return text
|
52 |
|
53 |
-
# Prediction functions for GRU, LSTM, and BiLSTM
|
54 |
def predict_with_gru(text):
|
55 |
cleaned = preprocess_text(text)
|
56 |
seq = gru_tokenizer.texts_to_sequences([cleaned])
|
@@ -75,13 +71,12 @@ def predict_with_bilstm(text):
|
|
75 |
predicted_class = np.argmax(probs, axis=1)[0]
|
76 |
return int(predicted_class + 1)
|
77 |
|
78 |
-
# Load other models
|
79 |
models = {
|
80 |
"DistilBERT": {
|
81 |
"tokenizer": DistilBertTokenizerFast.from_pretrained("nhull/distilbert-sentiment-model"),
|
82 |
"model": DistilBertForSequenceClassification.from_pretrained("nhull/distilbert-sentiment-model"),
|
83 |
},
|
84 |
-
"Logistic Regression": {},
|
85 |
"BERT Multilingual (NLP Town)": {
|
86 |
"tokenizer": AutoTokenizer.from_pretrained("nlptown/bert-base-multilingual-uncased-sentiment"),
|
87 |
"model": AutoModelForSequenceClassification.from_pretrained("nlptown/bert-base-multilingual-uncased-sentiment"),
|
@@ -96,7 +91,6 @@ models = {
|
|
96 |
}
|
97 |
}
|
98 |
|
99 |
-
# Logistic regression model and TF-IDF vectorizer
|
100 |
logistic_regression_repo = "nhull/logistic-regression-model"
|
101 |
log_reg_model_path = hf_hub_download(repo_id=logistic_regression_repo, filename="logistic_regression_model.pkl")
|
102 |
with open(log_reg_model_path, "rb") as model_file:
|
@@ -106,13 +100,11 @@ vectorizer_path = hf_hub_download(repo_id=logistic_regression_repo, filename="tf
|
|
106 |
with open(vectorizer_path, "rb") as vectorizer_file:
|
107 |
vectorizer = pickle.load(vectorizer_file)
|
108 |
|
109 |
-
# Move HuggingFace models to device
|
110 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
111 |
for model_data in models.values():
|
112 |
if "model" in model_data:
|
113 |
model_data["model"].to(device)
|
114 |
|
115 |
-
# Prediction functions for other models
|
116 |
def predict_with_distilbert(text):
|
117 |
tokenizer = models["DistilBERT"]["tokenizer"]
|
118 |
model = models["DistilBERT"]["model"]
|
@@ -158,7 +150,6 @@ def predict_with_roberta_ordek899(text):
|
|
158 |
predictions = logits.argmax(axis=-1).cpu().numpy()
|
159 |
return int(predictions[0] + 1)
|
160 |
|
161 |
-
# Unified function for analysis
|
162 |
def analyze_sentiment_and_statistics(text):
|
163 |
results = {
|
164 |
"Logistic Regression": predict_with_logistic_regression(text),
|
@@ -190,7 +181,6 @@ def analyze_sentiment_and_statistics(text):
|
|
190 |
}
|
191 |
return results, statistics
|
192 |
|
193 |
-
# Gradio Interface
|
194 |
with gr.Blocks(
|
195 |
css="""
|
196 |
.gradio-container {
|
@@ -252,10 +242,9 @@ with gr.Blocks(
|
|
252 |
}
|
253 |
"""
|
254 |
) as demo:
|
255 |
-
# Add the unicorn image at the start
|
256 |
gr.Image(
|
257 |
-
value=unicorn_image_path,
|
258 |
-
type="filepath",
|
259 |
elem_classes=["unicorn-image"]
|
260 |
)
|
261 |
|
@@ -269,7 +258,6 @@ with gr.Blocks(
|
|
269 |
"""
|
270 |
)
|
271 |
|
272 |
-
|
273 |
with gr.Row():
|
274 |
with gr.Column():
|
275 |
text_input = gr.Textbox(
|
@@ -285,12 +273,15 @@ with gr.Blocks(
|
|
285 |
"Terrible! The room was dirty, and the service was non-existent."
|
286 |
]
|
287 |
sample_dropdown = gr.Dropdown(
|
288 |
-
choices=sample_reviews,
|
289 |
label="Or select a sample review:",
|
|
|
290 |
interactive=True
|
291 |
)
|
292 |
|
293 |
def update_textbox(selected_sample):
|
|
|
|
|
294 |
return selected_sample
|
295 |
|
296 |
sample_dropdown.change(
|
@@ -318,12 +309,16 @@ with gr.Blocks(
|
|
318 |
tinybert_output = gr.Textbox(label="TinyBERT", interactive=False)
|
319 |
roberta_output = gr.Textbox(label="RoBERTa", interactive=False)
|
320 |
|
|
|
|
|
|
|
|
|
|
|
321 |
with gr.Row():
|
322 |
with gr.Column():
|
323 |
gr.Markdown("### Statistics")
|
324 |
stats_output = gr.Textbox(label="Statistics", interactive=False)
|
325 |
|
326 |
-
# Add footer
|
327 |
gr.Markdown(
|
328 |
"""
|
329 |
<footer>
|
@@ -335,80 +330,69 @@ with gr.Blocks(
|
|
335 |
</footer>
|
336 |
"""
|
337 |
)
|
|
|
|
|
|
|
|
|
338 |
def process_input_and_analyze(text_input):
|
339 |
-
# Check for empty input
|
340 |
if not text_input.strip():
|
341 |
funny_message = "Are you sure you wrote something? Try again! 🧐"
|
342 |
return (
|
343 |
-
|
344 |
-
funny_message,
|
345 |
-
|
346 |
-
funny_message, # BiLSTM
|
347 |
-
funny_message, # DistilBERT
|
348 |
-
funny_message, # BERT Multilingual
|
349 |
-
funny_message, # TinyBERT
|
350 |
-
funny_message, # RoBERTa
|
351 |
-
"No statistics to display, as nothing was input. 🤷♀️"
|
352 |
)
|
353 |
|
354 |
-
# Check for one letter/number input
|
355 |
if len(text_input.strip()) == 1 or text_input.strip().isdigit():
|
356 |
funny_message = "Why not write something that makes sense? 🤔"
|
357 |
return (
|
358 |
-
|
359 |
-
funny_message,
|
360 |
-
|
361 |
-
funny_message, # BiLSTM
|
362 |
-
funny_message, # DistilBERT
|
363 |
-
funny_message, # BERT Multilingual
|
364 |
-
funny_message, # TinyBERT
|
365 |
-
funny_message, # RoBERTa
|
366 |
-
"No statistics to display for one letter or number. 😅"
|
367 |
)
|
368 |
|
369 |
-
# Check if the review is shorter than 5 words
|
370 |
if len(text_input.split()) < 5:
|
371 |
results, statistics = analyze_sentiment_and_statistics(text_input)
|
372 |
short_message = "Maybe try with some longer text next time. 😉"
|
|
|
|
|
|
|
|
|
|
|
373 |
return (
|
374 |
-
|
375 |
-
|
376 |
-
|
377 |
-
|
378 |
-
|
379 |
-
|
380 |
-
|
381 |
-
|
382 |
-
|
|
|
383 |
)
|
384 |
|
385 |
-
# Proceed with normal sentiment analysis if none of the above conditions apply
|
386 |
results, statistics = analyze_sentiment_and_statistics(text_input)
|
|
|
|
|
387 |
if "Message" in statistics:
|
388 |
-
|
389 |
-
results["Logistic Regression"],
|
390 |
-
results["GRU Model"],
|
391 |
-
results["LSTM Model"],
|
392 |
-
results["BiLSTM Model"],
|
393 |
-
results["DistilBERT"],
|
394 |
-
results["BERT Multilingual (NLP Town)"],
|
395 |
-
results["TinyBERT"],
|
396 |
-
results["RoBERTa"],
|
397 |
-
f"Statistics:\n{statistics['Message']}\nAverage Score: {statistics['Average Score']}"
|
398 |
-
)
|
399 |
else:
|
400 |
-
|
401 |
-
|
402 |
-
|
403 |
-
|
404 |
-
|
405 |
-
|
406 |
-
|
407 |
-
|
408 |
-
|
409 |
-
|
410 |
-
)
|
411 |
-
|
|
|
|
|
|
|
412 |
analyze_button.click(
|
413 |
process_input_and_analyze,
|
414 |
inputs=[text_input],
|
@@ -420,9 +404,10 @@ with gr.Blocks(
|
|
420 |
distilbert_output,
|
421 |
bert_output,
|
422 |
tinybert_output,
|
423 |
-
roberta_output,
|
424 |
-
|
|
|
425 |
]
|
426 |
)
|
427 |
|
428 |
-
demo.launch()
|
|
|
1 |
import os
|
2 |
+
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
|
3 |
|
4 |
from PIL import Image
|
5 |
from huggingface_hub import hf_hub_download
|
6 |
|
7 |
+
unicorn_image_path = "scripts/demo/unicorn.png"
|
|
|
8 |
|
9 |
import gradio as gr
|
10 |
from transformers import (
|
|
|
21 |
from tensorflow.keras.preprocessing.sequence import pad_sequences
|
22 |
import re
|
23 |
|
|
|
24 |
gru_repo_id = "arjahojnik/GRU-sentiment-model"
|
25 |
gru_model_path = hf_hub_download(repo_id=gru_repo_id, filename="best_GRU_tuning_model.h5")
|
26 |
gru_model = load_model(gru_model_path)
|
|
|
42 |
with open(bilstm_tokenizer_path, "rb") as f:
|
43 |
bilstm_tokenizer = pickle.load(f)
|
44 |
|
|
|
45 |
def preprocess_text(text):
|
46 |
text = text.lower()
|
47 |
text = re.sub(r"[^a-zA-Z\s]", "", text).strip()
|
48 |
return text
|
49 |
|
|
|
50 |
def predict_with_gru(text):
|
51 |
cleaned = preprocess_text(text)
|
52 |
seq = gru_tokenizer.texts_to_sequences([cleaned])
|
|
|
71 |
predicted_class = np.argmax(probs, axis=1)[0]
|
72 |
return int(predicted_class + 1)
|
73 |
|
|
|
74 |
models = {
|
75 |
"DistilBERT": {
|
76 |
"tokenizer": DistilBertTokenizerFast.from_pretrained("nhull/distilbert-sentiment-model"),
|
77 |
"model": DistilBertForSequenceClassification.from_pretrained("nhull/distilbert-sentiment-model"),
|
78 |
},
|
79 |
+
"Logistic Regression": {},
|
80 |
"BERT Multilingual (NLP Town)": {
|
81 |
"tokenizer": AutoTokenizer.from_pretrained("nlptown/bert-base-multilingual-uncased-sentiment"),
|
82 |
"model": AutoModelForSequenceClassification.from_pretrained("nlptown/bert-base-multilingual-uncased-sentiment"),
|
|
|
91 |
}
|
92 |
}
|
93 |
|
|
|
94 |
logistic_regression_repo = "nhull/logistic-regression-model"
|
95 |
log_reg_model_path = hf_hub_download(repo_id=logistic_regression_repo, filename="logistic_regression_model.pkl")
|
96 |
with open(log_reg_model_path, "rb") as model_file:
|
|
|
100 |
with open(vectorizer_path, "rb") as vectorizer_file:
|
101 |
vectorizer = pickle.load(vectorizer_file)
|
102 |
|
|
|
103 |
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
104 |
for model_data in models.values():
|
105 |
if "model" in model_data:
|
106 |
model_data["model"].to(device)
|
107 |
|
|
|
108 |
def predict_with_distilbert(text):
|
109 |
tokenizer = models["DistilBERT"]["tokenizer"]
|
110 |
model = models["DistilBERT"]["model"]
|
|
|
150 |
predictions = logits.argmax(axis=-1).cpu().numpy()
|
151 |
return int(predictions[0] + 1)
|
152 |
|
|
|
153 |
def analyze_sentiment_and_statistics(text):
|
154 |
results = {
|
155 |
"Logistic Regression": predict_with_logistic_regression(text),
|
|
|
181 |
}
|
182 |
return results, statistics
|
183 |
|
|
|
184 |
with gr.Blocks(
|
185 |
css="""
|
186 |
.gradio-container {
|
|
|
242 |
}
|
243 |
"""
|
244 |
) as demo:
|
|
|
245 |
gr.Image(
|
246 |
+
value=unicorn_image_path,
|
247 |
+
type="filepath",
|
248 |
elem_classes=["unicorn-image"]
|
249 |
)
|
250 |
|
|
|
258 |
"""
|
259 |
)
|
260 |
|
|
|
261 |
with gr.Row():
|
262 |
with gr.Column():
|
263 |
text_input = gr.Textbox(
|
|
|
273 |
"Terrible! The room was dirty, and the service was non-existent."
|
274 |
]
|
275 |
sample_dropdown = gr.Dropdown(
|
276 |
+
choices=["Select an option"] + sample_reviews,
|
277 |
label="Or select a sample review:",
|
278 |
+
value=None,
|
279 |
interactive=True
|
280 |
)
|
281 |
|
282 |
def update_textbox(selected_sample):
|
283 |
+
if selected_sample == "Select an option":
|
284 |
+
return ""
|
285 |
return selected_sample
|
286 |
|
287 |
sample_dropdown.change(
|
|
|
309 |
tinybert_output = gr.Textbox(label="TinyBERT", interactive=False)
|
310 |
roberta_output = gr.Textbox(label="RoBERTa", interactive=False)
|
311 |
|
312 |
+
with gr.Row():
|
313 |
+
with gr.Column():
|
314 |
+
gr.Markdown("### Feedback")
|
315 |
+
feedback_output = gr.Textbox(label="Feedback", interactive=False)
|
316 |
+
|
317 |
with gr.Row():
|
318 |
with gr.Column():
|
319 |
gr.Markdown("### Statistics")
|
320 |
stats_output = gr.Textbox(label="Statistics", interactive=False)
|
321 |
|
|
|
322 |
gr.Markdown(
|
323 |
"""
|
324 |
<footer>
|
|
|
330 |
</footer>
|
331 |
"""
|
332 |
)
|
333 |
+
|
334 |
+
def convert_to_stars(rating):
|
335 |
+
return "★" * rating + "☆" * (5 - rating)
|
336 |
+
|
337 |
def process_input_and_analyze(text_input):
|
|
|
338 |
if not text_input.strip():
|
339 |
funny_message = "Are you sure you wrote something? Try again! 🧐"
|
340 |
return (
|
341 |
+
"", "", "", "", "", "", "", "",
|
342 |
+
funny_message,
|
343 |
+
"No statistics can be shown."
|
|
|
|
|
|
|
|
|
|
|
|
|
344 |
)
|
345 |
|
|
|
346 |
if len(text_input.strip()) == 1 or text_input.strip().isdigit():
|
347 |
funny_message = "Why not write something that makes sense? 🤔"
|
348 |
return (
|
349 |
+
"", "", "", "", "", "", "", "",
|
350 |
+
funny_message,
|
351 |
+
"No statistics can be shown."
|
|
|
|
|
|
|
|
|
|
|
|
|
352 |
)
|
353 |
|
|
|
354 |
if len(text_input.split()) < 5:
|
355 |
results, statistics = analyze_sentiment_and_statistics(text_input)
|
356 |
short_message = "Maybe try with some longer text next time. 😉"
|
357 |
+
stats_text = (
|
358 |
+
f"Statistics:\n{statistics['Lowest Score']}\n{statistics['Highest Score']}\n"
|
359 |
+
f"Average Score: {statistics['Average Score']}"
|
360 |
+
if "Message" not in statistics else f"Statistics:\n{statistics['Message']}"
|
361 |
+
)
|
362 |
return (
|
363 |
+
convert_to_stars(results['Logistic Regression']),
|
364 |
+
convert_to_stars(results['GRU Model']),
|
365 |
+
convert_to_stars(results['LSTM Model']),
|
366 |
+
convert_to_stars(results['BiLSTM Model']),
|
367 |
+
convert_to_stars(results['DistilBERT']),
|
368 |
+
convert_to_stars(results['BERT Multilingual (NLP Town)']),
|
369 |
+
convert_to_stars(results['TinyBERT']),
|
370 |
+
convert_to_stars(results['RoBERTa']),
|
371 |
+
short_message,
|
372 |
+
stats_text
|
373 |
)
|
374 |
|
|
|
375 |
results, statistics = analyze_sentiment_and_statistics(text_input)
|
376 |
+
feedback_message = "Sentiment analysis completed successfully! 😊"
|
377 |
+
|
378 |
if "Message" in statistics:
|
379 |
+
stats_text = f"Statistics:\n{statistics['Message']}\nAverage Score: {statistics['Average Score']}"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
380 |
else:
|
381 |
+
stats_text = f"Statistics:\n{statistics['Lowest Score']}\n{statistics['Highest Score']}\nAverage Score: {statistics['Average Score']}"
|
382 |
+
|
383 |
+
return (
|
384 |
+
convert_to_stars(results["Logistic Regression"]),
|
385 |
+
convert_to_stars(results["GRU Model"]),
|
386 |
+
convert_to_stars(results["LSTM Model"]),
|
387 |
+
convert_to_stars(results["BiLSTM Model"]),
|
388 |
+
convert_to_stars(results["DistilBERT"]),
|
389 |
+
convert_to_stars(results["BERT Multilingual (NLP Town)"]),
|
390 |
+
convert_to_stars(results["TinyBERT"]),
|
391 |
+
convert_to_stars(results["RoBERTa"]),
|
392 |
+
feedback_message,
|
393 |
+
stats_text
|
394 |
+
)
|
395 |
+
|
396 |
analyze_button.click(
|
397 |
process_input_and_analyze,
|
398 |
inputs=[text_input],
|
|
|
404 |
distilbert_output,
|
405 |
bert_output,
|
406 |
tinybert_output,
|
407 |
+
roberta_output,
|
408 |
+
feedback_output,
|
409 |
+
stats_output
|
410 |
]
|
411 |
)
|
412 |
|
413 |
+
demo.launch()
|