Spaces:
Sleeping
Sleeping
Upload app.py
Browse files
app.py
ADDED
@@ -0,0 +1,284 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
"""app.ipynb
|
3 |
+
|
4 |
+
Automatically generated by Colab.
|
5 |
+
|
6 |
+
Original file is located at
|
7 |
+
https://colab.research.google.com/drive/1y3yISz14Lpsr131OIJCKA77lwbFmEJzB
|
8 |
+
"""
|
9 |
+
|
10 |
+
import streamlit as st
|
11 |
+
import os
|
12 |
+
import joblib
|
13 |
+
import torch
|
14 |
+
import numpy as np
|
15 |
+
import html
|
16 |
+
from transformers import AutoTokenizer, AutoModel, logging as hf_logging
|
17 |
+
|
18 |
+
# Hugging Face Transformers ๋ก๊น
๋ ๋ฒจ ์ค์ (์ค๋ฅ๋ง ํ์)
|
19 |
+
hf_logging.set_verbosity_error()
|
20 |
+
|
21 |
+
# โโโโโโโโโโ ์ค์ (Hugging Face Spaces ํ๊ฒฝ์ ๋ง๊ฒ ์กฐ์ ) โโโโโโโโโโ
|
22 |
+
MODEL_NAME = "bert-base-uncased"
|
23 |
+
DEVICE = "cpu" # Hugging Face Spaces ๋ฌด๋ฃ ํฐ์ด๋ CPU ์ฌ์ฉ
|
24 |
+
SAVE_DIR = "์ ์ฅ์ ์ฅ1" # ์
๋ก๋ํ ํด๋๋ช
๊ณผ ์ผ์นํด์ผ ํจ
|
25 |
+
LAYER_ID = 4 # ์๋ณธ ์ฝ๋์ SeparationScore ์ต๊ณ ๋ ์ด์ด
|
26 |
+
SEED = 0 # ์๋ณธ ์ฝ๋์ SEED ๊ฐ
|
27 |
+
CLF_NAME = "linear" # ์๋ณธ ์ฝ๋์ CLF_NAME
|
28 |
+
|
29 |
+
# โโโโโโโโโโ ๋ชจ๋ธ ๋ก๋ (Streamlit ์บ์ ์ฌ์ฉ์ผ๋ก ์ฑ ์ ์ฒด์์ ํ ๋ฒ๋ง ์คํ) โโโโโโโโโโ
|
30 |
+
@st.cache_resource
|
31 |
+
def load_all_models_and_data():
|
32 |
+
"""
|
33 |
+
LDA, ๋ถ๋ฅ๊ธฐ, ํ ํฌ๋์ด์ , BERT ๋ชจ๋ธ ๋ฐ ๊ด๋ จ ํ๋ ฌ๋ค์ ๋ก๋ํฉ๋๋ค.
|
34 |
+
Hugging Face Spaces์ ๋ฐฐํฌ ์ ํ์ผ ๊ฒฝ๋ก๊ฐ ์ ํํด์ผ ํฉ๋๋ค.
|
35 |
+
"""
|
36 |
+
lda_file_path = os.path.join(SAVE_DIR, f"lda_layer{LAYER_ID}_seed{SEED}.pkl")
|
37 |
+
clf_file_path = os.path.join(SAVE_DIR, f"{CLF_NAME}_layer{LAYER_ID}_projlda_seed{SEED}.pkl")
|
38 |
+
|
39 |
+
# ํ์ผ ์กด์ฌ ์ฌ๋ถ ํ์ธ (๋ฐฐํฌ ํ๊ฒฝ ๋๋ฒ๊น
์ฉ)
|
40 |
+
if not os.path.isdir(SAVE_DIR):
|
41 |
+
st.error(f"์ค๋ฅ: ๋ชจ๋ธ ์ ์ฅ ๋๋ ํ ๋ฆฌ '{SAVE_DIR}'๋ฅผ ์ฐพ์ ์ ์์ต๋๋ค. Spaces์ ํด๋๊ฐ ์ฌ๋ฐ๋ฅด๊ฒ ์
๋ก๋๋์๋์ง, ์ด๋ฆ์ด ์ผ์นํ๋์ง ํ์ธํ์ธ์.")
|
42 |
+
return None
|
43 |
+
if not os.path.exists(lda_file_path):
|
44 |
+
st.error(f"์ค๋ฅ: LDA ๋ชจ๋ธ ํ์ผ '{lda_file_path}'๋ฅผ ์ฐพ์ ์ ์์ต๋๋ค. ํ์ผ ์ด๋ฆ๊ณผ ๊ฒฝ๋ก๋ฅผ ํ์ธํ์ธ์.")
|
45 |
+
return None
|
46 |
+
if not os.path.exists(clf_file_path):
|
47 |
+
st.error(f"์ค๋ฅ: ๋ถ๋ฅ๊ธฐ ๋ชจ๋ธ ํ์ผ '{clf_file_path}'๋ฅผ ์ฐพ์ ์ ์์ต๋๋ค. ํ์ผ ์ด๋ฆ๊ณผ ๊ฒฝ๋ก๋ฅผ ํ์ธํ์ธ์.")
|
48 |
+
return None
|
49 |
+
|
50 |
+
try:
|
51 |
+
lda = joblib.load(lda_file_path)
|
52 |
+
clf = joblib.load(clf_file_path)
|
53 |
+
except Exception as e:
|
54 |
+
st.error(f"๋ชจ๋ธ ํ์ผ ๋ก๋ ์ค ์ค๋ฅ ๋ฐ์: {e}")
|
55 |
+
st.error("ํ์ผ์ด ์์๋์๊ฑฐ๋, joblib ๋ฒ์ ํธํ์ฑ ๋ฌธ์ ๊ฐ ์์ ์ ์์ต๋๋ค.")
|
56 |
+
return None
|
57 |
+
|
58 |
+
if hasattr(clf, "base_estimator"): # Calibrated Ridge ๊ฒฝ์ฐ
|
59 |
+
clf = clf.base_estimator
|
60 |
+
|
61 |
+
# LDA ํ๋ ฌยทํ๊ท , ๋ถ๋ฅ๊ธฐ ๊ฐ์ค์น๋ฅผ PyTorch Tensor๋ก ๋ณํ
|
62 |
+
W_tensor = torch.tensor(lda.scalings_, dtype=torch.float32, device=DEVICE)
|
63 |
+
mu_vector = torch.tensor(lda.xbar_, dtype=torch.float32, device=DEVICE)
|
64 |
+
w_p_tensor = torch.tensor(clf.coef_, dtype=torch.float32, device=DEVICE)
|
65 |
+
b_p_vector = torch.tensor(clf.intercept_, dtype=torch.float32, device=DEVICE)
|
66 |
+
|
67 |
+
# Hugging Face ํ ํฌ๋์ด์ ๋ฐ BERT ๋ชจ๋ธ ๋ก๋
|
68 |
+
try:
|
69 |
+
tokenizer_obj = AutoTokenizer.from_pretrained(MODEL_NAME, use_fast=True)
|
70 |
+
model_obj = AutoModel.from_pretrained(
|
71 |
+
MODEL_NAME, output_hidden_states=True
|
72 |
+
).to(DEVICE).eval()
|
73 |
+
except Exception as e:
|
74 |
+
st.error(f"Hugging Face ๋ชจ๋ธ ({MODEL_NAME}) ๋ก๋ ์ค ์ค๋ฅ: {e}")
|
75 |
+
st.error("์ธํฐ๋ท ์ฐ๊ฒฐ ๋๋ ๋ชจ๋ธ ์ด๋ฆ์ด ์ฌ๋ฐ๋ฅธ์ง ํ์ธํ์ธ์.")
|
76 |
+
return None
|
77 |
+
|
78 |
+
# ํด๋์ค ์ด๋ฆ ๊ฐ์ ธ์ค๊ธฐ ์๋
|
79 |
+
class_names = None
|
80 |
+
if hasattr(lda, 'classes_'): # scikit-learn LDA์ ๊ฒฝ์ฐ
|
81 |
+
class_names = lda.classes_
|
82 |
+
elif hasattr(clf, 'classes_'): # scikit-learn ๋ถ๋ฅ๊ธฐ์ ๊ฒฝ์ฐ
|
83 |
+
class_names = clf.classes_
|
84 |
+
|
85 |
+
return tokenizer_obj, model_obj, W_tensor, mu_vector, w_p_tensor, b_p_vector, class_names
|
86 |
+
|
87 |
+
# โโโโโโโโโโ ํต์ฌ ๋ถ์ ํจ์ (์๋ณธ ์ฝ๋ ๊ธฐ๋ฐ) โโโโโโโโโโ
|
88 |
+
def explain_sentence_streamlit(
|
89 |
+
text: str,
|
90 |
+
tokenizer, model, W, mu, w_p, b_p, # ๋ก๋๋ ๊ฐ์ฒด๋ค
|
91 |
+
layer_id_to_use: int, device_to_use: str, # ์ค์ ๊ฐ
|
92 |
+
top_k_tokens: int = 5
|
93 |
+
) -> tuple[str, int, float, list] | None: # ๊ฒฐ๊ณผ ํ์
๋ช
์ (์คํจ ์ None)
|
94 |
+
"""
|
95 |
+
์
๋ ฅ ๋ฌธ์ฅ์ ์์ธกํ๊ณ ํ ํฐ ์ค์๋๋ฅผ ๊ณ์ฐํ์ฌ ๊ฒฐ๊ณผ๋ฅผ ๋ฐํํฉ๋๋ค.
|
96 |
+
"""
|
97 |
+
try:
|
98 |
+
# 1) ํ ํฐํ (์ต๋ ๊ธธ์ด ๋ฐ ์๋ฆผ ์ฒ๋ฆฌ ์ถ๊ฐ)
|
99 |
+
enc = tokenizer(text, return_tensors="pt", truncation=True, max_length=510, padding=True) # BERT ์ต๋ ๊ธธ์ด 512 ๊ณ ๋ ค, CLS/SEP ๊ณต๊ฐ ํ๋ณด
|
100 |
+
input_ids = enc["input_ids"].to(device_to_use)
|
101 |
+
attn_mask = enc["attention_mask"].to(device_to_use)
|
102 |
+
|
103 |
+
if input_ids.shape[1] == 0: # ์
๋ ฅ์ด ๋๋ฌด ์งง๊ฑฐ๋ ๋ชจ๋ ํํฐ๋ง ๋ ๊ฒฝ์ฐ
|
104 |
+
# Streamlit ์ฑ์์๋ ์ฌ์ฉ์์๊ฒ ๊ฒฝ๊ณ ๋ฅผ ํ์ํ ์ ์์ต๋๋ค.
|
105 |
+
# st.warning("ํ ํฐํ ๊ฒฐ๊ณผ ์ ํจํ ํ ํฐ์ด ์์ต๋๋ค. ๋ค๋ฅธ ๋ฌธ์ฅ์ ์๋ํด๏ฟฝ๏ฟฝ๏ฟฝ์ธ์.")
|
106 |
+
return None
|
107 |
+
|
108 |
+
|
109 |
+
# 2) ์๋ฒ ๋ฉ์ gradient ์ถ์
|
110 |
+
input_embeds = model.embeddings.word_embeddings(input_ids).clone().detach()
|
111 |
+
input_embeds.requires_grad_(True)
|
112 |
+
|
113 |
+
# 3) Forward pass โ CLS ๋ฒกํฐ ์ถ์ถ
|
114 |
+
outputs = model(inputs_embeds=input_embeds,
|
115 |
+
attention_mask=attn_mask, # Attention mask ์ ๋ฌ
|
116 |
+
output_hidden_states=True)
|
117 |
+
cls_vec = outputs.hidden_states[layer_id_to_use][:, 0, :] # (1, 768)
|
118 |
+
|
119 |
+
# 4) LDA ํฌ์ โ ๋ถ๋ฅ logit ๊ณ์ฐ
|
120 |
+
z_projected = (cls_vec - mu) @ W # (1, d)
|
121 |
+
logit_output = z_projected @ w_p.T + b_p # (1, C)
|
122 |
+
|
123 |
+
probs = torch.softmax(logit_output, dim=1)
|
124 |
+
pred_idx = torch.argmax(probs, dim=1).item()
|
125 |
+
pred_prob = probs[0, pred_idx].item()
|
126 |
+
|
127 |
+
# 5) Gradient ๊ณ์ฐ
|
128 |
+
if input_embeds.grad is not None:
|
129 |
+
input_embeds.grad.zero_() # ์ด์ ๊ทธ๋๋์ธํธ ์ด๊ธฐํ
|
130 |
+
logit_output[0, pred_idx].backward() # ์ ํ๋ ์์ธก ํด๋์ค์ ๋ํ ๊ทธ๋๋์ธํธ ๊ณ์ฐ
|
131 |
+
|
132 |
+
if input_embeds.grad is None: # backward ํ์๋ grad๊ฐ ์๋ ์์ธ์ ์ํฉ ๋ฐฉ์ง
|
133 |
+
# st.error("๊ทธ๋๋์ธํธ๋ฅผ ๊ณ์ฐํ ์ ์์ต๋๋ค.") # Streamlit ์ฑ ๋ด์์ ์ค๋ฅ ํ์
|
134 |
+
return None
|
135 |
+
|
136 |
+
grads = input_embeds.grad.clone().detach()
|
137 |
+
|
138 |
+
# 6) Grad ร Input โ ์ค์๋ ์ ์ ๊ณ์ฐ
|
139 |
+
scores = (grads * input_embeds.detach()).norm(dim=2).squeeze(0)
|
140 |
+
scores_np = scores.cpu().numpy()
|
141 |
+
|
142 |
+
# ์ ํจํ ์ ์๋ง์ผ๋ก ์ ๊ทํ (NaN/Inf ๋ฐฉ์ง)
|
143 |
+
valid_scores = scores_np[np.isfinite(scores_np)]
|
144 |
+
if len(valid_scores) > 0 and valid_scores.max() > 0:
|
145 |
+
scores_np = scores_np / (valid_scores.max() + 1e-9) # 0~1 ์ ๊ทํ
|
146 |
+
else: # ๋ชจ๋ ์ ์๊ฐ 0์ด๊ฑฐ๋ ์ ํจํ์ง ์์ ๊ฒฝ์ฐ
|
147 |
+
scores_np = np.zeros_like(scores_np)
|
148 |
+
|
149 |
+
|
150 |
+
# 7) HTML ํ์ด๋ผ์ดํธ ์์ฑ
|
151 |
+
tokens = tokenizer.convert_ids_to_tokens(input_ids[0], skip_special_tokens=False) # ์คํ์
ํ ํฐ ํฌํจ
|
152 |
+
html_tokens_list = []
|
153 |
+
|
154 |
+
# CLS, SEP, PAD ํ ํฐ ID ํ์ธ
|
155 |
+
cls_token_id = tokenizer.cls_token_id
|
156 |
+
sep_token_id = tokenizer.sep_token_id
|
157 |
+
pad_token_id = tokenizer.pad_token_id
|
158 |
+
|
159 |
+
for i, tok_str in enumerate(tokens):
|
160 |
+
if input_ids[0, i] == pad_token_id: # PAD ํ ํฐ์ ๊ฑด๋๋ฐ๊ธฐ
|
161 |
+
continue
|
162 |
+
|
163 |
+
clean_tok_str = tok_str.replace("##", "") if "##" not in tok_str else tok_str[2:]
|
164 |
+
|
165 |
+
# ์คํ์
ํ ํฐ์ ๋ค๋ฅธ ์คํ์ผ ์ ์ฉ ๋๋ ์ค์๋ ๊ณ์ฐ์์ ์ ์ธ ๊ฐ๋ฅ
|
166 |
+
if input_ids[0, i] == cls_token_id or input_ids[0, i] == sep_token_id:
|
167 |
+
html_tokens_list.append(f"<span style='font-weight:bold;'>{html.escape(clean_tok_str)}</span>")
|
168 |
+
else:
|
169 |
+
score_val = scores_np[i] if i < len(scores_np) else 0 # ์ ์ ๋ฐฐ์ด ๋ฒ์ ํ์ธ
|
170 |
+
color = f"rgba(255, 0, 0, {max(0, min(1, score_val)):.2f})" # ์ ์ ๋ฒ์ 0~1๋ก ํด๋ฆฌํ
|
171 |
+
html_tokens_list.append(
|
172 |
+
f"<span style='background-color:{color}; padding: 1px 2px; margin: 1px; border-radius: 3px; display:inline-block;'>{html.escape(clean_tok_str)}</span>"
|
173 |
+
)
|
174 |
+
|
175 |
+
html_output_str = " ".join(html_tokens_list)
|
176 |
+
# ๋ถํ์ํ ๊ณต๋ฐฑ ์ ๋ฆฌ (์: subword ์ฌ์ด ๊ณต๋ฐฑ)
|
177 |
+
html_output_str = html_output_str.replace(" ##", "")
|
178 |
+
|
179 |
+
# Top-K ์ค์ ํ ํฐ ์ ๋ณด (์คํ์
ํ ํฐ ๋ฐ PAD ํ ํฐ ์ ์ธ)
|
180 |
+
top_tokens_info_list = []
|
181 |
+
valid_indices_for_top_k = [
|
182 |
+
idx for idx, token_id in enumerate(input_ids[0].tolist())
|
183 |
+
if token_id not in [cls_token_id, sep_token_id, pad_token_id] and idx < len(scores_np)
|
184 |
+
]
|
185 |
+
|
186 |
+
# ์ ์๊ฐ ๋์ ์์ผ๋ก ์ ๋ ฌ
|
187 |
+
sorted_valid_indices = sorted(valid_indices_for_top_k, key=lambda idx: -scores_np[idx])
|
188 |
+
|
189 |
+
for token_idx in sorted_valid_indices[:top_k_tokens]:
|
190 |
+
top_tokens_info_list.append({
|
191 |
+
"token": tokens[token_idx],
|
192 |
+
"score": f"{scores_np[token_idx]:.3f}"
|
193 |
+
})
|
194 |
+
|
195 |
+
return html_output_str, pred_idx, pred_prob, top_tokens_info_list
|
196 |
+
|
197 |
+
except Exception as e:
|
198 |
+
# Streamlit ์ฑ ๋ด์์ ์ค๋ฅ๋ฅผ ๋ ์ ํ์ํ๋๋ก ์์
|
199 |
+
# st.error(f"๋ฌธ์ฅ ๋ถ์ ์ค ์๊ธฐ์น ์์ ์ค๋ฅ ๋ฐ์: {e}")
|
200 |
+
# import traceback
|
201 |
+
# st.text_area("์ค๋ฅ ์์ธ ์ ๋ณด (๋๋ฒ๊น
์ฉ):", traceback.format_exc(), height=200)
|
202 |
+
# print(f"๋ฌธ์ฅ ๋ถ์ ์ค ์๊ธฐ์น ์์ ์ค๋ฅ ๋ฐ์: {e}") # ์ฝ์ ๋ก๊น
(Spaces ๋ก๊ทธ์์ ํ์ธ ๊ฐ๋ฅ)
|
203 |
+
# import traceback
|
204 |
+
# print(traceback.format_exc()) # ์ฝ์ ๋ก๊น
|
205 |
+
raise # ์ค๋ฅ๋ฅผ ๋ค์ ๋ฐ์์์ผ Streamlit์ด ์ฒ๋ฆฌํ๋๋ก ํ๊ฑฐ๋, ์๋์์ None์ ๋ฐํ
|
206 |
+
# return None
|
207 |
+
|
208 |
+
|
209 |
+
# โโโโโโโโโโ Streamlit UI ๊ตฌ์ฑ โโโโโโโโโโ
|
210 |
+
st.set_page_config(page_title="๋ฌธ์ฅ ํ ํฐ ์ค์๋ ๋ถ์๊ธฐ", layout="wide")
|
211 |
+
st.title("๐ ๋ฌธ์ฅ ํ ํฐ ์ค์๋ ๋ถ์๊ธฐ")
|
212 |
+
st.markdown("BERT์ LDA๋ฅผ ํ์ฉํ์ฌ ๋ฌธ์ฅ ๋ด ๊ฐ ํ ํฐ์ ์ค์๋๋ฅผ ์๊ฐํํฉ๋๋ค.")
|
213 |
+
|
214 |
+
# ๋ชจ๋ธ ๋ก๋ ์๋
|
215 |
+
loaded_data_tuple = load_all_models_and_data()
|
216 |
+
|
217 |
+
if loaded_data_tuple:
|
218 |
+
tokenizer, model, W, mu, w_p, b_p, class_names = loaded_data_tuple
|
219 |
+
|
220 |
+
# ์ฌ์ด๋๋ฐ์ ๋ชจ๋ธ ์ ๋ณด ํ์
|
221 |
+
st.sidebar.header("โ๏ธ ๋ชจ๋ธ ๋ฐ ์ค์ ์ ๋ณด")
|
222 |
+
st.sidebar.info(f"**BERT ๋ชจ๋ธ:** `{MODEL_NAME}`\n\n"
|
223 |
+
f"**์ฌ์ฉ๋ ๋ ์ด์ด ID:** `{LAYER_ID}`\n\n"
|
224 |
+
f"**๋ถ๋ฅ๊ธฐ ์ข
๋ฅ:** `{CLF_NAME}` (LDA ํฌ์ ๊ธฐ๋ฐ)\n\n"
|
225 |
+
f"**์คํ ์ฅ์น:** `{DEVICE.upper()}`")
|
226 |
+
if class_names is not None:
|
227 |
+
st.sidebar.markdown(f"**์์ธก ๊ฐ๋ฅ ํด๋์ค:** `{', '.join(map(str, class_names))}`")
|
228 |
+
|
229 |
+
|
230 |
+
# ์ฌ์ฉ์ ์
๋ ฅ
|
231 |
+
st.subheader("๐ ๋ถ์ํ ์์ด ๋ฌธ์ฅ์ ์
๋ ฅํ์ธ์:")
|
232 |
+
user_sentence = st.text_area("๋ฌธ์ฅ ์
๋ ฅ:", "This movie is exceptionally good and I highly recommend it.", height=100)
|
233 |
+
|
234 |
+
top_k_slider = st.slider("ํ์ํ Top-K ์ค์ ํ ํฐ ์:", min_value=1, max_value=10, value=5, step=1)
|
235 |
+
|
236 |
+
if st.button("๋ถ์ ์คํํ๊ธฐ ๐", type="primary"):
|
237 |
+
if user_sentence:
|
238 |
+
with st.spinner("๋ฌธ์ฅ์ ๋ถ์ํ๊ณ ์์ต๋๋ค... ์กฐ๊ธ๋ง ๊ธฐ๋ค๋ ค์ฃผ์ธ์...โณ"):
|
239 |
+
analysis_results = None
|
240 |
+
try:
|
241 |
+
analysis_results = explain_sentence_streamlit(
|
242 |
+
user_sentence, tokenizer, model, W, mu, w_p, b_p,
|
243 |
+
LAYER_ID, DEVICE, top_k_tokens=top_k_slider
|
244 |
+
)
|
245 |
+
except Exception as e: # explain_sentence_streamlit ๋ด๋ถ์์ raise๋ ์ค๋ฅ ์ฒ๋ฆฌ
|
246 |
+
st.error(f"๋ถ์ ์ฒ๋ฆฌ ์ค ์ค๋ฅ ๋ฐ์: {e}")
|
247 |
+
st.info("์
๋ ฅ ๋ฌธ์ฅ์ด๋ ๋ชจ๋ธ ํธํ์ฑ ๋ฌธ์ ๋ฅผ ํ์ธํด๋ณด์ธ์. ๋ฌธ์ ๊ฐ ์ง์๋๋ฉด ๊ด๋ฆฌ์์๊ฒ ๋ฌธ์ํ์ธ์.")
|
248 |
+
# ๋ ์์ธํ ์ค๋ฅ๋ Spaces์ ๋ก๊ทธ์์ ํ์ธ ๊ฐ๋ฅ (print๋ฌธ ์ฌ์ฉ ์)
|
249 |
+
|
250 |
+
|
251 |
+
if analysis_results: # ์ฑ๊ณต์ ์ผ๋ก ๊ฒฐ๊ณผ ๋ฐํ ์
|
252 |
+
html_viz, predicted_idx, probability, top_k_list = analysis_results
|
253 |
+
|
254 |
+
st.markdown("---")
|
255 |
+
st.subheader("๐ ๋ถ์ ๊ฒฐ๊ณผ")
|
256 |
+
|
257 |
+
predicted_class_label = str(predicted_idx) # ๊ธฐ๋ณธ๊ฐ: ์ธ๋ฑ์ค
|
258 |
+
if class_names is not None and 0 <= predicted_idx < len(class_names):
|
259 |
+
predicted_class_label = str(class_names[predicted_idx]) # ํด๋์ค ์ด๋ฆ ์ฌ์ฉ
|
260 |
+
|
261 |
+
st.success(f"**์์ธก๋ ํด๋์ค:** **`{predicted_class_label}`** (์ ๋ขฐ๋: **{probability:.2f}**)")
|
262 |
+
|
263 |
+
st.subheader("๐จ ํ ํฐ๋ณ ์ค์๋ ์๊ฐํ")
|
264 |
+
st.markdown(html_viz, unsafe_allow_html=True)
|
265 |
+
|
266 |
+
st.subheader(f"โญ Top-{top_k_slider} ์ค์ ํ ํฐ")
|
267 |
+
if top_k_list:
|
268 |
+
cols = st.columns(len(top_k_list) if len(top_k_list) <=5 else 5 ) # ํ ์ค์ ์ต๋ 5๊ฐ
|
269 |
+
for i, item in enumerate(top_k_list):
|
270 |
+
with cols[i % len(cols)]:
|
271 |
+
st.metric(label=item['token'], value=item['score'])
|
272 |
+
else:
|
273 |
+
st.info("์ค์๋ ๋์ ํ ํฐ์ ์ฐพ์ ์ ์์ต๋๋ค (์คํ์
ํ ํฐ ๋ฑ ์ ์ธ).")
|
274 |
+
# 'analysis_results is None' ์ด๊ณ ์์ธ์ฒ๋ฆฌ๋ก st.error๊ฐ ์ด๋ฏธ ํ์๋ ๊ฒฝ์ฐ๋ ์ถ๊ฐ ๋ฉ์์ง ๋ถํ์
|
275 |
+
elif analysis_results is None and not user_sentence: # ๋ฌธ์ฅ ์
๋ ฅ ์์ด ๋ฒํผ ๋๋ฅธ ๊ฒฝ์ฐ (์ฌ์ค์ ์์์ ์ฒ๋ฆฌ)
|
276 |
+
pass # ์ด๋ฏธ st.warning์ผ๋ก ์ฒ๋ฆฌ๋จ
|
277 |
+
|
278 |
+
else: # ๋ฌธ์ฅ ์
๋ ฅ ์์ด ๋ฒํผ ๋๋ฅธ ๊ฒฝ์ฐ
|
279 |
+
st.warning("๋ถ์ํ ๋ฌธ์ฅ์ ์
๋ ฅํด์ฃผ์ธ์.")
|
280 |
+
else:
|
281 |
+
st.error("๋ชจ๋ธ ๋ก๋ฉ์ ์คํจํ์ฌ ์ ํ๋ฆฌ์ผ์ด์
์ ์์ํ ์ ์์ต๋๋ค. ์
๋ก๋๋ ํ์ผ๊ณผ ๊ฒฝ๋ก ์ค์ ์ ํ์ธํด์ฃผ์ธ์. Hugging Face Spaces์ 'Logs' ํญ์์ ์์ธ ์ค๋ฅ๋ฅผ ํ์ธํ ์ ์์ต๋๋ค.")
|
282 |
+
|
283 |
+
st.markdown("---")
|
284 |
+
st.markdown("<p style='text-align: center; color: grey;'>BERT ๊ธฐ๋ฐ ๋ฌธ์ฅ ๋ถ์ ๋ฐ๋ชจ</p>", unsafe_allow_html=True)
|