Spaces:
Running
on
Zero
Running
on
Zero
add suggestion cleanning
Browse files
app.py
CHANGED
@@ -1,10 +1,9 @@
|
|
1 |
-
# app.py
|
2 |
import spaces
|
3 |
import gradio as gr
|
4 |
from gradio import update
|
5 |
from functools import lru_cache
|
6 |
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
7 |
-
from opencc import OpenCC #
|
8 |
|
9 |
# 初始化簡體到繁體轉換器
|
10 |
cc = OpenCC('s2t')
|
@@ -26,30 +25,19 @@ MODEL_LIST = [
|
|
26 |
]
|
27 |
|
28 |
def merge_common_prefixes(suggestions, min_len=2):
|
29 |
-
"""
|
30 |
-
合併具有共同前綴的建議:
|
31 |
-
- 找出所有長度 ≥ min_len 的共同前綴
|
32 |
-
- 將這些前綴作為新建議,移除原有被合併的項目
|
33 |
-
"""
|
34 |
prefixes = []
|
35 |
to_remove = set()
|
36 |
-
|
37 |
for i in range(len(suggestions)):
|
38 |
for j in range(i+1, len(suggestions)):
|
39 |
s1, s2 = suggestions[i], suggestions[j]
|
40 |
-
# 計算字元級共同前綴
|
41 |
common = ''.join(c1 for c1, c2 in zip(s1, s2) if c1 == c2)
|
42 |
if len(common) >= min_len:
|
43 |
prefixes.append(common)
|
44 |
to_remove.update([s1, s2])
|
45 |
-
|
46 |
-
# 去重前綴
|
47 |
unique_prefixes = []
|
48 |
for p in prefixes:
|
49 |
if p not in unique_prefixes:
|
50 |
unique_prefixes.append(p)
|
51 |
-
|
52 |
-
# 剩下沒有被合併的建議
|
53 |
remainder = [s for s in suggestions if s not in to_remove]
|
54 |
return unique_prefixes + remainder
|
55 |
|
@@ -64,14 +52,7 @@ def get_pipeline(model_name):
|
|
64 |
|
65 |
@spaces.GPU
|
66 |
def suggest_next(text, model_name, k, m, num_beam_groups, diversity_penalty):
|
67 |
-
"""
|
68 |
-
使用 Diverse Beam Search 產生 m 條候選:
|
69 |
-
- num_beams = m
|
70 |
-
- num_beam_groups, diversity_penalty 可調整多樣性
|
71 |
-
之後轉繁體、去重、合併共同前綴後回傳。
|
72 |
-
"""
|
73 |
gen_pipe = get_pipeline(model_name)
|
74 |
-
# 構造 generate 參數字典,僅在 penalty>0 時加入 diversity 相關
|
75 |
gen_kwargs = {
|
76 |
"max_new_tokens": k,
|
77 |
"num_beams": m,
|
@@ -81,110 +62,54 @@ def suggest_next(text, model_name, k, m, num_beam_groups, diversity_penalty):
|
|
81 |
}
|
82 |
if diversity_penalty and diversity_penalty > 0:
|
83 |
gen_kwargs["num_beam_groups"] = num_beam_groups
|
84 |
-
gen_kwargs["diversity_penalty"] =
|
85 |
|
86 |
outs = gen_pipe(text, **gen_kwargs)
|
87 |
|
88 |
-
#
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
if
|
93 |
-
|
|
|
|
|
94 |
|
95 |
-
# 去重
|
96 |
unique_suggestions = []
|
97 |
-
|
98 |
-
|
99 |
-
|
|
|
|
|
|
|
100 |
|
101 |
# 合併共同前綴
|
102 |
-
|
103 |
|
104 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
105 |
|
|
|
106 |
|
107 |
-
def append_suggestion(
|
108 |
-
|
109 |
-
return current
|
110 |
-
# 直接插入選中的候選文字
|
111 |
-
return current + choice
|
112 |
|
113 |
-
|
114 |
-
|
115 |
-
#suggestions-bar {
|
116 |
-
width: 100%;
|
117 |
-
margin-bottom: 8px;
|
118 |
-
}
|
119 |
#suggestions-bar .candidate-list {
|
120 |
-
display: flex;
|
121 |
-
|
122 |
-
|
123 |
-
border: 1px solid #999;
|
124 |
-
border-radius: 4px;
|
125 |
-
padding: 6px;
|
126 |
-
overflow-x: auto;
|
127 |
-
white-space: nowrap;
|
128 |
-
}
|
129 |
-
#suggestions-bar .candidate-list label {
|
130 |
-
cursor: pointer;
|
131 |
-
padding: 6px 10px;
|
132 |
-
font-size: 16px;
|
133 |
-
}
|
134 |
-
#suggestions-bar .candidate-list label:hover {
|
135 |
-
background: #f5f5f5;
|
136 |
-
}
|
137 |
-
#suggestions-bar .candidate-list input[type=radio]:checked + label {
|
138 |
-
background: #e6f7ff;
|
139 |
-
border: 1px solid #1890ff;
|
140 |
-
}
|
141 |
-
#input-box textarea {
|
142 |
-
width: 100%;
|
143 |
-
font-size: 16px;
|
144 |
-
padding: 6px;
|
145 |
-
box-sizing: border-box;
|
146 |
-
overflow: hidden;
|
147 |
-
resize: none;
|
148 |
}
|
149 |
-
#
|
150 |
-
|
151 |
-
width: 100%;
|
152 |
-
}
|
153 |
-
/* 手機響應式 */
|
154 |
-
@media only screen and (max-width: 600px) {
|
155 |
-
#suggestions-bar .candidate-list label {
|
156 |
-
padding: 8px;
|
157 |
-
font-size: 18px;
|
158 |
-
}
|
159 |
-
#predict-button {
|
160 |
-
font-size: 18px;
|
161 |
-
}
|
162 |
-
}
|
163 |
-
"""
|
164 |
-
|
165 |
-
# 自動增高腳本
|
166 |
-
auto_height_js = """
|
167 |
-
<script>
|
168 |
-
window.addEventListener('load', () => {
|
169 |
-
const textarea = document.querySelector('#input-box textarea');
|
170 |
-
if (!textarea) return;
|
171 |
-
textarea.style.height = 'auto';
|
172 |
-
textarea.addEventListener('input', function() {
|
173 |
-
this.style.height = 'auto';
|
174 |
-
this.style.height = this.scrollHeight + 'px';
|
175 |
-
});
|
176 |
-
});
|
177 |
-
</script>
|
178 |
-
"""
|
179 |
-
|
180 |
-
with gr.Blocks(css=custom_css) as demo:
|
181 |
-
gr.HTML(auto_height_js)
|
182 |
-
gr.Markdown(
|
183 |
-
"## 🇹🇼 繁體中文 IME 加速器 \
|
184 |
-
"
|
185 |
-
"結合小型語言模型與 ZeroGPU,提供即時輸入法風格候選欄。"
|
186 |
-
)
|
187 |
-
|
188 |
with gr.Column():
|
189 |
suggestions = gr.Radio(
|
190 |
[], label="", interactive=True, type="value",
|
@@ -195,14 +120,11 @@ with gr.Blocks(css=custom_css) as demo:
|
|
195 |
lines=1, max_lines=20, elem_id="input-box"
|
196 |
)
|
197 |
|
198 |
-
# 永遠顯示預測按鈕
|
199 |
with gr.Row():
|
200 |
auto_predict = gr.Checkbox(
|
201 |
value=True, label="自動預測(內容變更時觸發)", elem_id="auto-predict"
|
202 |
)
|
203 |
-
predict_button = gr.Button(
|
204 |
-
"預測", elem_id="predict-button"
|
205 |
-
)
|
206 |
|
207 |
with gr.Accordion("進階設定", open=False):
|
208 |
model_selector = gr.Dropdown(
|
@@ -215,15 +137,14 @@ with gr.Blocks(css=custom_css) as demo:
|
|
215 |
minimum=1, maximum=30, step=1, value=30, label="M(建議數/Beam 數)"
|
216 |
)
|
217 |
group_slider = gr.Slider(
|
218 |
-
minimum=1, maximum=30, step=1, value=
|
219 |
label="Beam 群組數 (num_beam_groups)"
|
220 |
)
|
221 |
diversity_penalty_slider = gr.Slider(
|
222 |
-
minimum=0.0, maximum=2.0, step=0.1, value=
|
223 |
label="多樣性懲罰 (diversity_penalty)"
|
224 |
)
|
225 |
|
226 |
-
# 綁定事件
|
227 |
predict_button.click(
|
228 |
fn=suggest_next,
|
229 |
inputs=[
|
@@ -258,4 +179,4 @@ with gr.Blocks(css=custom_css) as demo:
|
|
258 |
outputs=input_text,
|
259 |
)
|
260 |
|
261 |
-
demo.launch()
|
|
|
|
|
1 |
import spaces
|
2 |
import gradio as gr
|
3 |
from gradio import update
|
4 |
from functools import lru_cache
|
5 |
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
6 |
+
from opencc import OpenCC # 用於簡體到繁體轉換
|
7 |
|
8 |
# 初始化簡體到繁體轉換器
|
9 |
cc = OpenCC('s2t')
|
|
|
25 |
]
|
26 |
|
27 |
def merge_common_prefixes(suggestions, min_len=2):
|
|
|
|
|
|
|
|
|
|
|
28 |
prefixes = []
|
29 |
to_remove = set()
|
|
|
30 |
for i in range(len(suggestions)):
|
31 |
for j in range(i+1, len(suggestions)):
|
32 |
s1, s2 = suggestions[i], suggestions[j]
|
|
|
33 |
common = ''.join(c1 for c1, c2 in zip(s1, s2) if c1 == c2)
|
34 |
if len(common) >= min_len:
|
35 |
prefixes.append(common)
|
36 |
to_remove.update([s1, s2])
|
|
|
|
|
37 |
unique_prefixes = []
|
38 |
for p in prefixes:
|
39 |
if p not in unique_prefixes:
|
40 |
unique_prefixes.append(p)
|
|
|
|
|
41 |
remainder = [s for s in suggestions if s not in to_remove]
|
42 |
return unique_prefixes + remainder
|
43 |
|
|
|
52 |
|
53 |
@spaces.GPU
|
54 |
def suggest_next(text, model_name, k, m, num_beam_groups, diversity_penalty):
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
gen_pipe = get_pipeline(model_name)
|
|
|
56 |
gen_kwargs = {
|
57 |
"max_new_tokens": k,
|
58 |
"num_beams": m,
|
|
|
62 |
}
|
63 |
if diversity_penalty and diversity_penalty > 0:
|
64 |
gen_kwargs["num_beam_groups"] = num_beam_groups
|
65 |
+
gen_kwargs["diversity_penalty"] = diversity_penalty
|
66 |
|
67 |
outs = gen_pipe(text, **gen_kwargs)
|
68 |
|
69 |
+
# 提取純下文、過濾空字串、繁體化、確保 strip 處理
|
70 |
+
raw_suggestions = []
|
71 |
+
for out in outs:
|
72 |
+
snippet = out["generated_text"][len(text):].strip()
|
73 |
+
if not snippet:
|
74 |
+
continue
|
75 |
+
converted = cc.convert(snippet).strip()
|
76 |
+
raw_suggestions.append(converted)
|
77 |
|
78 |
+
# 去重 (基於 strip 後內容)
|
79 |
unique_suggestions = []
|
80 |
+
seen = set()
|
81 |
+
for s in raw_suggestions:
|
82 |
+
key = s
|
83 |
+
if key not in seen:
|
84 |
+
seen.add(key)
|
85 |
+
unique_suggestions.append(key)
|
86 |
|
87 |
# 合併共同前綴
|
88 |
+
merged_prefixes = merge_common_prefixes(unique_suggestions, min_len=2)
|
89 |
|
90 |
+
# 最終去重並移除空項 (基於 strip 後內容)
|
91 |
+
final_suggestions = []
|
92 |
+
seen_final = set()
|
93 |
+
for s in merged_prefixes:
|
94 |
+
key = s.strip()
|
95 |
+
if key and key not in seen_final:
|
96 |
+
seen_final.add(key)
|
97 |
+
final_suggestions.append(key)
|
98 |
|
99 |
+
return update(choices=final_suggestions, value=None)
|
100 |
|
101 |
+
def append_suggestion(text, choice):
|
102 |
+
return text + choice
|
|
|
|
|
|
|
103 |
|
104 |
+
with gr.Blocks(css="""
|
105 |
+
#suggestions-bar { width: 100%; margin-bottom: 8px; }
|
|
|
|
|
|
|
|
|
106 |
#suggestions-bar .candidate-list {
|
107 |
+
display: flex; gap: 8px; background: #fff;
|
108 |
+
border: 1px solid #999; border-radius: 4px;
|
109 |
+
padding: 6px; overflow-x: auto; white-space: nowrap;
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
110 |
}
|
111 |
+
#suggestions-bar .candidate-list label { cursor: pointer; }
|
112 |
+
""") as demo:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
113 |
with gr.Column():
|
114 |
suggestions = gr.Radio(
|
115 |
[], label="", interactive=True, type="value",
|
|
|
120 |
lines=1, max_lines=20, elem_id="input-box"
|
121 |
)
|
122 |
|
|
|
123 |
with gr.Row():
|
124 |
auto_predict = gr.Checkbox(
|
125 |
value=True, label="自動預測(內容變更時觸發)", elem_id="auto-predict"
|
126 |
)
|
127 |
+
predict_button = gr.Button("預測", elem_id="predict-button")
|
|
|
|
|
128 |
|
129 |
with gr.Accordion("進階設定", open=False):
|
130 |
model_selector = gr.Dropdown(
|
|
|
137 |
minimum=1, maximum=30, step=1, value=30, label="M(建議數/Beam 數)"
|
138 |
)
|
139 |
group_slider = gr.Slider(
|
140 |
+
minimum=1, maximum=30, step=1, value=5,
|
141 |
label="Beam 群組數 (num_beam_groups)"
|
142 |
)
|
143 |
diversity_penalty_slider = gr.Slider(
|
144 |
+
minimum=0.0, maximum=2.0, step=0.1, value=0.3,
|
145 |
label="多樣性懲罰 (diversity_penalty)"
|
146 |
)
|
147 |
|
|
|
148 |
predict_button.click(
|
149 |
fn=suggest_next,
|
150 |
inputs=[
|
|
|
179 |
outputs=input_text,
|
180 |
)
|
181 |
|
182 |
+
demo.launch()
|