layout

run.py

@@ -1,7 +1,7 @@
 import gradio as gr
 from huggingface_hub import list_models
 from datasets import load_dataset
-from type.dataset_type import AutoGeneratedJapaneseQAData
+from type.dataset_type import TanukiPhase2AnnotationDataset
 
 js = """
 function blockEnter(event) {
@@ -16,11 +16,17 @@ function blockEnter(event) {
 }
 """
 
+target_dataset = TanukiPhase2AnnotationDataset()
 
-def hello(profile: gr.OAuthProfile | None) -> str:
+def load_css():
+    with open("style.css", "r") as file:
+        css_content = file.read()
+    return css_content
+
+def set_name(profile: gr.OAuthProfile | None) -> str:
     if profile is None:
         return "プライベートデータセット取得のためにログインしてください。"
-    return f"{profile.username}さん、よろしくお願いいたします。"
+    return f'{profile.username}さん、よろしくお願いいたします。'
 
 
 def list_private_models(profile: gr.OAuthProfile | None, oauth_token: gr.OAuthToken | None) -> str:
@@ -59,39 +65,60 @@ def display_dataset(profile: gr.OAuthProfile | None, oauth_token: gr.OAuthToken
     else:
         return gr.update(visible=True, value="データセットのロードに失敗しました。"), None, None
 
+def switch_theme(theme):
+    if theme == "Dark":
+        return gr.themes.Default()
+    else:
+        return gr.themes.Monochrome()
+
+theme_ = gr.State("Light")
+
+with gr.Blocks(theme=theme_, css=load_css()) as demo:
 
-with gr.Blocks() as demo:
     gr.Markdown("# 自動生成データセット アノテーション for Tanuki 2Phase")
 
-    with gr.Row(equal_height=True):
+    with gr.Tab("アノテーション"):
+
+        def update_theme():
+            new_theme = "Dark" if theme_.value == "Light" else "Light"
+            theme_.value = new_theme
+            return switch_theme(new_theme)
+
+        with gr.Row(equal_height=True):
+
+            login_btn = gr.LoginButton(visible=True, scale=1)
+            logout_btn = gr.LogoutButton(visible=False, scale=1)
+
+            profile = gr.Markdown()
+            data_load_btn = gr.Button("データセットを読み込む")
+            theme_button = gr.Button("Switch Theme")
+            theme_button.click(fn=update_theme, outputs=None)
+
+            demo.load(set_name, inputs=None, outputs=profile)
+            demo.load(toggle_buttons, inputs=None, outputs=[login_btn, logout_btn])
+
+        dataset_display = gr.Markdown(visible=False)
+        question_text = gr.Textbox(label="質問: ", interactive=False)
 
-        login_btn = gr.LoginButton(visible=True, scale=1)
-        logout_btn = gr.LogoutButton(visible=False, scale=1)
+        with gr.Row(equal_height=True):
+            good_btn = gr.Button("良い")
+            bad_btn = gr.Button("悪い")
 
-        m1 = gr.Markdown()
-        # m2 = gr.Markdown(scale=1)
-        data_load_btn = gr.Button("データセットを読み込む")
+        answer_text = gr.Textbox(label="回答: 改行はShift+Enterです。", elem_id="answer", interactive=True)
 
-    demo.load(hello, inputs=None, outputs=m1)
-    # demo.load(list_private_models, inputs=None, outputs=m2)
-    demo.load(toggle_buttons, inputs=None, outputs=[login_btn, logout_btn])
+        data_load_btn.click(
+            display_dataset,
+            inputs=None,
+            outputs=[dataset_display, question_text, answer_text],
+        )
 
-    with gr.Row(equal_height=True):
-        good_btn = gr.Button("良い")
-        bad_btn = gr.Button("悪い")
+        def on_submit(answer_text):
+            return f" {answer_text}"
 
-    dataset_display = gr.Markdown(visible=False)
-    question_text = gr.Textbox(label="質問: ", interactive=False)
-    answer_text = gr.Textbox(label="回答: 改行はShift+Enterです。", elem_id="answer", interactive=True)
 
-    data_load_btn.click(
-        display_dataset,
-        inputs=None,
-        outputs=[dataset_display, question_text, answer_text],
-    )
+    with gr.Tab("アノテ済みデータセット"):
 
-    def on_submit(answer_text):
-        return f" {answer_text}"
+        gr.Textbox("データセットID", lines=1, placeholder="データセットIDを入力してください。")
 
 
 if __name__ == "__main__":

script/custom.js

@@ -1,4 +1,3 @@
 function prevent (event)  {
-    alert('Hello, World!');
-    
-};
+    // alert('Hello, World!');
+};

target_dataset.txt

@@ -0,0 +1,70 @@
+## hatakeyama-llm-team/WikiBookJa
+fileSize 29.3MB
+- index int64
+- text string
+
+## hatakeyama-llm-team/AutoGeneratedJapaneseQA
+fileSize 33.1MB
+- question string
+- database string
+- answer string
+- score float64
+
+## hatakeyama-llm-team/AutoGeneratedJapaneseQA-other
+fileSize 12.9MB
+- question string
+- database string
+- answer string
+- score float64
+
+## kanhatakeyama/AutoWikiQA
+fileSize 232MB
+- question string
+- inst_question string
+- inst_answer_0 string
+- text string
+- database string
+- answer string
+- answer_0 string
+- answer_1 string
+
+## kanhatakeyama/ChatbotArenaJaMixtral8x22b
+fileSize 13.9MB
+- question string
+- inst_question string
+- inst_answer_0 string
+- text string
+- database string
+- answer string
+
+# kanhatakeyama/OrcaJaMixtral8x22b
+fileSize 1.24GB
+- question string
+- inst_question string
+- inst_answer_0 string
+- text string
+- database string
+- answer string
+
+# kanhatakeyama/AutoMultiTurnByMixtral8x22b
+fileSize 270MB
+- q1 string
+- a1 string
+- q2 string
+- a2 string
+- database string
+
+# kanhatakeyama/LogicalDatasetsByMixtral8x22b
+fileSize 116MB
+- question string
+- answer string
+
+## output_dataset
+- id int64
+- dataset string
+- dataset_id int64
+- who string
+- good bool
+- bad bool
+- is_proofreading string
+- t_proofreading string

type/dataset_type.py

@@ -1,10 +1,66 @@
 # データセットのファイル構造に合わせて型定義
-class WikiBookJaData:
+
+class WikiBookJa:
   index: int
   text: str
 
-class AutoGeneratedJapaneseQAData:
+class AutoGeneratedJapaneseQA:
+  question: str
+  database: str
+  answer: str
+  score: float
+
+class AutoGeneratedJapaneseQA_other:
+  question: str
+  database: str
+  answer: str
+  score: float
+
+class AutoWikiQA:
   question: str
+  answer: str
+  inst_question: str
+  inst_answer_0: str
+  text : str
   database: str
   answer: str
-  score: float
+  answer_0: str
+  answer_1: str
+
+class ChatbotArenaJaMixtral8x22b:
+  question: str
+  inst_question: str
+  inst_answer_0: str
+  text: str
+  database: str
+  answer: str
+
+class OrcaJaMixtral8x22b:
+  question: str
+  inst_question: str
+  inst_answer_0: str
+  text: str
+  database: str
+  answer: str
+
+class AutoMultiTurnByMixtral8x22b:
+  q1: str
+  a1: str
+  q2: str
+  a2: str
+  database: str
+
+class LogicalDatasetsByMixtral8x22b:
+  question: str
+  answer: str
+
+
+class TanukiPhase2AnnotationDataset:
+  id: int
+  dataset: str
+  database_id: int
+  who: str
+  good: bool
+  bad: bool
+  is_proofreading: bool
+  t_proofreading: str