update logo

README.md

@@ -17,8 +17,12 @@ hf_oauth_scopes:
 - manage-repos
 - inference-api
 ---
-
-<img src="https://raw.githubusercontent.com/argilla-io/synthetic-data-generator/main/assets/logo-sdg.svg" alt="Synthetic Data Generator Logo" style="display: block; margin-left: auto; margin-right: auto; width: 75%; margin-bottom: -200px;"/>
+<p align="center">
+<img src="https://raw.githubusercontent.com/argilla-io/synthetic-data-generator/main/assets/logo.svg" alt="Synthetic Data Generator Logo" style="width: 80%;"/>
+</p>
+<p align="center">
+  <h3>Build datasets using natural language</h3>
+</p>
 
 ![Synthetic Data Generator](https://huggingface.co/spaces/argilla/synthetic-data-generator/resolve/main/assets/ui-full.png)
 

src/synthetic_dataset_generator/_tabbedinterface.py

@@ -8,7 +8,6 @@ from collections.abc import Sequence
 
 import gradio as gr
 from gradio.blocks import Blocks
-from gradio.components import HTML
 from gradio.layouts import Tab, Tabs
 from gradio.themes import ThemeClass as Theme
 from gradio_client.documentation import document
@@ -61,16 +60,11 @@ class TabbedInterface(Blocks):
             tab_names = [f"Tab {i}" for i in range(len(interface_list))]
         with self:
             if title:
-                HTML(value=title)
-                with gr.Row():
-                    with gr.Column(scale=2):
-                        gr.Markdown("### Build datasets using natural language")
-                    with gr.Column(scale=3):
-                        pass
-                    with gr.Column(scale=2):
-                        gr.LoginButton(
-                            value="Sign in", variant="primary", scale=2
-                        )
+                gr.HTML(value=title)
+            gr.HTML(
+                "<div style='text-align: center;'><h3>Build datasets using natural language</h3></div>"
+            )
+            gr.LoginButton(value="Sign in", variant="primary", scale=2)
             with Tabs():
                 for interface, tab_name in zip(interface_list, tab_names, strict=False):
                     with Tab(label=tab_name):

src/synthetic_dataset_generator/app.py

@@ -10,11 +10,13 @@ css = """
 .main_ui_logged_out{opacity: 0.3; pointer-events: none}
 """
 
+image = """<img src="https://raw.githubusercontent.com/argilla-io/synthetic-data-generator/main/assets/logo-sdg.svg" alt="Synthetic Data Generator Logo" style="display: block; margin-left: auto; margin-right: auto; width: 75%; margin-bottom: -400px;"/>"""
+
 demo = TabbedInterface(
     [textcat_app, sft_app, eval_app, faq_app],
     ["Text Classification", "Supervised Fine-Tuning", "Evaluation", "FAQ"],
     css=css,
-    title="Synthetic Data Generator",
+    title=image,
     head="Synthetic Data Generator",
     theme=theme,
 )

src/synthetic_dataset_generator/apps/eval.py

@@ -13,8 +13,9 @@ from datasets import (
     load_dataset,
 )
 from distilabel.distiset import Distiset
+from gradio.oauth import OAuthToken  #
 from gradio_huggingfacehub_search import HuggingfaceHubSearch
-from huggingface_hub import HfApi
+from huggingface_hub import HfApi, repo_exists
 
 from synthetic_dataset_generator.apps.base import (
     hide_success_message,
@@ -45,7 +46,10 @@ from synthetic_dataset_generator.utils import (
 
 def get_iframe(hub_repo_id: str) -> str:
     if not hub_repo_id:
-        raise gr.Error("Hub repository ID is required.")
+        return ""
+
+    if not repo_exists(repo_id=hub_repo_id, repo_type="dataset"):
+        return ""
 
     url = f"https://huggingface.co/datasets/{hub_repo_id}/embed/viewer"
     iframe = f"""
@@ -79,12 +83,14 @@ def get_valid_columns(dataframe: pd.DataFrame):
     return instruction_valid_columns, response_valid_columns
 
 
-def load_dataset_from_hub(repo_id: str, num_rows: int = 10):
+def load_dataset_from_hub(
+    repo_id: str, num_rows: int = 10, token: Union[OAuthToken, None] = None
+):
     if not repo_id:
         raise gr.Error("Hub repo id is required")
-    subsets = get_dataset_config_names(repo_id)
-    ds_dict = load_dataset(repo_id, subsets[0])
-    splits = get_dataset_split_names(repo_id, subsets[0])
+    subsets = get_dataset_config_names(repo_id, token=token)
+    ds_dict = load_dataset(repo_id, subsets[0], token=token)
+    splits = get_dataset_split_names(repo_id, subsets[0], token=token)
     ds = ds_dict[splits[0]]
     if num_rows:
         ds = ds.select(range(num_rows))
@@ -601,7 +607,10 @@ with gr.Blocks() as app:
                     search_type="dataset",
                     sumbit_on_select=True,
                 )
-                load_btn = gr.Button("Load dataset", variant="primary")
+                with gr.Row():
+                    load_btn = gr.Button("Load", variant="primary")
+                    clear_btn_part = gr.Button("Clear", variant="secondary")
+
             with gr.Column(scale=3):
                 search_out = gr.HTML(label="Dataset preview")
 
@@ -666,9 +675,9 @@ with gr.Blocks() as app:
                         inputs=[],
                         outputs=[eval_type],
                     )
-                btn_apply_to_sample_dataset = gr.Button(
-                    "Refresh dataset", variant="primary"
-                )
+                with gr.Row():
+                    btn_apply_to_sample_dataset = gr.Button("Save", variant="primary")
+                    clear_btn_full = gr.Button("Clear", variant="secondary")
             with gr.Column(scale=3):
                 dataframe = gr.Dataframe(
                     headers=["prompt", "completion", "evaluation"],
@@ -724,7 +733,11 @@ with gr.Blocks() as app:
                         label="Distilabel Pipeline Code",
                     )
 
-    search_in.submit(fn=get_iframe, inputs=search_in, outputs=search_out)
+    search_in.submit(fn=get_iframe, inputs=search_in, outputs=search_out).then(
+        fn=lambda df: pd.DataFrame(columns=df.columns),
+        inputs=[dataframe],
+        outputs=[dataframe],
+    )
 
     load_btn.click(
         fn=load_dataset_from_hub,
@@ -793,12 +806,8 @@ with gr.Blocks() as app:
         fn=generate_pipeline_code,
         inputs=[
             search_in,
-            aspects_instruction_response,
-            instruction_instruction_response,
-            response_instruction_response,
             prompt_template,
             structured_output,
-            num_rows,
             eval_type,
         ],
         outputs=[pipeline_code],
@@ -808,5 +817,16 @@ with gr.Blocks() as app:
         outputs=[pipeline_code_ui],
     )
 
+    clear_btn_part.click(fn=lambda x: "", inputs=[], outputs=[search_in])
+    clear_btn_full.click(
+        fn=lambda df: ("", "", pd.DataFrame(columns=df.columns)),
+        inputs=[dataframe],
+        outputs=[
+            search_in,
+            instruction_instruction_response,
+            response_instruction_response,
+        ],
+    )
+
     app.load(fn=swap_visibility, outputs=main_ui)
     app.load(fn=get_org_dropdown, outputs=[org_name])

src/synthetic_dataset_generator/apps/sft.py

@@ -78,6 +78,15 @@ def generate_sample_dataset(system_prompt, num_turns, progress=gr.Progress()):
     return dataframe
 
 
+def _get_dataframe():
+    return gr.Dataframe(
+        headers=["prompt", "completion"],
+        wrap=True,
+        height=500,
+        interactive=False,
+    )
+
+
 def generate_dataset(
     system_prompt: str,
     num_turns: int = 1,
@@ -368,7 +377,7 @@ with gr.Blocks() as app:
                             "Create",
                             variant="primary",
                         )
-                        clear_btn = gr.Button(
+                        clear_btn_part = gr.Button(
                             "Clear",
                             variant="secondary",
                         )
@@ -401,17 +410,12 @@ with gr.Blocks() as app:
                         btn_apply_to_sample_dataset = gr.Button(
                             "Save", variant="primary"
                         )
-                        clear_btn = gr.Button(
+                        clear_btn_full = gr.Button(
                             "Clear",
                             variant="secondary",
                         )
                 with gr.Column(scale=3):
-                    dataframe = gr.Dataframe(
-                        headers=["prompt", "completion"],
-                        wrap=True,
-                        height=500,
-                        interactive=False,
-                    )
+                    dataframe = _get_dataframe()
 
             gr.HTML(value="<hr>")
             gr.Markdown(value="## 3. Generate your dataset")
@@ -527,6 +531,12 @@ with gr.Blocks() as app:
             inputs=[],
             outputs=[pipeline_code_ui],
         )
+        gr.on(
+            triggers=[clear_btn_part.click, clear_btn_full.click],
+            fn=lambda _: ("", "", 1, _get_dataframe()),
+            inputs=[dataframe],
+            outputs=[dataset_description, system_prompt, num_turns, dataframe],
+        )
 
         app.load(fn=swap_visibility, outputs=main_ui)
         app.load(fn=get_org_dropdown, outputs=[org_name])

src/synthetic_dataset_generator/apps/textcat.py

@@ -35,6 +35,12 @@ from src.synthetic_dataset_generator.utils import (
 from synthetic_dataset_generator.constants import DEFAULT_BATCH_SIZE
 
 
+def _get_dataframe():
+    return gr.Dataframe(
+        headers=["labels", "text"], wrap=True, height=500, interactive=False
+    )
+
+
 def generate_system_prompt(dataset_description, temperature, progress=gr.Progress()):
     progress(0.0, desc="Generating text classification task")
     progress(0.3, desc="Initializing text generation")
@@ -345,7 +351,7 @@ with gr.Blocks() as app:
                         "Create",
                         variant="primary",
                     )
-                    clear_btn = gr.Button(
+                    clear_btn_part = gr.Button(
                         "Clear",
                         variant="secondary",
                     )
@@ -411,11 +417,9 @@ with gr.Blocks() as app:
                 )
                 with gr.Row():
                     btn_apply_to_sample_dataset = gr.Button("Save", variant="primary")
-                    clear_btn = gr.Button("Clear", variant="secondary")
+                    clear_btn_full = gr.Button("Clear", variant="secondary")
             with gr.Column(scale=3):
-                dataframe = gr.Dataframe(
-                    headers=["labels", "text"], wrap=True, height=500, interactive=False
-                )
+                dataframe = _get_dataframe()
 
         gr.HTML("<hr>")
         gr.Markdown("## 3. Generate your dataset")
@@ -553,5 +557,17 @@ with gr.Blocks() as app:
         outputs=[pipeline_code_ui],
     )
 
+    gr.on(
+        triggers=[clear_btn_part.click, clear_btn_full.click],
+        fn=lambda _: (
+            "",
+            "",
+            [],
+            _get_dataframe(),
+        ),
+        inputs=[dataframe],
+        outputs=[dataset_description, system_prompt, labels, dataframe],
+    )
+
     app.load(fn=swap_visibility, outputs=main_ui)
     app.load(fn=get_org_dropdown, outputs=[org_name])