Spaces:
Sleeping
Sleeping
edbeeching
commited on
Commit
Β·
684d1a6
1
Parent(s):
3e3c42b
fixing descriptions
Browse files
app.py
CHANGED
|
@@ -546,25 +546,65 @@ def main():
|
|
| 546 |
gr.Image("dataforge.png", show_label=False, show_download_button=False, container=False, height=300)
|
| 547 |
# Store the current oauth token for use in submit_request
|
| 548 |
current_oauth_token = gr.State(None)
|
| 549 |
-
|
| 550 |
-
|
| 551 |
-
|
|
|
|
| 552 |
|
|
|
|
|
|
|
| 553 |
# Main description
|
| 554 |
gr.Markdown("""
|
| 555 |
-
This tool allows you to **generate synthetic data from existing datasets
|
| 556 |
-
""")
|
| 557 |
|
| 558 |
-
|
| 559 |
-
|
| 560 |
|
| 561 |
-
#
|
| 562 |
with gr.Row():
|
| 563 |
-
gr.
|
| 564 |
-
|
| 565 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 566 |
|
| 567 |
-
|
|
|
|
| 568 |
main_interface = gr.Column(visible=False)
|
| 569 |
|
| 570 |
with main_interface:
|
|
|
|
| 546 |
gr.Image("dataforge.png", show_label=False, show_download_button=False, container=False, height=300)
|
| 547 |
# Store the current oauth token for use in submit_request
|
| 548 |
current_oauth_token = gr.State(None)
|
| 549 |
+
with gr.Row():
|
| 550 |
+
gr.Markdown("") # Empty space for alignment
|
| 551 |
+
login_button = gr.LoginButton(value="π Sign in", size="sm")
|
| 552 |
+
gr.Markdown("") # Empty space for alignment
|
| 553 |
|
| 554 |
+
signin_message = gr.Markdown("## π Sign In Required\n\nPlease sign in with your Hugging Face account to access the synthetic data generation service. Click the **Sign in** button above to continue.", visible=True)
|
| 555 |
+
|
| 556 |
# Main description
|
| 557 |
gr.Markdown("""
|
| 558 |
+
This tool allows you to **generate synthetic data from existing datasets**, for all your **fine-tuning/research/data augmentation** needs!
|
|
|
|
| 559 |
|
| 560 |
+
DataForge is built on top of [DataTrove](https://github.com/huggingface/datatrove), our backend data generation script is open-source and available on [GitHub](https://github.com/huggingface/dataforge). DataForge is **FREE** for HuggingFace PRO users (10,000 samples) β’ 100 samples for free users.
|
| 561 |
+
""")
|
| 562 |
|
| 563 |
+
# Usage guide and examples (right below description)
|
| 564 |
with gr.Row():
|
| 565 |
+
with gr.Column(scale=1):
|
| 566 |
+
with gr.Accordion("Usage Guide", open=False):
|
| 567 |
+
gr.Markdown("""
|
| 568 |
+
**Step-by-Step Process:**
|
| 569 |
+
1. **Load Dataset**: Enter a HF dataset name
|
| 570 |
+
2. **Load Info**: Click "Load Dataset Info"
|
| 571 |
+
3. **Choose Model**: Select from 20+ models
|
| 572 |
+
4. **Configure**: Set generation parameters
|
| 573 |
+
5. **Submit**: Monitor progress in Statistics tab
|
| 574 |
+
|
| 575 |
+
**Requirements:**
|
| 576 |
+
- Input dataset must be public on HF Hub
|
| 577 |
+
- Model must be publicly accessible
|
| 578 |
+
- Free users: 100 samples max, PRO: 10K max
|
| 579 |
+
- Token limit: 8,192 per sample
|
| 580 |
+
""")
|
| 581 |
+
with gr.Column(scale=1):
|
| 582 |
+
with gr.Accordion("Examples", open=False):
|
| 583 |
+
gr.Markdown("""
|
| 584 |
+
**Popular Use Cases:**
|
| 585 |
+
|
| 586 |
+
**Educational**: Q&A datasets
|
| 587 |
+
- Models: Qwen3-4B, Phi-3.5-mini
|
| 588 |
+
- Temperature: 0.3-0.5
|
| 589 |
+
|
| 590 |
+
**Conversational**: Multi-turn dialogues
|
| 591 |
+
- Models: Llama-3.2-3B, Mistral-7B
|
| 592 |
+
- Temperature: 0.7-0.9
|
| 593 |
+
|
| 594 |
+
**Code**: Problem β Solution
|
| 595 |
+
- Models: Qwen2.5-Coder, DeepSeek-Coder
|
| 596 |
+
- Temperature: 0.1-0.3
|
| 597 |
+
|
| 598 |
+
**Example datasets to try:**
|
| 599 |
+
```
|
| 600 |
+
simplescaling/s1K-1.1
|
| 601 |
+
HuggingFaceH4/ultrachat_200k
|
| 602 |
+
iamtarun/python_code_instructions_18k_alpaca
|
| 603 |
+
```
|
| 604 |
+
""")
|
| 605 |
|
| 606 |
+
|
| 607 |
+
# Sign in button
|
| 608 |
main_interface = gr.Column(visible=False)
|
| 609 |
|
| 610 |
with main_interface:
|