Spaces:
Build error
Build error
Update app.py
Browse files
app.py
CHANGED
|
@@ -568,11 +568,173 @@ with gr.Blocks() as demo:
|
|
| 568 |
outputs=[combo_output, combo_info]
|
| 569 |
)
|
| 570 |
|
| 571 |
-
|
| 572 |
-
|
| 573 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 574 |
|
| 575 |
-
|
|
|
|
|
|
|
| 576 |
|
| 577 |
# with gr.Tab("Assignment 1"):
|
| 578 |
# # Assignment content here
|
|
|
|
| 568 |
outputs=[combo_output, combo_info]
|
| 569 |
)
|
| 570 |
|
| 571 |
+
with gr.Tab("Week 5: Advanced Prompting"):
|
| 572 |
+
gr.Markdown("# Advanced Prompt Engineering Techniques")
|
| 573 |
+
gr.Markdown("Explore sophisticated prompting strategies and visualise reasoning patterns")
|
| 574 |
+
|
| 575 |
+
with gr.Tabs() as week5_tabs:
|
| 576 |
+
|
| 577 |
+
with gr.Tab("Tree of Thought Explorer"):
|
| 578 |
+
gr.Markdown("""
|
| 579 |
+
### Visualise Multi-Path Reasoning
|
| 580 |
+
The model will break down your problem into multiple approaches, evaluate each one, and select the best path.
|
| 581 |
+
""")
|
| 582 |
+
|
| 583 |
+
with gr.Row():
|
| 584 |
+
tot_input = gr.Textbox(
|
| 585 |
+
label="Enter a problem to solve",
|
| 586 |
+
placeholder="e.g., How can we improve user engagement on a mobile app?",
|
| 587 |
+
lines=3,
|
| 588 |
+
value="How should a startup decide between building a mobile app or a web application first?"
|
| 589 |
+
)
|
| 590 |
+
|
| 591 |
+
with gr.Row():
|
| 592 |
+
generate_tot = gr.Button("Generate Tree of Thought", variant="primary")
|
| 593 |
+
|
| 594 |
+
with gr.Row():
|
| 595 |
+
tot_output = gr.Textbox(
|
| 596 |
+
label="Reasoning Process",
|
| 597 |
+
lines=12
|
| 598 |
+
)
|
| 599 |
+
|
| 600 |
+
with gr.Row():
|
| 601 |
+
tot_visualization = gr.HTML(
|
| 602 |
+
label="Tree Visualisation"
|
| 603 |
+
)
|
| 604 |
+
|
| 605 |
+
with gr.Tab("Self-Consistency Testing"):
|
| 606 |
+
gr.Markdown("""
|
| 607 |
+
### Test Response Consistency
|
| 608 |
+
Run the same prompt multiple times to identify consistent patterns and areas of uncertainty.
|
| 609 |
+
""")
|
| 610 |
+
|
| 611 |
+
with gr.Row():
|
| 612 |
+
consistency_input = gr.Textbox(
|
| 613 |
+
label="Enter your prompt",
|
| 614 |
+
placeholder="Ask a question that requires reasoning...",
|
| 615 |
+
lines=3,
|
| 616 |
+
value="What are the three most important factors in choosing a database system?"
|
| 617 |
+
)
|
| 618 |
+
|
| 619 |
+
with gr.Row():
|
| 620 |
+
num_runs = gr.Slider(
|
| 621 |
+
minimum=3,
|
| 622 |
+
maximum=5,
|
| 623 |
+
value=3,
|
| 624 |
+
step=1,
|
| 625 |
+
label="Number of generations"
|
| 626 |
+
)
|
| 627 |
+
consistency_temp = gr.Slider(
|
| 628 |
+
minimum=0.3,
|
| 629 |
+
maximum=0.9,
|
| 630 |
+
value=0.7,
|
| 631 |
+
step=0.1,
|
| 632 |
+
label="Temperature"
|
| 633 |
+
)
|
| 634 |
+
|
| 635 |
+
with gr.Row():
|
| 636 |
+
generate_consistency = gr.Button("Generate Multiple Responses", variant="primary")
|
| 637 |
+
|
| 638 |
+
with gr.Row():
|
| 639 |
+
consistency_analysis = gr.Textbox(
|
| 640 |
+
label="Analysis Guide",
|
| 641 |
+
lines=4
|
| 642 |
+
)
|
| 643 |
+
|
| 644 |
+
with gr.Row():
|
| 645 |
+
consistency_output1 = gr.Textbox(label="Response 1", lines=5)
|
| 646 |
+
consistency_output2 = gr.Textbox(label="Response 2", lines=5)
|
| 647 |
+
|
| 648 |
+
with gr.Row():
|
| 649 |
+
consistency_output3 = gr.Textbox(label="Response 3", lines=5)
|
| 650 |
+
consistency_output4 = gr.Textbox(label="Response 4 (if selected)", lines=5, visible=True)
|
| 651 |
+
|
| 652 |
+
with gr.Row():
|
| 653 |
+
consistency_output5 = gr.Textbox(label="Response 5 (if selected)", lines=5, visible=True)
|
| 654 |
+
|
| 655 |
+
with gr.Tab("Prompt Structure Comparison"):
|
| 656 |
+
gr.Markdown("""
|
| 657 |
+
### Compare Structural Strategies
|
| 658 |
+
Test how different prompt structures affect model attention and output quality.
|
| 659 |
+
""")
|
| 660 |
+
|
| 661 |
+
with gr.Row():
|
| 662 |
+
structure_input = gr.Textbox(
|
| 663 |
+
label="Enter your task",
|
| 664 |
+
placeholder="Enter a task or question...",
|
| 665 |
+
lines=3,
|
| 666 |
+
value=""
|
| 667 |
+
)
|
| 668 |
+
|
| 669 |
+
with gr.Row():
|
| 670 |
+
gr.Markdown("### Select ONE structure to test:")
|
| 671 |
+
|
| 672 |
+
with gr.Row():
|
| 673 |
+
structure_radio = gr.Radio(
|
| 674 |
+
choices=[
|
| 675 |
+
"Baseline (no special structure)",
|
| 676 |
+
"Front-loading (critical instruction first)",
|
| 677 |
+
"Delimiter strategy (section separation)",
|
| 678 |
+
"Sandwich technique (instruction at start and end)"
|
| 679 |
+
],
|
| 680 |
+
label="Prompt Structure",
|
| 681 |
+
value="Baseline (no special structure)"
|
| 682 |
+
)
|
| 683 |
+
|
| 684 |
+
with gr.Row():
|
| 685 |
+
generate_structure = gr.Button("Generate Response", variant="primary")
|
| 686 |
+
|
| 687 |
+
with gr.Row():
|
| 688 |
+
structure_output = gr.Textbox(
|
| 689 |
+
label="Response",
|
| 690 |
+
lines=8
|
| 691 |
+
)
|
| 692 |
+
structure_info = gr.Textbox(
|
| 693 |
+
label="Structure Information",
|
| 694 |
+
lines=8
|
| 695 |
+
)
|
| 696 |
+
|
| 697 |
+
# Week 5 Event Handlers
|
| 698 |
+
def handle_tot(task):
|
| 699 |
+
text_output, svg_output = generate_tot_response(task)
|
| 700 |
+
return text_output, svg_output
|
| 701 |
+
|
| 702 |
+
def handle_consistency(prompt, runs, temp):
|
| 703 |
+
responses, analysis = generate_self_consistency(prompt, int(runs), temp)
|
| 704 |
+
while len(responses) < 5:
|
| 705 |
+
responses.append("")
|
| 706 |
+
return analysis, responses[0], responses[1], responses[2], responses[3], responses[4]
|
| 707 |
+
|
| 708 |
+
def handle_structure(task, structure_choice):
|
| 709 |
+
use_frontload = "Front-loading" in structure_choice
|
| 710 |
+
use_delimiters = "Delimiter" in structure_choice
|
| 711 |
+
use_sandwich = "Sandwich" in structure_choice
|
| 712 |
+
|
| 713 |
+
output, info = compare_prompt_structures(task, use_frontload, use_delimiters, use_sandwich)
|
| 714 |
+
return output, info
|
| 715 |
+
|
| 716 |
+
generate_tot.click(
|
| 717 |
+
handle_tot,
|
| 718 |
+
inputs=[tot_input],
|
| 719 |
+
outputs=[tot_output, tot_visualization]
|
| 720 |
+
)
|
| 721 |
+
|
| 722 |
+
generate_consistency.click(
|
| 723 |
+
handle_consistency,
|
| 724 |
+
inputs=[consistency_input, num_runs, consistency_temp],
|
| 725 |
+
outputs=[consistency_analysis, consistency_output1, consistency_output2,
|
| 726 |
+
consistency_output3, consistency_output4, consistency_output5]
|
| 727 |
+
)
|
| 728 |
+
|
| 729 |
+
generate_structure.click(
|
| 730 |
+
handle_structure,
|
| 731 |
+
inputs=[structure_input, structure_radio],
|
| 732 |
+
outputs=[structure_output, structure_info]
|
| 733 |
+
)
|
| 734 |
|
| 735 |
+
# with gr.Tab("Week 8: Error Detection"):
|
| 736 |
+
# # Week 8 content here
|
| 737 |
+
# pass
|
| 738 |
|
| 739 |
# with gr.Tab("Assignment 1"):
|
| 740 |
# # Assignment content here
|