Spaces:
Build error
Build error
Update app_dialogue.py
Browse files- app_dialogue.py +207 -201
app_dialogue.py
CHANGED
@@ -561,7 +561,7 @@ textbox = gr.Textbox(
|
|
561 |
|
562 |
css2 = """
|
563 |
#chatbot {
|
564 |
-
background-image: url('https://huggingface.co/spaces/ysharma/
|
565 |
background-repeat: repeat;}
|
566 |
"""
|
567 |
|
@@ -580,194 +580,49 @@ with gr.Blocks(title="IDEFICS-Chat", theme=gr.themes.Base(), css=css2) as demo:
|
|
580 |
"""
|
581 |
)
|
582 |
|
583 |
-
with gr.Row():
|
584 |
-
|
585 |
-
|
586 |
-
|
587 |
-
|
588 |
-
|
589 |
-
|
590 |
-
|
591 |
-
|
592 |
-
|
593 |
-
|
594 |
-
|
595 |
-
|
596 |
-
imagebox = gr.Image(type="filepath", label="Image input", visible=False)
|
597 |
-
|
598 |
-
with gr.Accordion("Advanced parameters", open=False, visible=True) as parameter_row:
|
599 |
-
max_new_tokens = gr.Slider(
|
600 |
-
minimum=0,
|
601 |
-
maximum=2048,
|
602 |
-
value=512,
|
603 |
-
step=1,
|
604 |
-
interactive=True,
|
605 |
-
label="Maximum number of new tokens to generate",
|
606 |
-
)
|
607 |
-
min_length = gr.Slider(
|
608 |
-
minimum=0,
|
609 |
-
maximum=50,
|
610 |
-
value=0,
|
611 |
-
step=1,
|
612 |
-
interactive=True,
|
613 |
-
label="Minimum number of new tokens to generate",
|
614 |
-
)
|
615 |
-
repetition_penalty = gr.Slider(
|
616 |
-
minimum=0.0,
|
617 |
-
maximum=5.0,
|
618 |
-
value=1.0,
|
619 |
-
step=0.1,
|
620 |
-
interactive=True,
|
621 |
-
label="Repetition penalty",
|
622 |
-
info="1.0 means no penalty",
|
623 |
-
)
|
624 |
-
no_repeat_ngram_size = gr.Slider(
|
625 |
-
minimum=0,
|
626 |
-
maximum=10,
|
627 |
-
value=0,
|
628 |
-
step=1,
|
629 |
-
interactive=True,
|
630 |
-
label="N-gram repetition threshold",
|
631 |
-
info="If set to int > 0, all ngrams of that size can only occur once.",
|
632 |
-
)
|
633 |
-
decoding_strategy = gr.Radio(
|
634 |
-
[
|
635 |
-
"Greedy",
|
636 |
-
# "beam_search",
|
637 |
-
# "beam_sampling",
|
638 |
-
# "sampling_top_k",
|
639 |
-
"Top P Sampling",
|
640 |
-
],
|
641 |
-
value="Top P Sampling",
|
642 |
-
label="Decoding strategy",
|
643 |
-
interactive=True,
|
644 |
-
)
|
645 |
-
temperature = gr.Slider(
|
646 |
-
minimum=0.0,
|
647 |
-
maximum=5.0,
|
648 |
-
value=1.2,
|
649 |
-
step=0.1,
|
650 |
-
interactive=True,
|
651 |
-
label="Sampling temperature",
|
652 |
-
)
|
653 |
-
decoding_strategy.change(
|
654 |
-
fn=lambda selection: gr.Slider.update(
|
655 |
-
visible=(
|
656 |
-
selection in ["contrastive_sampling", "beam_sampling", "Top P Sampling", "sampling_top_k"]
|
657 |
-
)
|
658 |
-
),
|
659 |
-
inputs=decoding_strategy,
|
660 |
-
outputs=temperature,
|
661 |
-
)
|
662 |
-
num_beams = gr.Slider(
|
663 |
-
minimum=0,
|
664 |
-
maximum=20,
|
665 |
-
value=3.0,
|
666 |
-
step=1.0,
|
667 |
-
interactive=True,
|
668 |
-
visible=False,
|
669 |
-
label="Number of beams",
|
670 |
-
info="Only used if `decoding_strategy` is `beam_search` or `beam_sampling`.",
|
671 |
-
)
|
672 |
-
decoding_strategy.change(
|
673 |
-
fn=lambda selection: gr.Slider.update(visible=(selection in ["beam_search", "beam_sampling"])),
|
674 |
-
inputs=decoding_strategy,
|
675 |
-
outputs=num_beams,
|
676 |
-
)
|
677 |
-
top_p = gr.Slider(
|
678 |
-
minimum=0.0,
|
679 |
-
maximum=1.0,
|
680 |
-
value=0.8,
|
681 |
-
step=0.01,
|
682 |
-
interactive=True,
|
683 |
-
visible=True,
|
684 |
-
label="Top P",
|
685 |
-
info=(
|
686 |
-
"If set to float < 1, only the smallest set of most probable tokens with probabilities that"
|
687 |
-
" add up to top_p or higher are kept for generation."
|
688 |
-
),
|
689 |
-
)
|
690 |
-
decoding_strategy.change(
|
691 |
-
fn=lambda selection: gr.Slider.update(visible=(selection in ["Top P Sampling"])),
|
692 |
-
inputs=decoding_strategy,
|
693 |
-
outputs=top_p,
|
694 |
-
)
|
695 |
-
top_k = gr.Slider(
|
696 |
-
minimum=0,
|
697 |
-
maximum=500,
|
698 |
-
value=50,
|
699 |
-
step=1,
|
700 |
-
interactive=True,
|
701 |
-
visible=False,
|
702 |
-
label="Top K",
|
703 |
-
info="The number of highest probability vocabulary tokens to keep for top-k-filtering.",
|
704 |
-
)
|
705 |
-
decoding_strategy.change(
|
706 |
-
fn=lambda selection: gr.Slider.update(visible=(selection in ["sampling_top_k"])),
|
707 |
-
inputs=decoding_strategy,
|
708 |
-
outputs=top_k,
|
709 |
-
)
|
710 |
-
length_penalty = gr.Slider(
|
711 |
-
minimum=-1000.0,
|
712 |
-
maximum=1000.0,
|
713 |
-
value=1.0,
|
714 |
-
step=0.1,
|
715 |
-
interactive=True,
|
716 |
-
visible=False,
|
717 |
-
label="Length penalty",
|
718 |
-
info=(
|
719 |
-
"length_penalty > 0.0 promotes longer sequences, while length_penalty < 0.0 encourages shorter"
|
720 |
-
" sequences. Only used if `decoding_strategy` is `beam_search` or `beam_sampling`."
|
721 |
-
),
|
722 |
-
)
|
723 |
-
decoding_strategy.change(
|
724 |
-
fn=lambda selection: gr.Slider.update(visible=(selection in ["beam_search", "beam_sampling"])),
|
725 |
-
inputs=decoding_strategy,
|
726 |
-
outputs=length_penalty,
|
727 |
-
)
|
728 |
-
penalty_alpha = gr.Slider(
|
729 |
-
minimum=0.0,
|
730 |
-
maximum=5.0,
|
731 |
-
value=0.95,
|
732 |
-
step=0.05,
|
733 |
-
interactive=True,
|
734 |
-
visible=False,
|
735 |
-
label="Penalty alpha",
|
736 |
-
info="Only used if `decoding_strategy` is `contrastive_sampling`.",
|
737 |
-
)
|
738 |
-
decoding_strategy.change(
|
739 |
-
fn=lambda selection: gr.Slider.update(visible=(selection in ["contrastive_sampling"])),
|
740 |
-
inputs=decoding_strategy,
|
741 |
-
outputs=penalty_alpha,
|
742 |
-
)
|
743 |
-
|
744 |
-
with gr.Column(scale=6):
|
745 |
|
746 |
-
|
747 |
-
"""
|
748 |
-
Pre-fetch the images that are passed in the chatbot default history.
|
749 |
-
"""
|
750 |
-
return user_prompt_list_to_markdown(handle_manual_images_in_user_prompt(user_prompt_str))
|
751 |
|
752 |
-
|
753 |
-
|
754 |
-
|
755 |
-
|
756 |
-
|
757 |
-
|
758 |
-
|
759 |
-
|
760 |
-
|
761 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
762 |
There should be drama and most importantly romance.
|
763 |
Character 1:<fake_token_around_image><image:https://m.media-amazon.com/images/M/MV5BODg1OTczMWEtNTU3MS00OTUzLThjODEtNDg1MWQwZmExYmFlXkEyXkFqcGdeQWFybm8@._V1_QL75_UX500_CR0,0,500,281_.jpg><fake_token_around_image>
|
764 |
Character 2:<fake_token_around_image><image:https://static.wikia.nocookie.net/dreamworks/images/0/09/Shrek_dreamworks.png/revision/latest?cb=20170921225439><fake_token_around_image>
|
765 |
Character 3:<fake_token_around_image><image:https://upload.wikimedia.org/wikipedia/en/e/ef/Marsupilami_1.jpg><fake_token_around_image>
|
766 |
The plot should take place in the world pictured here:<fake_token_around_image><image:https://www.astronomy.com/wp-content/uploads/sites/2/2021/09/ON1.jpg><fake_token_around_image>"""
|
767 |
-
|
768 |
-
|
769 |
-
|
770 |
-
|
771 |
|
772 |
Mulan, being the kind-hearted person she was, decided to help Shrek find his way back. As they traveled together, they began to develop a strong bond. Shrek was impressed by Mulan's bravery and determination, while Mulan admired Shrek's loyalty and sense of humor.
|
773 |
|
@@ -780,27 +635,178 @@ Finally, they reached Shrek's home, and he was reunited with his family and frie
|
|
780 |
Mulan was overjoyed, and they shared a passionate kiss. From that day on, they lived happily ever after, exploring the world together and facing any challenges that came their way.
|
781 |
|
782 |
And so, the story of Mulan and Shrek's romance came to an end, leaving a lasting impression on all who heard it."""
|
783 |
-
|
784 |
-
],
|
785 |
],
|
786 |
-
|
787 |
-
|
788 |
-
with gr.Row():
|
789 |
-
with gr.Column(scale=7):
|
790 |
-
textbox.render()
|
791 |
-
with gr.Column(scale=1, min_width=80):
|
792 |
-
submit_btn = gr.Button(value="Submit", visible=True)
|
793 |
-
with gr.Column(scale=1, min_width=10):
|
794 |
-
clear_btn = gr.ClearButton([textbox, chatbot])
|
795 |
-
with gr.Column(scale=1, min_width=10):
|
796 |
-
upload_btn = gr.UploadButton("📁", file_types=["image"])
|
797 |
-
with gr.Group():
|
798 |
-
with gr.Row():
|
799 |
-
with gr.Column(scale=1, min_width=50):
|
800 |
-
dope_bttn = gr.Button("Dope🔥")
|
801 |
-
with gr.Column(scale=1, min_width=50):
|
802 |
-
problematic_bttn = gr.Button("Problematic😬")
|
803 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
804 |
def model_inference(
|
805 |
user_prompt_str,
|
806 |
chat_history,
|
|
|
561 |
|
562 |
css2 = """
|
563 |
#chatbot {
|
564 |
+
background-image: url('https://huggingface.co/spaces/ysharma/dummy_m4/resolve/main/idefics_transparent20.png');
|
565 |
background-repeat: repeat;}
|
566 |
"""
|
567 |
|
|
|
580 |
"""
|
581 |
)
|
582 |
|
583 |
+
#with gr.Row():
|
584 |
+
# with gr.Column(): #(scale=3):
|
585 |
+
with gr.Row(elem_id="model_selector_row"):
|
586 |
+
model_selector = gr.Dropdown(
|
587 |
+
choices=MODELS,
|
588 |
+
value="HuggingFaceM4/idefics-9b-instruct",
|
589 |
+
interactive=True,
|
590 |
+
show_label=False,
|
591 |
+
container=False,
|
592 |
+
label="Model"
|
593 |
+
)
|
594 |
+
processor, tokenizer, model = load_processor_tokenizer_model(model_selector.value)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
595 |
|
596 |
+
imagebox = gr.Image(type="filepath", label="Image input", visible=False)
|
|
|
|
|
|
|
|
|
597 |
|
598 |
+
with gr.Row():
|
599 |
+
#with gr.Column(scale=6):
|
600 |
+
|
601 |
+
def prefetch_images_in_history(user_prompt_str):
|
602 |
+
"""
|
603 |
+
Pre-fetch the images that are passed in the chatbot default history.
|
604 |
+
"""
|
605 |
+
return user_prompt_list_to_markdown(handle_manual_images_in_user_prompt(user_prompt_str))
|
606 |
+
|
607 |
+
chatbot = gr.Chatbot(
|
608 |
+
elem_id="chatbot",
|
609 |
+
label="ChatIDEFICS",
|
610 |
+
visible=True,
|
611 |
+
height=550,
|
612 |
+
value=[
|
613 |
+
[
|
614 |
+
(
|
615 |
+
prefetch_images_in_history(
|
616 |
+
"""Write a funny story including the following characters.
|
617 |
There should be drama and most importantly romance.
|
618 |
Character 1:<fake_token_around_image><image:https://m.media-amazon.com/images/M/MV5BODg1OTczMWEtNTU3MS00OTUzLThjODEtNDg1MWQwZmExYmFlXkEyXkFqcGdeQWFybm8@._V1_QL75_UX500_CR0,0,500,281_.jpg><fake_token_around_image>
|
619 |
Character 2:<fake_token_around_image><image:https://static.wikia.nocookie.net/dreamworks/images/0/09/Shrek_dreamworks.png/revision/latest?cb=20170921225439><fake_token_around_image>
|
620 |
Character 3:<fake_token_around_image><image:https://upload.wikimedia.org/wikipedia/en/e/ef/Marsupilami_1.jpg><fake_token_around_image>
|
621 |
The plot should take place in the world pictured here:<fake_token_around_image><image:https://www.astronomy.com/wp-content/uploads/sites/2/2021/09/ON1.jpg><fake_token_around_image>"""
|
622 |
+
)
|
623 |
+
),
|
624 |
+
(
|
625 |
+
"""Once upon a time, in a world where the sky was a dome and the ground was covered in grass, there lived a beautiful princess named Mulan. She was kind, brave, and always ready to help others. One day, while walking through the forest, she stumbled upon a strange creature. It was a green ogre named Shrek, who was lost and looking for his way back home.
|
626 |
|
627 |
Mulan, being the kind-hearted person she was, decided to help Shrek find his way back. As they traveled together, they began to develop a strong bond. Shrek was impressed by Mulan's bravery and determination, while Mulan admired Shrek's loyalty and sense of humor.
|
628 |
|
|
|
635 |
Mulan was overjoyed, and they shared a passionate kiss. From that day on, they lived happily ever after, exploring the world together and facing any challenges that came their way.
|
636 |
|
637 |
And so, the story of Mulan and Shrek's romance came to an end, leaving a lasting impression on all who heard it."""
|
638 |
+
),
|
|
|
639 |
],
|
640 |
+
],
|
641 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
642 |
|
643 |
+
with gr.Row():
|
644 |
+
with gr.Group():
|
645 |
+
with gr.Row():
|
646 |
+
with gr.Column(scale=7):
|
647 |
+
textbox.render()
|
648 |
+
with gr.Column(scale=1, min_width=80):
|
649 |
+
submit_btn = gr.Button(value="Submit", visible=True)
|
650 |
+
with gr.Column(scale=1, min_width=10):
|
651 |
+
clear_btn = gr.ClearButton([textbox, chatbot])
|
652 |
+
with gr.Column(scale=1, min_width=10):
|
653 |
+
upload_btn = gr.UploadButton("📁", file_types=["image"])
|
654 |
+
with gr.Row():
|
655 |
+
with gr.Group():
|
656 |
+
with gr.Row():
|
657 |
+
with gr.Column(scale=1, min_width=50):
|
658 |
+
dope_bttn = gr.Button("Dope🔥")
|
659 |
+
with gr.Column(scale=1, min_width=50):
|
660 |
+
problematic_bttn = gr.Button("Problematic😬")
|
661 |
+
|
662 |
+
with gr.Row():
|
663 |
+
with gr.Accordion("Advanced parameters", open=False, visible=True) as parameter_row:
|
664 |
+
max_new_tokens = gr.Slider(
|
665 |
+
minimum=0,
|
666 |
+
maximum=2048,
|
667 |
+
value=512,
|
668 |
+
step=1,
|
669 |
+
interactive=True,
|
670 |
+
label="Maximum number of new tokens to generate",
|
671 |
+
)
|
672 |
+
min_length = gr.Slider(
|
673 |
+
minimum=0,
|
674 |
+
maximum=50,
|
675 |
+
value=0,
|
676 |
+
step=1,
|
677 |
+
interactive=True,
|
678 |
+
label="Minimum number of new tokens to generate",
|
679 |
+
)
|
680 |
+
repetition_penalty = gr.Slider(
|
681 |
+
minimum=0.0,
|
682 |
+
maximum=5.0,
|
683 |
+
value=1.0,
|
684 |
+
step=0.1,
|
685 |
+
interactive=True,
|
686 |
+
label="Repetition penalty",
|
687 |
+
info="1.0 means no penalty",
|
688 |
+
)
|
689 |
+
no_repeat_ngram_size = gr.Slider(
|
690 |
+
minimum=0,
|
691 |
+
maximum=10,
|
692 |
+
value=0,
|
693 |
+
step=1,
|
694 |
+
interactive=True,
|
695 |
+
label="N-gram repetition threshold",
|
696 |
+
info="If set to int > 0, all ngrams of that size can only occur once.",
|
697 |
+
)
|
698 |
+
decoding_strategy = gr.Radio(
|
699 |
+
[
|
700 |
+
"Greedy",
|
701 |
+
# "beam_search",
|
702 |
+
# "beam_sampling",
|
703 |
+
# "sampling_top_k",
|
704 |
+
"Top P Sampling",
|
705 |
+
],
|
706 |
+
value="Top P Sampling",
|
707 |
+
label="Decoding strategy",
|
708 |
+
interactive=True,
|
709 |
+
)
|
710 |
+
temperature = gr.Slider(
|
711 |
+
minimum=0.0,
|
712 |
+
maximum=5.0,
|
713 |
+
value=1.2,
|
714 |
+
step=0.1,
|
715 |
+
interactive=True,
|
716 |
+
label="Sampling temperature",
|
717 |
+
)
|
718 |
+
decoding_strategy.change(
|
719 |
+
fn=lambda selection: gr.Slider.update(
|
720 |
+
visible=(
|
721 |
+
selection in ["contrastive_sampling", "beam_sampling", "Top P Sampling", "sampling_top_k"]
|
722 |
+
)
|
723 |
+
),
|
724 |
+
inputs=decoding_strategy,
|
725 |
+
outputs=temperature,
|
726 |
+
)
|
727 |
+
num_beams = gr.Slider(
|
728 |
+
minimum=0,
|
729 |
+
maximum=20,
|
730 |
+
value=3.0,
|
731 |
+
step=1.0,
|
732 |
+
interactive=True,
|
733 |
+
visible=False,
|
734 |
+
label="Number of beams",
|
735 |
+
info="Only used if `decoding_strategy` is `beam_search` or `beam_sampling`.",
|
736 |
+
)
|
737 |
+
decoding_strategy.change(
|
738 |
+
fn=lambda selection: gr.Slider.update(visible=(selection in ["beam_search", "beam_sampling"])),
|
739 |
+
inputs=decoding_strategy,
|
740 |
+
outputs=num_beams,
|
741 |
+
)
|
742 |
+
top_p = gr.Slider(
|
743 |
+
minimum=0.0,
|
744 |
+
maximum=1.0,
|
745 |
+
value=0.8,
|
746 |
+
step=0.01,
|
747 |
+
interactive=True,
|
748 |
+
visible=True,
|
749 |
+
label="Top P",
|
750 |
+
info=(
|
751 |
+
"If set to float < 1, only the smallest set of most probable tokens with probabilities that"
|
752 |
+
" add up to top_p or higher are kept for generation."
|
753 |
+
),
|
754 |
+
)
|
755 |
+
decoding_strategy.change(
|
756 |
+
fn=lambda selection: gr.Slider.update(visible=(selection in ["Top P Sampling"])),
|
757 |
+
inputs=decoding_strategy,
|
758 |
+
outputs=top_p,
|
759 |
+
)
|
760 |
+
top_k = gr.Slider(
|
761 |
+
minimum=0,
|
762 |
+
maximum=500,
|
763 |
+
value=50,
|
764 |
+
step=1,
|
765 |
+
interactive=True,
|
766 |
+
visible=False,
|
767 |
+
label="Top K",
|
768 |
+
info="The number of highest probability vocabulary tokens to keep for top-k-filtering.",
|
769 |
+
)
|
770 |
+
decoding_strategy.change(
|
771 |
+
fn=lambda selection: gr.Slider.update(visible=(selection in ["sampling_top_k"])),
|
772 |
+
inputs=decoding_strategy,
|
773 |
+
outputs=top_k,
|
774 |
+
)
|
775 |
+
length_penalty = gr.Slider(
|
776 |
+
minimum=-1000.0,
|
777 |
+
maximum=1000.0,
|
778 |
+
value=1.0,
|
779 |
+
step=0.1,
|
780 |
+
interactive=True,
|
781 |
+
visible=False,
|
782 |
+
label="Length penalty",
|
783 |
+
info=(
|
784 |
+
"length_penalty > 0.0 promotes longer sequences, while length_penalty < 0.0 encourages shorter"
|
785 |
+
" sequences. Only used if `decoding_strategy` is `beam_search` or `beam_sampling`."
|
786 |
+
),
|
787 |
+
)
|
788 |
+
decoding_strategy.change(
|
789 |
+
fn=lambda selection: gr.Slider.update(visible=(selection in ["beam_search", "beam_sampling"])),
|
790 |
+
inputs=decoding_strategy,
|
791 |
+
outputs=length_penalty,
|
792 |
+
)
|
793 |
+
penalty_alpha = gr.Slider(
|
794 |
+
minimum=0.0,
|
795 |
+
maximum=5.0,
|
796 |
+
value=0.95,
|
797 |
+
step=0.05,
|
798 |
+
interactive=True,
|
799 |
+
visible=False,
|
800 |
+
label="Penalty alpha",
|
801 |
+
info="Only used if `decoding_strategy` is `contrastive_sampling`.",
|
802 |
+
)
|
803 |
+
decoding_strategy.change(
|
804 |
+
fn=lambda selection: gr.Slider.update(visible=(selection in ["contrastive_sampling"])),
|
805 |
+
inputs=decoding_strategy,
|
806 |
+
outputs=penalty_alpha,
|
807 |
+
)
|
808 |
+
|
809 |
+
|
810 |
def model_inference(
|
811 |
user_prompt_str,
|
812 |
chat_history,
|