app_text_classification.py

import uuid

import gradio as gr

from io_utils import read_scanners, write_scanners
from text_classification_ui_helpers import (
    get_related_datasets_from_leaderboard,
    align_columns_and_show_prediction,
    check_dataset,
    precheck_model_ds_enable_example_btn,
    try_submit,
    write_column_mapping_to_config,
)

from text_classification import (
  get_example_prediction, 
  check_hf_token_validity, 
  HuggingFaceInferenceAPIResponse
)
from wordings import (
  CONFIRM_MAPPING_DETAILS_MD, 
  INTRODUCTION_MD, 
  USE_INFERENCE_API_TIP, 
  CHECK_LOG_SECTION_RAW,
  HF_TOKEN_INVALID_STYLED
)

MAX_LABELS = 40
MAX_FEATURES = 20

EXAMPLE_MODEL_ID = "cardiffnlp/twitter-roberta-base-sentiment-latest"
CONFIG_PATH = "./config.yaml"


def get_demo():
    with gr.Row():
        gr.Markdown(INTRODUCTION_MD)
        uid_label = gr.Textbox(
            label="Evaluation ID:", value=uuid.uuid4, visible=False, interactive=False
        )
    with gr.Row():
        model_id_input = gr.Textbox(
            label="Hugging Face model id",
            placeholder=EXAMPLE_MODEL_ID + " (press enter to confirm)",
        )

        with gr.Column():
            dataset_id_input = gr.Dropdown(
                choices=[],
                value="",
                allow_custom_value=True,
                label="Hugging Face Dataset id",
            )

    with gr.Row():
        dataset_config_input = gr.Dropdown(label="Dataset Config", visible=False, allow_custom_value=True)
        dataset_split_input = gr.Dropdown(label="Dataset Split", visible=False, allow_custom_value=True)

    with gr.Row():
        first_line_ds = gr.DataFrame(label="Dataset preview", visible=False)
    with gr.Row():
        loading_status = gr.HTML(visible=True)
    with gr.Row():
        example_btn = gr.Button(
            "Validate model & dataset",
            visible=True,
            variant="primary",
            interactive=False,
        )

    with gr.Row():
        example_input = gr.HTML(visible=False)
    with gr.Row():
        example_prediction = gr.Label(label="Model Prediction Sample", visible=False)

    with gr.Row():
        with gr.Accordion(
            label="Label and Feature Mapping", visible=False, open=False
        ) as column_mapping_accordion:
            with gr.Row():
                gr.Markdown(CONFIRM_MAPPING_DETAILS_MD)
            column_mappings = []
            with gr.Row():
                with gr.Column():
                    gr.Markdown("# Label Mapping")
                    for _ in range(MAX_LABELS):
                        column_mappings.append(gr.Dropdown(visible=False))
                with gr.Column():
                    gr.Markdown("# Feature Mapping")
                    for _ in range(MAX_LABELS, MAX_LABELS + MAX_FEATURES):
                        column_mappings.append(gr.Dropdown(visible=False))

    with gr.Accordion(label="Model Wrap Advance Config", open=True):
        gr.HTML(USE_INFERENCE_API_TIP)

        run_inference = gr.Checkbox(value=True, label="Run with Inference API")
        inference_token = gr.Textbox(
            placeholder="hf-xxxxxxxxxxxxxxxxxxxx",
            value="",
            label="HF Token for Inference API",
            visible=True,
            interactive=True,
        )
        inference_token_info = gr.HTML(value=HF_TOKEN_INVALID_STYLED, visible=False)

        inference_token.change(
            lambda token: gr.update(visible=lambda: check_hf_token_validity(token)),
            inputs=[inference_token],
            outputs=[inference_token_info],
        )

    with gr.Accordion(label="Scanner Advance Config (optional)", open=False):
        scanners = gr.CheckboxGroup(label="Scan Settings", visible=True)

        @gr.on(triggers=[uid_label.change], inputs=[uid_label], outputs=[scanners])
        def get_scanners(uid):
            selected = read_scanners(uid)
            # we remove data_leakage from the default scanners
            # Reason: data_leakage barely raises any issues and takes too many requests
            # when using inference API, causing rate limit error
            scan_config = selected + ["data_leakage"]
            return gr.update(
                choices=scan_config, value=selected, label="Scan Settings", visible=True
            )

    with gr.Row():
        run_btn = gr.Button(
            "Get Evaluation Result",
            variant="primary",
            interactive=False,
            size="lg",
        )

    with gr.Row():
        logs = gr.Textbox(
            value=CHECK_LOG_SECTION_RAW,
            label="Giskard Bot Evaluation Guide:",
            visible=False,
            every=0.5,
        )

    
    scanners.change(write_scanners, inputs=[scanners, uid_label])

    gr.on(
        triggers=[model_id_input.change],
        fn=get_related_datasets_from_leaderboard,
        inputs=[model_id_input],
        outputs=[dataset_id_input],
    ).then(
        fn=check_dataset, 
        inputs=[dataset_id_input], 
        outputs=[dataset_config_input, dataset_split_input, loading_status]
    )
    
    gr.on(
        triggers=[dataset_id_input.change],
        fn=check_dataset,
        inputs=[dataset_id_input],
        outputs=[dataset_config_input, dataset_split_input, loading_status]
    )

    gr.on(
        triggers=[label.change for label in column_mappings],
        fn=write_column_mapping_to_config,
        inputs=[
            uid_label,
            *column_mappings,
        ],
    )

    # label.change sometimes does not pass the changed value
    gr.on(
        triggers=[label.input for label in column_mappings],
        fn=write_column_mapping_to_config,
        inputs=[
            uid_label,
            *column_mappings,
        ],
    )

    gr.on(
        triggers=[
            model_id_input.change,
            dataset_id_input.change,
            dataset_config_input.change,
            dataset_split_input.change,
        ],
        fn=precheck_model_ds_enable_example_btn,
        inputs=[
            model_id_input,
            dataset_id_input,
            dataset_config_input,
            dataset_split_input,
        ],
        outputs=[example_btn, first_line_ds, loading_status],
    )

    gr.on(
        triggers=[
            example_btn.click,
        ],
        fn=align_columns_and_show_prediction,
        inputs=[
            model_id_input,
            dataset_id_input,
            dataset_config_input,
            dataset_split_input,
            uid_label,
            run_inference,
            inference_token,
        ],
        outputs=[
            example_input,
            example_prediction,
            column_mapping_accordion,
            run_btn,
            loading_status,
            *column_mappings,
        ],
    )

    gr.on(
        triggers=[
            run_btn.click,
        ],
        fn=try_submit,
        inputs=[
            model_id_input,
            dataset_id_input,
            dataset_config_input,
            dataset_split_input,
            run_inference,
            inference_token,
            uid_label,
        ],
        outputs=[run_btn, logs, uid_label],
    )

    def enable_run_btn(run_inference, inference_token, model_id, dataset_id, dataset_config, dataset_split):
        if not run_inference or inference_token == "":
            return gr.update(interactive=False)
        if model_id == "" or dataset_id == "" or dataset_config == "" or dataset_split == "":
            return gr.update(interactive=False)
        if not column_mapping_accordion.visible:
            return gr.update(interactive=False)
        _, prediction_response = get_example_prediction(
            model_id, dataset_id, dataset_config, dataset_split, inference_token
        )
        if not isinstance(prediction_response, HuggingFaceInferenceAPIResponse):
            gr.warning("Your HF token is invalid. Please check your token.")
            return gr.update(interactive=False)
        return gr.update(interactive=True)

    gr.on(
        triggers=[
            run_inference.input,
            inference_token.input,
            scanners.input,
        ],
        fn=enable_run_btn,
        inputs=[
            run_inference, 
            inference_token, 
            model_id_input, 
            dataset_id_input, 
            dataset_config_input, 
            dataset_split_input
        ],
        outputs=[run_btn],
    )

    gr.on(
        triggers=[label.input for label in column_mappings],
        fn=enable_run_btn,
        inputs=[
            run_inference, 
            inference_token, 
            model_id_input, 
            dataset_id_input, 
            dataset_config_input, 
            dataset_split_input
        ],  # FIXME
        outputs=[run_btn],
    )