app.py

import gradio as gr
import datasets
import huggingface_hub
import os
import time
import subprocess
import logging

import json

from transformers.pipelines import TextClassificationPipeline

from text_classification import check_column_mapping_keys_validity, text_classification_fix_column_mapping
from utils import read_scanners, write_scanners, read_inference_type, write_inference_type, convert_column_mapping_to_json

HF_REPO_ID = 'HF_REPO_ID'
HF_SPACE_ID = 'SPACE_ID'
HF_WRITE_TOKEN = 'HF_WRITE_TOKEN'

theme = gr.themes.Soft(
    primary_hue="green",
)

def check_model(model_id):
    try:
        task = huggingface_hub.model_info(model_id).pipeline_tag
    except Exception:
        return None, None

    try:
        from transformers import pipeline
        ppl = pipeline(task=task, model=model_id)

        return model_id, ppl
    except Exception as e:
        return model_id, e


def check_dataset(dataset_id, dataset_config="default", dataset_split="test"):
    try:
        configs = datasets.get_dataset_config_names(dataset_id)
    except Exception:
        # Dataset may not exist
        return None, dataset_config, dataset_split

    if dataset_config not in configs:
        # Need to choose dataset subset (config)
        return dataset_id, configs, dataset_split

    ds = datasets.load_dataset(dataset_id, dataset_config)

    if isinstance(ds, datasets.DatasetDict):
        # Need to choose dataset split
        if dataset_split not in ds.keys():
            return dataset_id, None, list(ds.keys())
    elif not isinstance(ds, datasets.Dataset):
        # Unknown type
        return dataset_id, None, None
    return dataset_id, dataset_config, dataset_split

def try_validate(m_id, ppl, dataset_id, dataset_config, dataset_split, column_mapping='{}'):
    # Validate model
    if m_id is None:
        gr.Warning('Model is not accessible. Please set your HF_TOKEN if it is a private model.')
        return (
            gr.update(interactive=False),   # Submit button
            gr.update(visible=True),       # Loading row
            gr.update(visible=False),        # Preview row
            gr.update(visible=False),       # Model prediction input
            gr.update(visible=False),       # Model prediction preview
            gr.update(visible=False),       # Label mapping preview
            gr.update(visible=False),       # feature mapping preview
        )
    if isinstance(ppl, Exception):
        gr.Warning(f'Failed to load model": {ppl}')
        return (
            gr.update(interactive=False),   # Submit button
            gr.update(visible=True),       # Loading row
            gr.update(visible=False),        # Preview row
            gr.update(visible=False),       # Model prediction input
            gr.update(visible=False),       # Model prediction preview
            gr.update(visible=False),       # Label mapping preview
            gr.update(visible=False),       # feature mapping preview
        )

    # Validate dataset
    d_id, config, split = check_dataset(dataset_id=dataset_id, dataset_config=dataset_config, dataset_split=dataset_split)

    dataset_ok = False
    if d_id is None:
        gr.Warning(f'Dataset "{dataset_id}" is not accessible. Please set your HF_TOKEN if it is a private dataset.')
    elif isinstance(config, list):
        gr.Warning(f'Dataset "{dataset_id}" does not have "{dataset_config}" config. Please choose a valid config.')
        config = gr.update(choices=config, value=config[0])
    elif isinstance(split, list):
        gr.Warning(f'Dataset "{dataset_id}" does not have "{dataset_split}" split. Please choose a valid split.')
        split = gr.update(choices=split, value=split[0])
    else:
        dataset_ok = True

    if not dataset_ok:
        return (
            gr.update(interactive=False),   # Submit button
            gr.update(visible=True),       # Loading row
            gr.update(visible=False),        # Preview row
            gr.update(visible=False),       # Model prediction input
            gr.update(visible=False),       # Model prediction preview
            gr.update(visible=False),       # Label mapping preview
            gr.update(visible=False),       # feature mapping preview
        )

    # TODO: Validate column mapping by running once
    prediction_result = None
    id2label_df = None
    if isinstance(ppl, TextClassificationPipeline):
        try:
            print('validating phase, ', column_mapping)
            column_mapping = json.loads(column_mapping)
        except Exception:
            column_mapping = {}

        column_mapping, prediction_input, prediction_result, id2label_df, feature_df = \
            text_classification_fix_column_mapping(column_mapping, ppl, d_id, config, split)

        column_mapping = json.dumps(column_mapping, indent=2)

    if prediction_result is None and id2label_df is not None:
        gr.Warning('The model failed to predict with the first row in the dataset. Please provide column mappings in "Advance" settings.')
        return (
            gr.update(interactive=False),   # Submit button
            gr.update(visible=False),       # Loading row
            gr.update(visible=True),        # Preview row
            gr.update(value=f'**Sample Input**: {prediction_input}', visible=True),       # Model prediction input
            gr.update(visible=False),   # Model prediction preview
            gr.update(value=id2label_df, visible=True, interactive=True),   # Label mapping preview
            gr.update(value=feature_df, visible=True, interactive=True),   # feature mapping preview
        )
    elif id2label_df is None:
        gr.Warning('The prediction result does not conform the labels in the dataset. Please provide label mappings in "Advance" settings.')
        return (
            gr.update(interactive=False),   # Submit button
            gr.update(visible=False),       # Loading row
            gr.update(visible=True),        # Preview row
            gr.update(value=f'**Sample Input**: {prediction_input}', visible=True),       # Model prediction input
            gr.update(value=prediction_result, visible=True),   # Model prediction preview
            gr.update(visible=True, interactive=True),   # Label mapping preview
            gr.update(visible=True, interactive=True),   # feature mapping preview
        )

    gr.Info("Model and dataset validations passed. Your can submit the evaluation task.")

    return (
        gr.update(interactive=True),    # Submit button
        gr.update(visible=False),       # Loading row
        gr.update(visible=True),        # Preview row
        gr.update(value=f'**Sample Input**: {prediction_input}', visible=True),       # Model prediction input
        gr.update(value=prediction_result, visible=True),   # Model prediction preview
        gr.update(value=id2label_df, visible=True, interactive=True), # Label mapping preview
        gr.update(value=feature_df, visible=True, interactive=True),   # feature mapping preview
    )


def try_submit(m_id, d_id, config, split, id2label_mapping_dataframe, feature_mapping_dataframe, local):
    label_mapping = {}
    for i, label in id2label_mapping_dataframe["Model Prediction Labels"].items():
        label_mapping.update({str(i): label})
    
    feature_mapping = {}
    for i, feature in feature_mapping_dataframe["Dataset Features"].items():
        feature_mapping.update({feature_mapping_dataframe["Model Input Features"][i]: feature})

    # TODO: Set column mapping for some dataset such as `amazon_polarity`

    if local:
        command = [
            "python",
            "cli.py",
            "--loader", "huggingface",
            "--model", m_id,
            "--dataset", d_id,
            "--dataset_config", config,
            "--dataset_split", split,
            "--hf_token", os.environ.get(HF_WRITE_TOKEN),
            "--discussion_repo", os.environ.get(HF_REPO_ID) or os.environ.get(HF_SPACE_ID),
            "--output_format", "markdown",
            "--output_portal", "huggingface",
            "--feature_mapping", json.dumps(feature_mapping),
            "--label_mapping", json.dumps(label_mapping),
            "--scan_config", "./config.yaml",
        ]

        eval_str = f"[{m_id}]<{d_id}({config}, {split} set)>"
        start = time.time()
        logging.info(f"Start local evaluation on {eval_str}")

        evaluator = subprocess.Popen(
            command,
            cwd=os.path.join(os.path.dirname(os.path.realpath(__file__)), "cicd"),
            stderr=subprocess.STDOUT,
        )
        result = evaluator.wait()

        logging.info(f"Finished local evaluation exit code {result} on {eval_str}: {time.time() - start:.2f}s")

        gr.Info(f"Finished local evaluation exit code {result} on {eval_str}: {time.time() - start:.2f}s")
    else:
        gr.Info("TODO: Submit task to an endpoint")
    
    return gr.update(interactive=True)  # Submit button


with gr.Blocks(theme=theme) as iface:
    with gr.Tab("Text Classification"):
        def check_dataset_and_get_config(dataset_id):
            try:
                configs = datasets.get_dataset_config_names(dataset_id)
                return gr.Dropdown(configs, value=configs[0], visible=True)
            except Exception:
                # Dataset may not exist
                pass

        def check_dataset_and_get_split(dataset_config, dataset_id):
            try:
                splits = list(datasets.load_dataset(dataset_id, dataset_config).keys())
                return gr.Dropdown(splits, value=splits[0], visible=True)
            except Exception as e:
                # Dataset may not exist
                gr.Warning(f"Failed to load dataset {dataset_id} with config {dataset_config}: {e}")
                pass
        
        def gate_validate_btn(model_id, dataset_id, dataset_config, dataset_split, id2label_mapping_dataframe=None, feature_mapping_dataframe=None):
            column_mapping = '{}'
            _, ppl = check_model(model_id=model_id)

            if id2label_mapping_dataframe is not None:
                labels = convert_column_mapping_to_json(id2label_mapping_dataframe.value, label="data")
                features = convert_column_mapping_to_json(feature_mapping_dataframe.value, label="text")
                column_mapping = json.dumps({**labels, **features}, indent=2)

            if check_column_mapping_keys_validity(column_mapping, ppl) is False:
                gr.Warning('Label mapping table has invalid contents. Please check again.')
                return (gr.update(interactive=False), 
                        gr.update(),
                        gr.update(),
                        gr.update(),
                        gr.update(),
                        gr.update(),
                        gr.update())
            else:
                if model_id and dataset_id and dataset_config and dataset_split:
                    return try_validate(model_id, ppl, dataset_id, dataset_config, dataset_split, column_mapping)
                else:
                    return (gr.update(interactive=False), 
                            gr.update(visible=True),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False))
        with gr.Row():
            gr.Markdown('''
                <h1 style="text-align: center;">
                Giskard Evaluator
                </h1>
                Welcome to Giskard Evaluator Space! Get your report immediately by simply input your model id and dataset id below. Follow our leads and improve your model in no time.
                ''')
        with gr.Row():
            run_local = gr.Checkbox(value=True, label="Run in this Space")
            use_inference = read_inference_type('./config.yaml')[0] == 'hf_inference_api'
            run_inference = gr.Checkbox(value=use_inference, label="Run with Inference API")
        
        with gr.Row() as advanced_row:
            selected = read_scanners('./config.yaml')
            scan_config = selected + ['data_leakage']
            scanners = gr.CheckboxGroup(choices=scan_config, value=selected, label='Scan Settings', visible=True)

        with gr.Row():
            model_id_input = gr.Textbox(
                label="Hugging Face model id",
                placeholder="cardiffnlp/twitter-roberta-base-sentiment-latest",
            )

            dataset_id_input = gr.Textbox(
                label="Hugging Face Dataset id",
                placeholder="tweet_eval",
            )
        with gr.Row():
            dataset_config_input = gr.Dropdown(['default'], value='default', label='Dataset Config', visible=False)
            dataset_split_input = gr.Dropdown(['default'], value='default', label='Dataset Split', visible=False)

            dataset_id_input.blur(check_dataset_and_get_config, dataset_id_input, dataset_config_input)
            dataset_id_input.submit(check_dataset_and_get_config, dataset_id_input, dataset_config_input)

            dataset_config_input.blur(
                check_dataset_and_get_split, 
                inputs=[dataset_config_input, dataset_id_input], 
                outputs=[dataset_split_input])
        
        with gr.Row(visible=True) as loading_row:
            gr.Markdown('''
                        <p style="text-align: center;">
                        🚀🐢Please validate your model and dataset first...
                        </p>
                        ''')
            
        with gr.Row(visible=False) as preview_row:
            gr.Markdown('''
                <h1 style="text-align: center;">
                Confirm Pre-processing Details
                </h1>
                Base on your model and dataset, we inferred this label mapping and feature mapping. <b>If the mapping is incorrect, please modify it in the table below.</b>
                ''')
        
        with gr.Row():
            id2label_mapping_dataframe = gr.DataFrame(label="Preview of label mapping", interactive=True, visible=False)
            feature_mapping_dataframe = gr.DataFrame(label="Preview of feature mapping", interactive=True, visible=False)
        with gr.Row():
            example_input = gr.Markdown('Sample Input: ', visible=False)
        
        with gr.Row():
            example_labels = gr.Label(label='Model Prediction Sample', visible=False)
        
        run_btn = gr.Button(
            "Get Evaluation Result",
            variant="primary",
            interactive=False,
            size="lg",
        )

        model_id_input.blur(gate_validate_btn, 
                                inputs=[model_id_input, dataset_id_input, dataset_config_input, dataset_split_input], 
                                outputs=[run_btn, loading_row, preview_row, example_input, example_labels, id2label_mapping_dataframe, feature_mapping_dataframe])
        dataset_id_input.blur(gate_validate_btn, 
                                inputs=[model_id_input, dataset_id_input, dataset_config_input, dataset_split_input], 
                                outputs=[run_btn, loading_row, preview_row, example_input,  example_labels, id2label_mapping_dataframe, feature_mapping_dataframe])
        dataset_config_input.input(gate_validate_btn,
                                inputs=[model_id_input, dataset_id_input, dataset_config_input, dataset_split_input],
                                outputs=[run_btn, loading_row, preview_row, example_input, example_labels, id2label_mapping_dataframe, feature_mapping_dataframe])
        dataset_split_input.input(gate_validate_btn,
                                inputs=[model_id_input, dataset_id_input, dataset_config_input, dataset_split_input],
                                outputs=[run_btn, loading_row, preview_row, example_input, example_labels, id2label_mapping_dataframe, feature_mapping_dataframe])
        id2label_mapping_dataframe.input(gate_validate_btn,
                                inputs=[model_id_input, dataset_id_input, dataset_config_input, dataset_split_input, id2label_mapping_dataframe, feature_mapping_dataframe],
                                outputs=[run_btn, loading_row, preview_row, example_input, example_labels, id2label_mapping_dataframe, feature_mapping_dataframe])
        feature_mapping_dataframe.input(gate_validate_btn,
                                inputs=[model_id_input, dataset_id_input, dataset_config_input, dataset_split_input, id2label_mapping_dataframe, feature_mapping_dataframe],
                                outputs=[run_btn, loading_row, preview_row, example_input, example_labels, id2label_mapping_dataframe, feature_mapping_dataframe])
        scanners.change(write_scanners, inputs=scanners)
        run_inference.change(
            write_inference_type,
            inputs=[run_inference]
        )

        run_btn.click(
            try_submit,
            inputs=[
                model_id_input,
                dataset_id_input,
                dataset_config_input,
                dataset_split_input,
                id2label_mapping_dataframe,
                feature_mapping_dataframe,
                run_local,
            ],
            outputs=[
                run_btn,
            ],
        )

    with gr.Tab("More"):
        pass

if __name__ == "__main__":
    iface.queue(max_size=20).launch()