first try

app.py

@@ -0,0 +1,5 @@
+import evaluate
+from gradio_tst import launch_gradio_widget2
+
+module = evaluate.load("textgen_evaluator.py")
+launch_gradio_widget2(module)

gradio_tst.py

@@ -0,0 +1,130 @@
+import json
+import os
+import re
+import sys
+from pathlib import Path
+
+import numpy as np
+from datasets import Value
+
+import logging
+
+
+
+REGEX_YAML_BLOCK = re.compile(r"---[\n\r]+([\S\s]*?)[\n\r]+---[\n\r]")
+
+
+def infer_gradio_input_types(feature_types):
+    """
+    Maps metric feature types to input types for gradio Dataframes:
+        - float/int -> numbers
+        - string -> strings
+        - any other -> json
+    Note that json is not a native gradio type but will be treated as string that
+    is then parsed as a json.
+    """
+    input_types = []
+    for feature_type in feature_types:
+        input_type = "json"
+        if isinstance(feature_type, Value):
+            if feature_type.dtype.startswith("int") or feature_type.dtype.startswith("float"):
+                input_type = "number"
+            elif feature_type.dtype == "string":
+                input_type = "str"
+        input_types.append(input_type)
+    return input_types
+
+
+def json_to_string_type(input_types):
+    """Maps json input type to str."""
+    return ["str" if i == "json" else i for i in input_types]
+
+
+def parse_readme(filepath):
+    """Parses a repositories README and removes"""
+    if not os.path.exists(filepath):
+        return "No README.md found."
+    with open(filepath, "r") as f:
+        text = f.read()
+        match = REGEX_YAML_BLOCK.search(text)
+        if match:
+            text = text[match.end() :]
+    return text
+
+
+def parse_gradio_data(data, input_types):
+    """Parses data from gradio Dataframe for use in metric."""
+    metric_inputs = {}
+    data.replace("", np.nan, inplace=True)
+    data.dropna(inplace=True)
+    for feature_name, input_type in zip(data, input_types):
+        if input_type == "json":
+            metric_inputs[feature_name] = [json.loads(d) for d in data[feature_name].to_list()]
+        elif input_type == "str":
+            metric_inputs[feature_name] = [d.strip('"') for d in data[feature_name].to_list()]
+        else:
+            metric_inputs[feature_name] = data[feature_name]
+    return metric_inputs
+
+
+def parse_test_cases(test_cases, feature_names, input_types):
+    """
+    Parses test cases to be used in gradio Dataframe. Note that an apostrophe is added
+    to strings to follow the format in json.
+    """
+    if len(test_cases) == 0:
+        return None
+    examples = []
+    for test_case in test_cases:
+        parsed_cases = []
+        for feat, input_type in zip(feature_names, input_types):
+            if input_type == "json":
+                parsed_cases.append([str(element) for element in test_case[feat]])
+            elif input_type == "str":
+                parsed_cases.append(['"' + element + '"' for element in test_case[feat]])
+            else:
+                parsed_cases.append(test_case[feat])
+        examples.append([list(i) for i in zip(*parsed_cases)])
+    return examples
+
+
+def launch_gradio_widget2(metric):
+    """Launches `metric` widget with Gradio."""
+
+    try:
+        import gradio as gr
+    except ImportError as error:
+        logging.error("To create a metric widget with Gradio make sure gradio is installed.")
+        raise error
+
+    local_path = Path(sys.path[0])
+    # if there are several input types, use first as default.
+    if isinstance(metric.features, list):
+        (feature_names, feature_types) = zip(*metric.features[0].items())
+    else:
+        (feature_names, feature_types) = zip(*metric.features.items())
+    gradio_input_types = infer_gradio_input_types(feature_types)
+
+    def compute(data):
+        return metric.compute(**parse_gradio_data(data, gradio_input_types))
+
+    iface = gr.Interface(
+        fn=compute,
+        inputs=gr.Dataframe(
+            headers=feature_names,
+            col_count=len(feature_names),
+            row_count=1,
+            datatype=json_to_string_type(gradio_input_types),
+        ),
+        outputs=gr.Textbox(label=metric.name),
+        description=(
+            metric.info.description + "\nIf this is a text-based metric, make sure to wrap you input in double quotes."
+            " Alternatively you can use a JSON-formatted list as input."
+        ),
+        title=f"Metric: {metric.name}",
+        article=parse_readme(local_path / "README.md"),
+        # TODO: load test cases and use them to populate examples
+        # examples=[parse_test_cases(test_cases, feature_names, gradio_input_types)]
+    )
+
+    iface.launch(share=True)

requirements.txt

@@ -0,0 +1,4 @@
+evaluate
+datasets
+scikit-learn
+gradio

textgen_evaluator.py

@@ -0,0 +1,88 @@
+import datasets
+import evaluate
+
+_CITATION = """\
+@inproceedings{lin-2004-rouge,
+    title = "{ROUGE}: A Package for Automatic Evaluation of Summaries",
+    author = "Lin, Chin-Yew",
+    booktitle = "Text Summarization Branches Out",
+    month = jul,
+    year = "2004",
+    address = "Barcelona, Spain",
+    publisher = "Association for Computational Linguistics",
+    url = "https://www.aclweb.org/anthology/W04-1013",
+    pages = "74--81",
+}
+"""
+
+_DESCRIPTION = """\
+ROUGE, or Recall-Oriented Understudy for Gisting Evaluation, is a set of metrics and a software package used for
+evaluating automatic summarization and machine translation software in natural language processing.
+The metrics compare an automatically produced summary or translation against a reference or a set of references (human-produced) summary or translation.
+
+Note that ROUGE is case insensitive, meaning that upper case letters are treated the same way as lower case letters.
+
+This metrics is a wrapper around Google Research reimplementation of ROUGE:
+https://github.com/google-research/google-research/tree/master/rouge
+"""
+
+_KWARGS_DESCRIPTION = """
+Calculates average rouge scores for a list of hypotheses and references
+Args:
+    predictions: list of predictions to score. Each prediction
+        should be a string with tokens separated by spaces.
+    references: list of reference for each prediction. Each
+        reference should be a string with tokens separated by spaces.
+    rouge_types: A list of rouge types to calculate.
+        Valid names:
+        `"rouge{n}"` (e.g. `"rouge1"`, `"rouge2"`) where: {n} is the n-gram based scoring,
+        `"rougeL"`: Longest common subsequence based scoring.
+        `"rougeLSum"`: rougeLsum splits text using `"\n"`.
+        See details in https://github.com/huggingface/datasets/issues/617
+    use_stemmer: Bool indicating whether Porter stemmer should be used to strip word suffixes.
+    use_aggregator: Return aggregates if this is set to True
+Returns:
+    rouge1: rouge_1 (precision, recall, f1),
+    rouge2: rouge_2 (precision, recall, f1),
+    rougeL: rouge_l (precision, recall, f1),
+    rougeLsum: rouge_lsum (precision, recall, f1)
+"""
+
+
+@datasets.utils.file_utils.add_start_docstrings(
+    _DESCRIPTION, _KWARGS_DESCRIPTION
+)
+class TextGenEvaluatorTest(datasets.Metric):
+    def _info(self):
+        return datasets.MetricInfo(
+            description=_DESCRIPTION,
+            citation=_CITATION,
+            inputs_description=_KWARGS_DESCRIPTION,
+            features=datasets.Features(
+                {
+                    "predictions": datasets.Value("string"),
+                    "references": datasets.Value("string"),
+                }
+            ),
+            codebase_urls=[
+                "https://github.com/google-research/google-research/tree/master/rouge"
+            ],
+            reference_urls=[
+                "https://en.wikipedia.org/wiki/ROUGE_(metric)",
+                "https://github.com/google-research/google-research/tree/master/rouge",
+            ],
+        )
+
+    def _compute(self, predictions, references):
+
+        rouge_score = evaluate.load("rouge")
+
+        scores = rouge_score.compute(
+            predictions=predictions, references=references
+        )
+        bleu_score = evaluate.load("bleu")
+        results = bleu_score.compute(
+            predictions=predictions, references=references
+        )
+
+        return {"ROUGE": scores, "BLEU": results}