fix: resolve unexpected 'out of memory' issue in multi-GPU setup (#2373)

BienBoy

2025-02-01 c1e365fea09aafda387cac12fdff43d28c598979

 fun_text_processing/text_normalization/run_evaluate.py

@@ -1,5 +1,3 @@


from argparse import ArgumentParser

from fun_text_processing.text_normalization.data_loader_utils import (
@@ -12,18 +10,22 @@
from fun_text_processing.text_normalization.normalize import Normalizer


'''
"""
Runs Evaluation on data in the format of : <semiotic class>\t<unnormalized text>\t<`self` if trivial class or normalized text>
like the Google text normalization data https://www.kaggle.com/richardwilliamsproat/text-normalization-for-english-russian-and-polish
'''
"""


def parse_args():
    parser = ArgumentParser()
    parser.add_argument("--input", help="input file path", type=str)
    parser.add_argument("--lang", help="language", choices=['en'], default="en", type=str)
    parser.add_argument("--lang", help="language", choices=["en"], default="en", type=str)
    parser.add_argument(
        "--input_case", help="input capitalization", choices=["lower_cased", "cased"], default="cased", type=str
        "--input_case",
        help="input capitalization",
        choices=["lower_cased", "cased"],
        default="cased",
        type=str,
    )
    parser.add_argument(
        "--cat",
@@ -33,7 +35,9 @@
        default=None,
        choices=known_types,
    )
    parser.add_argument("--filter", action='store_true', help="clean data for normalization purposes")
    parser.add_argument(
        "--filter", action="store_true", help="clean data for normalization purposes"
    )
    return parser.parse_args()


@@ -41,7 +45,7 @@
    # Example usage:
    # python run_evaluate.py --input=<INPUT> --cat=<CATEGORY> --filter
    args = parse_args()
    if args.lang == 'en':
    if args.lang == "en":
        from fun_text_processing.text_normalization.en.clean_eval_data import filter_loaded_data
    file_path = args.input
    normalizer = Normalizer(input_case=args.input_case, lang=args.lang)
@@ -76,30 +80,35 @@
            preds=tokens_prediction, labels=tokens_normalized, input=tokens_un_normalized
        )
        print("  - Accuracy: " + str(token_accuracy[token_type]))
    token_count_per_type = {token_type: len(tokens_per_type[token_type][0]) for token_type in tokens_per_type}
    token_count_per_type = {
        token_type: len(tokens_per_type[token_type][0]) for token_type in tokens_per_type
    }
    token_weighted_accuracy = [
        token_count_per_type[token_type] * accuracy for token_type, accuracy in token_accuracy.items()
        token_count_per_type[token_type] * accuracy
        for token_type, accuracy in token_accuracy.items()
    ]
    print("- Accuracy: " + str(sum(token_weighted_accuracy) / sum(token_count_per_type.values())))
    print(" - Total: " + str(sum(token_count_per_type.values())), '\n')
    print(" - Total: " + str(sum(token_count_per_type.values())), "\n")

    print(" - Total: " + str(sum(token_count_per_type.values())), '\n')
    print(" - Total: " + str(sum(token_count_per_type.values())), "\n")

    for token_type in token_accuracy:
        if token_type not in known_types:
            raise ValueError("Unexpected token type: " + token_type)

    if args.category is None:
        c1 = ['Class', 'sent level'] + known_types
        c2 = ['Num Tokens', len(sentences_normalized)] + [
            token_count_per_type[known_type] if known_type in tokens_per_type else '0' for known_type in known_types
        c1 = ["Class", "sent level"] + known_types
        c2 = ["Num Tokens", len(sentences_normalized)] + [
            token_count_per_type[known_type] if known_type in tokens_per_type else "0"
            for known_type in known_types
        ]
        c3 = ['Normalization', sentences_accuracy] + [
            token_accuracy[known_type] if known_type in token_accuracy else '0' for known_type in known_types
        c3 = ["Normalization", sentences_accuracy] + [
            token_accuracy[known_type] if known_type in token_accuracy else "0"
            for known_type in known_types
        ]

        for i in range(len(c1)):
            print(f'{str(c1[i]):10s} | {str(c2[i]):10s} | {str(c3[i]):5s}')
            print(f"{str(c1[i]):10s} | {str(c2[i]):10s} | {str(c3[i]):5s}")
    else:
        print(f'numbers\t{token_count_per_type[args.category]}')
        print(f'Normalization\t{token_accuracy[args.category]}')
        print(f"numbers\t{token_count_per_type[args.category]}")
        print(f"Normalization\t{token_accuracy[args.category]}")