#!/usr/bin/env python3 from datasets import load_dataset, load_metric, Audio, Dataset from transformers import pipeline, AutoFeatureExtractor, AutoTokenizer, AutoConfig, AutoModelForCTC, Wav2Vec2Processor, Wav2Vec2ProcessorWithLM import re import torch import argparse from typing import Dict def log_results(result: Dataset, args: Dict[str, str]): """ DO NOT CHANGE. This function computes and logs the result metrics. """ log_outputs = args.log_outputs dataset_id = "_".join(args.dataset.split("/") + [args.config, args.split]) # load metric wer = load_metric("wer") cer = load_metric("cer") # compute metrics wer_result = wer.compute(references=result["target"], predictions=result["prediction"]) cer_result = cer.compute(references=result["target"], predictions=result["prediction"]) # print & log results result_str = ( f"WER: {wer_result}\n" f"CER: {cer_result}" ) print(result_str) with open(f"{dataset_id}_eval_results.txt", "w") as f: f.write(result_str) # log all results in text file. Possibly interesting for analysis if log_outputs is not None: pred_file = f"log_{dataset_id}_predictions.txt" target_file = f"log_{dataset_id}_targets.txt" with open(pred_file, "w") as p, open(target_file, "w") as t: # mapping function to write output def write_to_file(batch, i): p.write(f"{i}" + "\n") p.write(batch["prediction"] + "\n") t.write(f"{i}" + "\n") t.write(batch["target"] + "\n") result.map(write_to_file, with_indices=True) def normalize_text(text: str, invalid_chars_regex: str, to_lower: bool) -> str: # remove special characters chars_to_ignore_regex = '[\µ\я\ひ\ⱎ\ⱅ\ḥ\ӌ\џ\ŵ\ʋ\λ\φ\χ\г\и\к\п\ц\ч\э\я\џ\ӌ\ቀ\ከ\ጀ\ḥ\牡\津\宇\厳\保\丹\三\む\も\ⱎ\ⱅ\⋅\⊨\↔\ℚ\э\п\к\и\г\|\£\§\·\½\º\ə\ơ\ǀ\ː\ʾ\ˢ\г\и\к\э\п\∨\„\,\?\.\!\—\―\–\;\:\"\‘\»\%\ł\_\€\×\ぬ\;\±\ß\Þ\«\Ø\°\…\”\“\`\ʿ\&\=\+\の\~\(\)\Σ\ı\ጠ\ℵ\馆\青\貴\西\美\甌\杜\术\星\文\扬\北\京\乃\ゔ\や\め\ま\へ\つ\た\う\い\☉\≥\®\/\∞\∆\∅\→\ℰ\ω\ψ\Μ\Θ\Κ\Π\Σ\Ω\α\γ\δ\ε\ζ\η\κ\ι\ν\μ\ρ\ς\σ\τ\υ\ℤ\ℝ\ℂ\ℕ\₽\∈\›\ο\‹\†\}\{\}\_\ደ\Δ\ወ\ي\و\ب\ة\د\ن\ن\ل\را\э\р\п\н\м\к\и\з\ψ\υ\θ\ṭ\ṯ\ḍ\*\^\∼\م\э\п\ǃ\$\Ꝑ]' chars_to_replace_a = '[\ɑ\ạ\ả\ầ\ậ\ắ\ẵ\а\ǎ\ā\ă\ą\á\ã\ä\å]' chars_to_replace_i = '[\ɨ\ị\ı\ī\ĩ\í\ì\і]' chars_to_replace_e = '[\ệ\ễ\ề\ě\ę\ė\ē\е\ế]' chars_to_replace_o = '[\ồ\ộ\ờ\ợ\ő\ö\ŏ\ō\ø\õ\ó\ò\ð\ǫ\ό\ớ\ổ\ố]' chars_to_replace_u = '[\ų\ʉ\ủ\ử\ù\ü\ư\ǔ\ů\ū\ũ\ú\ứ\ụ\ű\ŭ]' chars_to_replace_c = '[\ς\ć\ċ\č\ҫ]' chars_to_replace_y = '[\ÿ\ỳ\ÿ\ý]' chars_to_replace_n = '[\ṇ\ṅ\ǹ\ħ\ñ\ň\ņ\ń]' chars_to_replace_t = '[\ṭ\ț\ť\ţ]' chars_to_replace_s = '[\ṣ\ș\š\ş\ś]' chars_to_replace_q = '[\զ\գ\գ\զ]' chars_to_replace_j = '[\ј]' chars_to_replace_z = '[\ž\ż\ź\ẓ]' chars_to_replace_r = '[\ř]' chars_to_replace_l = '[\ł\ļ\ĺ]' chars_to_replace_k = '[\ķ]' chars_to_replace_g = '[\ġ\ğ]' chars_to_replace_d = '[\đ\ď]' chars_to_replace_b = '[\þ]' chars_to_replace_p = '[\р]' chars_to_replace_apostrophe = '[\´\′\ʼ\’\'\'\ʽ\ʻ\ʾ]' chars_to_replace_tirets = '[\─\−\‐]' if to_lower: text = re.sub(chars_to_ignore_regex, " ", text).lower() text = re.sub(chars_to_replace_a, "a", text) text = re.sub(chars_to_replace_i, "i", text) text = re.sub(chars_to_replace_e, "e", text) text = re.sub(chars_to_replace_o, "o", text) text = re.sub(chars_to_replace_u, "u", text) text = re.sub(chars_to_replace_c, "c", text) text = re.sub(chars_to_replace_y, "y", text) text = re.sub(chars_to_replace_n, "n", text) text = re.sub(chars_to_replace_t, "t", text) text = re.sub(chars_to_replace_s, "s", text) text = re.sub(chars_to_replace_q, "q", text) text = re.sub(chars_to_replace_j, "j", text) text = re.sub(chars_to_replace_z, "z", text) text = re.sub(chars_to_replace_r, "r", text) text = re.sub(chars_to_replace_l, "l", text) text = re.sub(chars_to_replace_k, "k", text) text = re.sub(chars_to_replace_g, "g", text) text = re.sub(chars_to_replace_d, "d", text) text = re.sub(chars_to_replace_b, "b", text) text = re.sub(chars_to_replace_q, "q", text) text = re.sub(chars_to_replace_p, "p", text) text = re.sub(chars_to_replace_apostrophe, "'", text) text = re.sub(chars_to_replace_tirets, "-", text) text = re.sub("β", "beta", text) text = re.sub("æ", "ae", text) text = re.sub("œ", "oe", text) text = re.sub("&", "et", text) text = re.sub("π", "pi", text) text = re.sub("ľ", "l'", text) text = re.sub(r"^\s+|\s+$", "", text) text = re.sub(" +", " ", text) text = re.sub("\n", " ", text) return text def main(args): # load dataset dataset = load_dataset(args.dataset, args.config, split=args.split, use_auth_token=True) # for testing: only process the first two examples as a test # dataset = dataset.select(range(10)) # load processor if args.greedy: processor = Wav2Vec2Processor.from_pretrained(args.model_id) decoder = None else: processor = Wav2Vec2ProcessorWithLM.from_pretrained(args.model_id) decoder = processor.decoder # load processor feature_extractor = processor.feature_extractor tokenizer = processor.tokenizer # resample audio dataset = dataset.cast_column("audio", Audio(sampling_rate=feature_extractor.sampling_rate)) # load eval pipeline if args.device is None: args.device = 0 if torch.cuda.is_available() else -1 config = AutoConfig.from_pretrained(args.model_id) model = AutoModelForCTC.from_pretrained(args.model_id) #asr = pipeline("automatic-speech-recognition", model=args.model_id, device=args.device, tokenizer=tokenizer, feature_extractor=feature_extractor, decoder=decoder) asr = pipeline("automatic-speech-recognition", config=config, model=model, tokenizer=tokenizer, feature_extractor=feature_extractor, decoder=decoder, device=args.device) # build normalizer config tokenizer = AutoTokenizer.from_pretrained(args.model_id) tokens = [x for x in tokenizer.convert_ids_to_tokens(range(0, tokenizer.vocab_size))] special_tokens = [ tokenizer.pad_token, tokenizer.word_delimiter_token, tokenizer.unk_token, tokenizer.bos_token, tokenizer.eos_token, ] non_special_tokens = [x for x in tokens if x not in special_tokens] invalid_chars_regex = f"[^\s{re.escape(''.join(set(non_special_tokens)))}]" normalize_to_lower = False for token in non_special_tokens: if token.isalpha() and token.islower(): normalize_to_lower = True break # map function to decode audio def map_to_pred(batch, args=args, asr=asr, invalid_chars_regex=invalid_chars_regex, normalize_to_lower=normalize_to_lower): prediction = asr(batch["audio"]["array"], chunk_length_s=args.chunk_length_s, stride_length_s=args.stride_length_s) batch["prediction"] = prediction["text"] batch["target"] = normalize_text(batch["sentence"], invalid_chars_regex, normalize_to_lower) return batch # run inference on all examples result = dataset.map(map_to_pred, remove_columns=dataset.column_names) # compute and log_results # do not change function below log_results(result, args) if __name__ == "__main__": parser = argparse.ArgumentParser() parser.add_argument( "--model_id", type=str, required=True, help="Model identifier. Should be loadable with 🤗 Transformers" ) parser.add_argument( "--dataset", type=str, required=True, help="Dataset name to evaluate the `model_id`. Should be loadable with 🤗 Datasets" ) parser.add_argument( "--config", type=str, required=True, help="Config of the dataset. *E.g.* `'en'` for Common Voice" ) parser.add_argument( "--split", type=str, required=True, help="Split of the dataset. *E.g.* `'test'`" ) parser.add_argument( "--chunk_length_s", type=float, default=None, help="Chunk length in seconds. Defaults to None. For long audio files a good value would be 5.0 seconds." ) parser.add_argument( "--stride_length_s", type=float, default=None, help="Stride of the audio chunks. Defaults to None. For long audio files a good value would be 1.0 seconds." ) parser.add_argument( "--log_outputs", action='store_true', help="If defined, write outputs to log file for analysis." ) parser.add_argument( "--greedy", action='store_true', help="If defined, the LM will be ignored during inference." ) parser.add_argument( "--device", type=int, default=None, help="The device to run the pipeline on. -1 for CPU (default), 0 for the first GPU and so on.", ) args = parser.parse_args() main(args)