From d5309db7f7cca2685cbb9727803d15cda2291c1e Mon Sep 17 00:00:00 2001 From: fenghaojin <157778956+fenghaojin@users.noreply.github.com> Date: Thu, 14 Mar 2024 17:00:16 +0800 Subject: [PATCH] Update decode.py Resolve issues with abnormal output formats and inaccurate error rates --- .../ASR/pruned_transducer_stateless7_streaming/decode.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/egs/aishell/ASR/pruned_transducer_stateless7_streaming/decode.py b/egs/aishell/ASR/pruned_transducer_stateless7_streaming/decode.py index 0e783e92b..a24d8d4d3 100755 --- a/egs/aishell/ASR/pruned_transducer_stateless7_streaming/decode.py +++ b/egs/aishell/ASR/pruned_transducer_stateless7_streaming/decode.py @@ -492,7 +492,7 @@ def save_results( for key, results in results_dict.items(): recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt" results = sorted(results) - store_transcripts(filename=recog_path, texts=results) + store_transcripts(filename=recog_path, texts=results, char_level=True) logging.info(f"The transcripts are stored in {recog_path}") # The following prints out WERs, per-word error statistics and aligned @@ -500,7 +500,7 @@ def save_results( errs_filename = params.res_dir / f"errs-{test_set_name}-{params.suffix}.txt" with open(errs_filename, "w") as f: wer = write_error_stats( - f, f"{test_set_name}-{key}", results, enable_log=True + f, f"{test_set_name}-{key}", results, enable_log=True, compute_CER=True ) test_set_wers[key] = wer