Merge pull request #325 from alibaba-damo-academy/dev_lzr

Lizerui9926

2023-04-04 a513d1ac95ae6f7d7cbc6dfefc9bbb3f78718163

Merge pull request #325 from alibaba-damo-academy/dev_lzr

fix compute cer problems

 egs/aishell/transformer/utils/compute_wer.py

@@ -45,8 +45,8 @@
           if out_item['wrong'] > 0:
               rst['wrong_sentences'] += 1
           cer_detail_writer.write(hyp_key + print_cer_detail(out_item) + '\n')
           cer_detail_writer.write("ref:" + '\t' + "".join(ref_dict[hyp_key]) + '\n')
           cer_detail_writer.write("hyp:" + '\t' + "".join(hyp_dict[hyp_key]) + '\n')
           cer_detail_writer.write("ref:" + '\t' + " ".join(list(map(lambda x: x.lower(), ref_dict[hyp_key]))) + '\n')
           cer_detail_writer.write("hyp:" + '\t' + " ".join(list(map(lambda x: x.lower(), hyp_dict[hyp_key]))) + '\n')

    if rst['Wrd'] > 0:
        rst['Err'] = round(rst['wrong_words'] * 100 / rst['Wrd'], 2)

 egs_modelscope/asr/data2vec/speech_data2vec_pretrain-paraformer-zh-cn-aishell2-16k/infer.py

@@ -74,7 +74,7 @@
    # If text exists, compute CER
    text_in = os.path.join(params["data_dir"], "text")
    if os.path.exists(text_in):
        text_proc_file = os.path.join(best_recog_path, "token")
        text_proc_file = os.path.join(best_recog_path, "text")
        compute_wer(text_in, text_proc_file, os.path.join(best_recog_path, "text.cer"))



 egs_modelscope/asr/data2vec/speech_data2vec_pretrain-paraformer-zh-cn-aishell2-16k/infer_after_finetune.py

@@ -38,7 +38,7 @@
    # computer CER if GT text is set
    text_in = os.path.join(params["data_dir"], "text")
    if os.path.exists(text_in):
        text_proc_file = os.path.join(decoding_path, "1best_recog/token")
        text_proc_file = os.path.join(decoding_path, "1best_recog/text")
        compute_wer(text_in, text_proc_file, os.path.join(decoding_path, "text.cer"))



 egs_modelscope/asr/data2vec/speech_data2vec_pretrain-zh-cn-aishell2-16k-pytorch/infer.py

@@ -74,7 +74,7 @@
    # If text exists, compute CER
    text_in = os.path.join(params["data_dir"], "text")
    if os.path.exists(text_in):
        text_proc_file = os.path.join(best_recog_path, "token")
        text_proc_file = os.path.join(best_recog_path, "text")
        compute_wer(text_in, text_proc_file, os.path.join(best_recog_path, "text.cer"))



 egs_modelscope/asr/data2vec/speech_data2vec_pretrain-zh-cn-aishell2-16k-pytorch/infer_after_finetune.py

@@ -38,7 +38,7 @@
    # computer CER if GT text is set
    text_in = os.path.join(params["data_dir"], "text")
    if os.path.exists(text_in):
        text_proc_file = os.path.join(decoding_path, "1best_recog/token")
        text_proc_file = os.path.join(decoding_path, "1best_recog/text")
        compute_wer(text_in, text_proc_file, os.path.join(decoding_path, "text.cer"))



 egs_modelscope/asr/paraformer/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/infer.sh

@@ -63,8 +63,8 @@

if [ $stage -le 2 ] && [ $stop_stage -ge 2 ];then
    echo "Computing WER ..."
    python utils/proce_text.py ${output_dir}/1best_recog/text ${output_dir}/1best_recog/text.proc
    python utils/proce_text.py ${data_dir}/text ${output_dir}/1best_recog/text.ref
    cp ${output_dir}/1best_recog/text ${output_dir}/1best_recog/text.proc
    cp ${data_dir}/text ${output_dir}/1best_recog/text.ref
    python utils/compute_wer.py ${output_dir}/1best_recog/text.ref ${output_dir}/1best_recog/text.proc ${output_dir}/1best_recog/text.cer
    tail -n 3 ${output_dir}/1best_recog/text.cer
fi

 egs_modelscope/asr/paraformer/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/infer_after_finetune.py

@@ -34,7 +34,7 @@
    # computer CER if GT text is set
    text_in = os.path.join(params["data_dir"], "text")
    if os.path.exists(text_in):
        text_proc_file = os.path.join(decoding_path, "1best_recog/token")
        text_proc_file = os.path.join(decoding_path, "1best_recog/text")
        compute_wer(text_in, text_proc_file, os.path.join(decoding_path, "text.cer"))



 egs_modelscope/asr/paraformer/speech_paraformer_asr_nat-zh-cn-8k-common-vocab8358-tensorflow1/infer.sh

@@ -63,8 +63,8 @@

if [ $stage -le 2 ] && [ $stop_stage -ge 2 ];then
    echo "Computing WER ..."
    python utils/proce_text.py ${output_dir}/1best_recog/text ${output_dir}/1best_recog/text.proc
    python utils/proce_text.py ${data_dir}/text ${output_dir}/1best_recog/text.ref
    cp ${output_dir}/1best_recog/text ${output_dir}/1best_recog/text.proc
    cp ${data_dir}/text ${output_dir}/1best_recog/text.ref
    python utils/compute_wer.py ${output_dir}/1best_recog/text.ref ${output_dir}/1best_recog/text.proc ${output_dir}/1best_recog/text.cer
    tail -n 3 ${output_dir}/1best_recog/text.cer
fi

 egs_modelscope/asr/paraformer/speech_paraformer_asr_nat-zh-cn-8k-common-vocab8358-tensorflow1/infer_after_finetune.py

@@ -34,7 +34,7 @@
    # computer CER if GT text is set
    text_in = os.path.join(params["data_dir"], "text")
    if os.path.exists(text_in):
        text_proc_file = os.path.join(decoding_path, "1best_recog/token")
        text_proc_file = os.path.join(decoding_path, "1best_recog/text")
        compute_wer(text_in, text_proc_file, os.path.join(decoding_path, "text.cer"))



 egs_modelscope/asr/uniasr/speech_UniASR_asr_2pass-zh-cn-8k-common-vocab3445-pytorch-offline/infer.py

@@ -75,7 +75,7 @@
    # If text exists, compute CER
    text_in = os.path.join(params["data_dir"], "text")
    if os.path.exists(text_in):
        text_proc_file = os.path.join(best_recog_path, "token")
        text_proc_file = os.path.join(best_recog_path, "text")
        compute_wer(text_in, text_proc_file, os.path.join(best_recog_path, "text.cer"))



 egs_modelscope/asr/uniasr/speech_UniASR_asr_2pass-zh-cn-8k-common-vocab3445-pytorch-offline/infer_after_finetune.py

@@ -39,7 +39,7 @@
    # computer CER if GT text is set
    text_in = os.path.join(params["data_dir"], "text")
    if os.path.exists(text_in):
        text_proc_file = os.path.join(decoding_path, "1best_recog/token")
        text_proc_file = os.path.join(decoding_path, "1best_recog/text")
        compute_wer(text_in, text_proc_file, os.path.join(decoding_path, "text.cer"))



 egs_modelscope/asr/uniasr/speech_UniASR_asr_2pass-zh-cn-8k-common-vocab3445-pytorch-online/infer.py

@@ -75,7 +75,7 @@
    # If text exists, compute CER
    text_in = os.path.join(params["data_dir"], "text")
    if os.path.exists(text_in):
        text_proc_file = os.path.join(best_recog_path, "token")
        text_proc_file = os.path.join(best_recog_path, "text")
        compute_wer(text_in, text_proc_file, os.path.join(best_recog_path, "text.cer"))



 egs_modelscope/asr/uniasr/speech_UniASR_asr_2pass-zh-cn-8k-common-vocab3445-pytorch-online/infer_after_finetune.py

@@ -39,7 +39,7 @@
    # computer CER if GT text is set
    text_in = os.path.join(params["data_dir"], "text")
    if os.path.exists(text_in):
        text_proc_file = os.path.join(decoding_path, "1best_recog/token")
        text_proc_file = os.path.join(decoding_path, "1best_recog/text")
        compute_wer(text_in, text_proc_file, os.path.join(decoding_path, "text.cer"))



 funasr/bin/asr_inference_paraformer.py

@@ -797,7 +797,7 @@
                        finish_count += 1
                        # asr_utils.print_progress(finish_count / file_count)
                        if writer is not None:
                            ibest_writer["text"][key] = text_postprocessed
                            ibest_writer["text"][key] = " ".join(word_lists)

                    logging.info("decoding, utt: {}, predictions: {}".format(key, text))
        rtf_avg = "decoding, feature length total: {}, forward_time total: {:.4f}, rtf avg: {:.4f}".format(length_total, forward_time_total, 100 * forward_time_total / (length_total * lfr_factor))

 funasr/bin/asr_inference_paraformer_vad.py

@@ -338,7 +338,7 @@
                    ibest_writer["token"][key] = " ".join(token)
                    ibest_writer["token_int"][key] = " ".join(map(str, token_int))
                    ibest_writer["vad"][key] = "{}".format(vadsegments)
                    ibest_writer["text"][key] = text_postprocessed
                    ibest_writer["text"][key] = " ".join(word_lists)
                    ibest_writer["text_with_punc"][key] = text_postprocessed_punc
                    if time_stamp_postprocessed is not None:
                        ibest_writer["time_stamp"][key] = "{}".format(time_stamp_postprocessed)

 funasr/bin/asr_inference_paraformer_vad_punc.py

@@ -670,7 +670,7 @@
                    ibest_writer["token"][key] = " ".join(token)
                    ibest_writer["token_int"][key] = " ".join(map(str, token_int))
                    ibest_writer["vad"][key] = "{}".format(vadsegments)
                    ibest_writer["text"][key] = text_postprocessed
                    ibest_writer["text"][key] = " ".join(word_lists)
                    ibest_writer["text_with_punc"][key] = text_postprocessed_punc
                    if time_stamp_postprocessed is not None:
                        ibest_writer["time_stamp"][key] = "{}".format(time_stamp_postprocessed)

 funasr/bin/asr_inference_rnnt.py

@@ -738,13 +738,13 @@
                        ibest_writer["rtf"][key] = rtf_cur

                    if text is not None:
                        text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
                        text_postprocessed, word_lists = postprocess_utils.sentence_postprocess(token)
                        item = {'key': key, 'value': text_postprocessed}
                        asr_result_list.append(item)
                        finish_count += 1
                        # asr_utils.print_progress(finish_count / file_count)
                        if writer is not None:
                            ibest_writer["text"][key] = text_postprocessed
                            ibest_writer["text"][key] = " ".join(word_lists)

                    logging.info("decoding, utt: {}, predictions: {}".format(key, text))
        rtf_avg = "decoding, feature length total: {}, forward_time total: {:.4f}, rtf avg: {:.4f}".format(length_total, forward_time_total, 100 * forward_time_total / (length_total * lfr_factor))

 funasr/bin/asr_inference_uniasr.py

@@ -507,13 +507,13 @@
                    ibest_writer["score"][key] = str(hyp.score)
    
                if text is not None:
                    text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
                    text_postprocessed, word_lists = postprocess_utils.sentence_postprocess(token)
                    item = {'key': key, 'value': text_postprocessed}
                    asr_result_list.append(item)
                    finish_count += 1
                    asr_utils.print_progress(finish_count / file_count)
                    if writer is not None:
                        ibest_writer["text"][key] = text_postprocessed
                        ibest_writer["text"][key] = " ".join(word_lists)
        return asr_result_list
    
    return _forward

 funasr/bin/asr_inference_uniasr_vad.py

@@ -507,13 +507,13 @@
                    ibest_writer["score"][key] = str(hyp.score)
    
                if text is not None:
                    text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
                    text_postprocessed, word_lists = postprocess_utils.sentence_postprocess(token)
                    item = {'key': key, 'value': text_postprocessed}
                    asr_result_list.append(item)
                    finish_count += 1
                    asr_utils.print_progress(finish_count / file_count)
                    if writer is not None:
                        ibest_writer["text"][key] = text_postprocessed
                        ibest_writer["text"][key] = " ".join(word_lists)
        return asr_result_list
    
    return _forward

 funasr/utils/compute_wer.py

@@ -45,8 +45,8 @@
           if out_item['wrong'] > 0:
               rst['wrong_sentences'] += 1
           cer_detail_writer.write(hyp_key + print_cer_detail(out_item) + '\n')
           cer_detail_writer.write("ref:" + '\t' + "".join(ref_dict[hyp_key]) + '\n')
           cer_detail_writer.write("hyp:" + '\t' + "".join(hyp_dict[hyp_key]) + '\n')
           cer_detail_writer.write("ref:" + '\t' + " ".join(list(map(lambda x: x.lower(), ref_dict[hyp_key]))) + '\n')
           cer_detail_writer.write("hyp:" + '\t' + " ".join(list(map(lambda x: x.lower(), hyp_dict[hyp_key]))) + '\n')

    if rst['Wrd'] > 0:
        rst['Err'] = round(rst['wrong_words'] * 100 / rst['Wrd'], 2)