fix bug

游雁

2024-06-09 1d27a1507b7a98d3d957f984bbab7e14523181fb

fix bug

 examples/industrial_data_pretraining/llm_asr/demo_speech2text.py

@@ -3,20 +3,37 @@
# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
#  MIT License  (https://opensource.org/licenses/MIT)

import json
import os
import sys

from funasr import AutoModel

model = AutoModel(
    model="/nfs/beinian.lzr/workspace/GPT-4o/Exp/exp6/4m-8gpu/exp6_speech2text_0607_linear_ddp",
)

ckpt_dir = "/nfs/beinian.lzr/workspace/GPT-4o/Exp/exp6/5m-8gpu/exp6_speech2text_linear_ddp_0609"
ckpt_id = "model.pt.ep0.90000"
jsonl = (
    "/nfs/beinian.lzr/workspace/GPT-4o/Data/Speech2Text/TestData/aishell1_test_speech2text.jsonl"
)
output_dir = f"{os.path.join(ckpt_dir, ckpt_id)}"

ckpt_dir = sys.argv[1]
ckpt_id = sys.argv[2]
jsonl = sys.argv[3]
output_dir = sys.argv[4]
device = sys.argv[5]

model = AutoModel(
    model=ckpt_dir,
    init_param=f"{os.path.join(ckpt_dir, ckpt_id)}",
    output_dir=output_dir,
    device=device,
)


with open(jsonl, "r") as f:
    lines = f.readlines()

tearchforing = True
tearchforing = False
for i, line in enumerate(lines):
    data_dict = json.loads(line.strip())
    data = data_dict["messages"]

 examples/industrial_data_pretraining/llm_asr/demo_speech2text.sh

New file
@@ -0,0 +1,63 @@



ckpt_dir="/nfs/zhifu.gzf/ckpt/saves/qwen_1.5_7b/full/sft/asr_tts_text_exp1_ds_z3/checkpoint-11000"
ckpt_id="model.pt.ep0.90000"
jsonl_dir="/nfs/beinian.lzr/workspace/GPT-4o/Data/Speech2Text/TestData"
out_dir="${ckpt_dir}/asr"
mkdir -p ${out_dir}

device="cuda:0"

for data_set in "librispeech_test_clean_speech2text.jsonl" "librispeech_test_other_speech2text.jsonl"; do
    jsonl=${jsonl_dir}/${data_set}
    output_dir=${out_dir}/${data_set}

    pred_file=${out_dir}/${data_set}/1best_recog/text_tn
    ref_file=${out_dir}/${data_set}/1best_recog/label

    python ./demo_speech2text.py ${ckpt_dir} ${ckpt_id} ${jsonl} ${output_dir} ${device}

    python /mnt/workspace/zhifu.gzf/codebase/FunASR/funasr/metrics/wer.py ++ref_file=${ref_file} ++hyp_file=${pred_file} ++cer_file=${pred_file}.cer ++cn_postprocess=false

done


for data_set in "aishell1_test_speech2text.jsonl" "aishell2_ios_test_speech2text.jsonl" "librispeech_test_other_speech2text.jsonl"; do
    jsonl=${jsonl_dir}/${data_set}
    output_dir=${out_dir}/${data_set}

    pred_file=${out_dir}/${data_set}/1best_recog/text_tn
    ref_file=${out_dir}/${data_set}/1best_recog/label

    python ./demo_speech2text.py ${ckpt_dir} ${ckpt_id} ${jsonl} ${output_dir}

    python /mnt/workspace/zhifu.gzf/codebase/FunASR/funasr/metrics/wer.py ++ref_file=${ref_file} ++hyp_file=${pred_file} ++cer_file=${pred_file}.cer ++cn_postprocess=true

done

for data_set in "s2tt_en2zh.v20240605.test.jsonl"; do
    jsonl=${jsonl_dir}/${data_set}
    output_dir=${out_dir}/${data_set}

    pred_file=${out_dir}/${data_set}/1best_recog/text_tn
    ref_file=${out_dir}/${data_set}/1best_recog/label

    python ./demo_speech2text.py ${ckpt_dir} ${ckpt_id} ${jsonl} ${output_dir}

    python /mnt/workspace/zhifu.gzf/codebase/FunASR/funasr/metrics/wer.py ++ref_file=${ref_file} ++hyp_file=${pred_file} ++cer_file=${pred_file}.cer ++cn_postprocess=true

done

for data_set in "s2tt_zh2en.v20240605.test.jsonl"; do
    jsonl=${jsonl_dir}/${data_set}
    output_dir=${out_dir}/${data_set}

    pred_file=${out_dir}/${data_set}/1best_recog/text_tn
    ref_file=${out_dir}/${data_set}/1best_recog/label

    python ./demo_speech2text.py ${ckpt_dir} ${ckpt_id} ${jsonl} ${output_dir}

    python /mnt/workspace/zhifu.gzf/codebase/FunASR/funasr/metrics/wer.py ++ref_file=${ref_file} ++hyp_file=${pred_file} ++cer_file=${pred_file}.cer ++cn_postprocess=true

done

 funasr/models/llm_asr/model.py

@@ -700,10 +700,10 @@
            generated_ids = self.llm.generate(
                inputs_embeds=inputs_embeds, max_new_tokens=kwargs.get("max_length", 512)
            )
            generated_ids = [
                output_ids[len(input_id) :]
                for input_id, output_ids in zip(input_ids, generated_ids)
            ]
            # generated_ids = [
            #     output_ids[len(input_id) :]
            #     for input_id, output_ids in zip(input_ids, generated_ids)
            # ]
            response = tokenizer.batch_decode(
                generated_ids, skip_special_tokens=kwargs.get("skip_special_tokens", True)
            )[0]
@@ -733,7 +733,8 @@
            ibest_writer = self.writer[f"{0 + 1}best_recog"]

        results = []
        result_i = {"key": key[0], "text": response, "label": label}
        response_clean = re.sub("[^\w\s\u3000\u4e00-\u9fff]+", "", response)
        result_i = {"key": key[0], "text": response, "text_tn": response_clean, "label": label}
        if loss is not None:
            result_i["loss"] = loss
        results.append(result_i)
@@ -741,5 +742,6 @@
        if ibest_writer is not None:
            ibest_writer["text"][key[0]] = response
            ibest_writer["label"][key[0]] = label
            ibest_writer["text_tn"][key[0]] = response_clean

        return results, meta_data