speech_recognition/FunASR
1
0
Fork 0
mirror of https://github.com/modelscope/FunASR synced 2025-09-15 14:48:36 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

update

Browse Source
This commit is contained in:
游雁 2024-07-11 11:41:03 +08:00
parent 9c5ea355ac
commit 10e95b07a1
1 changed files with 8 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
examples/industrial_data_pretraining/llm_asr/app_chatbot_audio_audio.py
View File

@ -22,6 +22,8 @@ import re
import os
import sys
import numpy as np
if len(sys.argv) > 1:
ckpt_dir = sys.argv[1]
ckpt_id = sys.argv[2]
@ -47,6 +49,9 @@ init_param_ckpt = f"{os.path.join(ckpt_dir, ckpt_id)}"
flow_init = "/data/zhifu.gzf/init_model/cosyvoice_flow_matching_for_streaming_with_prompt_random_cut_sft_zh_0630_25hz_1/60epoch.pth.prefix"
vocoder_init = "/data/zhifu.gzf/init_model/hiftnet_1400k_cvt/model.pth.prefix"
init_param = f"{init_param},{init_param_ckpt},{flow_init},{vocoder_init}"
spk_emb = np.load(
"/data/zhifu.gzf/init_model/cosyvoice_flow_matching_for_streaming_with_prompt_random_cut_sft_zh_0630_25hz_1/xvec/xiaoxia.npy"
)
model_llm = AutoModel(
model=ckpt_dir,
@ -119,12 +124,14 @@ def model_inference(input_wav, text_inputs, state, turn_num, history):
res = model_llm.generate(
input=[contents_i],
spk_emb=spk_emb,
tearchforing=False,
cache={},
key="test_demo",
)
print(res)
res_text = res[0]["text"]
history[-1][1] = gr.Audio((16000, res[0]["wav"].flatten()), autoplay=True)
history[-1][1] = gr.Audio((22050, res[0]["wav"].cpu().flatten().numpy()), autoplay=True)
out_his = state.get("out", "")
out = f"{out_his}" f"<br><br>" f"Q: {asr_out}" f"<br>" f"A: {res_text}"
# out = f"{res}"