FunASR java http client

VirtuosoQ

2024-04-26 e9d2cfc3a134b00f4e98271fbee3838d1ccecbcc

 funasr/models/qwen_audio/model.py

@@ -9,8 +9,7 @@
from torch import nn
import whisper
from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank
from transformers import AutoModelForCausalLM, AutoTokenizer
from transformers.generation import GenerationConfig


from funasr.register import tables

@@ -27,6 +26,8 @@
    """
    def __init__(self, *args, **kwargs):
        super().__init__()
        from transformers import AutoModelForCausalLM, AutoTokenizer
        from transformers.generation import GenerationConfig

        model_or_path = kwargs.get("model_path", "QwenAudio")
        model = AutoModelForCausalLM.from_pretrained(model_or_path, device_map="cpu",
@@ -54,14 +55,13 @@

        meta_data = {}
        # meta_data["batch_data_time"] = -1

        sp_prompt = "<|startoftranscription|><|en|><|transcribe|><|en|><|notimestamps|><|wo_itn|>"
        query = f"<audio>{data_in[0]}</audio>{sp_prompt}"
        prompt = kwargs.get("prompt", "<|startoftranscription|><|en|><|transcribe|><|en|><|notimestamps|><|wo_itn|>")
        query = f"<audio>{data_in[0]}</audio>{prompt}"
        audio_info = self.tokenizer.process_audio(query)
        inputs = self.tokenizer(query, return_tensors='pt', audio_info=audio_info)
        inputs = inputs.to(self.model.device)
        pred = self.model.generate(**inputs, audio_info=audio_info)
        response = tokenizer.decode(pred.cpu()[0], skip_special_tokens=False, audio_info=audio_info)
        response = self.tokenizer.decode(pred.cpu()[0], skip_special_tokens=False, audio_info=audio_info)

        results = []
        result_i = {"key": key[0], "text": response}
@@ -83,7 +83,8 @@
        Modified from https://github.com/QwenLM/Qwen-Audio
        """
        super().__init__()
        
        from transformers import AutoModelForCausalLM, AutoTokenizer
        from transformers.generation import GenerationConfig
        model_or_path = kwargs.get("model_path", "QwenAudio")
        bf16 = kwargs.get("bf16", False)
        fp16 = kwargs.get("fp16", False)