Dev gzf llm (#1493)

zhifu gao

2024-03-14 35b1c051f6db3649a818547902497d219c871b84

Dev gzf llm (#1493)

* update

* update

* update

* update onnx

* update with main (#1492)

* contextual&seaco ONNX export (#1481)

* contextual&seaco ONNX export

* update ContextualEmbedderExport2

* update ContextualEmbedderExport2

* update code

* onnx (#1482)

* qwenaudio qwenaudiochat

* qwenaudio qwenaudiochat

* whisper

* whisper

* llm

* llm

* llm

* llm

* llm

* llm

* llm

* llm

* export onnx

* export onnx

* export onnx

* dingding

* dingding

* llm

* doc

* onnx

* onnx

* onnx

* onnx

* onnx

* onnx

* v1.0.15

* qwenaudio

* qwenaudio

* issue doc

* update

* update

* bugfix

* onnx

* update export calling

* update codes

* remove useless code

* update code

---------

Co-authored-by: zhifu gao <zhifu.gzf@alibaba-inc.com>

* acknowledge

---------

Co-authored-by: Shi Xian <40013335+R1ckShi@users.noreply.github.com>

* update onnx

* update onnx

---------

Co-authored-by: Shi Xian <40013335+R1ckShi@users.noreply.github.com>

 funasr/auto/auto_model.py

@@ -164,22 +164,23 @@
            tokenizer_class = tables.tokenizer_classes.get(tokenizer)
            tokenizer_conf = kwargs.get("tokenizer_conf", {})
            tokenizer = tokenizer_class(**tokenizer_conf)
            kwargs["tokenizer"] = tokenizer
            

            kwargs["token_list"] = tokenizer.token_list if hasattr(tokenizer, "token_list") else None
            kwargs["token_list"] = tokenizer.get_vocab() if hasattr(tokenizer, "get_vocab") else kwargs["token_list"]
            vocab_size = len(kwargs["token_list"]) if kwargs["token_list"] is not None else -1
        else:
            vocab_size = -1
        kwargs["tokenizer"] = tokenizer
        
        # build frontend
        frontend = kwargs.get("frontend", None)
        kwargs["input_size"] = None
        if frontend is not None:
            frontend_class = tables.frontend_classes.get(frontend)
            frontend = frontend_class(**kwargs["frontend_conf"])
            kwargs["frontend"] = frontend
            kwargs["input_size"] = frontend.output_size() if hasattr(frontend, "output_size") else None
        
        kwargs["frontend"] = frontend
        # build model
        model_class = tables.model_classes.get(kwargs["model"])
        model = model_class(**kwargs, **kwargs.get("model_conf", {}), vocab_size=vocab_size)
@@ -469,13 +470,19 @@
        #                      f"time_escape_all: {time_escape_total_all_samples:0.3f}")
        return results_ret_list

    def export(self, input=None,
               type : str = "onnx",
               quantize: bool = False,
               fallback_num: int = 5,
               calib_num: int = 100,
               opset_version: int = 14,
               **cfg):
    def export(self, input=None, **cfg):
    
        """
        
        :param input:
        :param type:
        :param quantize:
        :param fallback_num:
        :param calib_num:
        :param opset_version:
        :param cfg:
        :return:
        """
    
        device = cfg.get("device", "cpu")
        model = self.model.to(device=device)
@@ -485,7 +492,7 @@
        del kwargs["model"]
        model.eval()

        batch_size = 1
        type = kwargs.get("type", "onnx")

        key_list, data_list = prepare_data_iterator(input, input_len=None, data_type=kwargs.get("data_type", None), key=None)

@@ -495,19 +502,11 @@
                export_dir = export_utils.export_onnx(
                                        model=model,
                                        data_in=data_list,
                                        quantize=quantize,
                                        fallback_num=fallback_num,
                                        calib_num=calib_num,
                                        opset_version=opset_version,
                                        **kwargs)
            else:
                export_dir = export_utils.export_torchscripts(
                                        model=model,
                                        data_in=data_list,
                                        quantize=quantize,
                                        fallback_num=fallback_num,
                                        calib_num=calib_num,
                                        opset_version=opset_version,
                                        **kwargs)

        return export_dir

 funasr/bin/train_llm.py

@@ -26,7 +26,7 @@
# from funasr.tokenizer.build_tokenizer import build_tokenizer
# from funasr.tokenizer.token_id_converter import TokenIDConverter
# from funasr.tokenizer.funtoken import build_tokenizer

from funasr import AutoModel

@hydra.main(config_name=None, version_base=None)
def main_hydra(kwargs: DictConfig):
@@ -60,6 +60,16 @@
    if use_ddp or use_fsdp:
        dist.init_process_group(backend=kwargs.get("backend", "nccl"), init_method='env://')
        torch.cuda.set_device(local_rank)
        
    device = kwargs.get("device", "cpu")
    kwargs["device"] = "cpu"
    model = AutoModel(**kwargs)
    kwargs["device"] = device
    model = model.model
    tokenizer = kwargs["tokenizer"]
    frontend = kwargs["frontend"]
    
    
    
    # save config.yaml
    if (use_ddp or use_fsdp) and dist.get_rank() == 0 or not (use_ddp or use_fsdp) and local_rank == 0:
@@ -68,28 +78,9 @@
        OmegaConf.save(config=kwargs, f=yaml_file)
        logging.info("config.yaml is saved to: %s", yaml_file)

    tokenizer = kwargs.get("tokenizer", None)
    if tokenizer is not None:
        tokenizer_class = tables.tokenizer_classes.get(tokenizer)
        tokenizer = tokenizer_class(**kwargs["tokenizer_conf"])
        kwargs["tokenizer"] = tokenizer

    
    # build frontend if frontend is none None
    frontend = kwargs.get("frontend", None)
    if frontend is not None:
        frontend_class = tables.frontend_classes.get(frontend)
        frontend = frontend_class(**kwargs["frontend_conf"])
        kwargs["frontend"] = frontend
        kwargs["input_size"] = frontend.output_size()


    # build model
    model_class = tables.model_classes.get(kwargs["model"])
    vocab_size = len(tokenizer.token_list) if hasattr(tokenizer, "token_list") else None
    vocab_size = len(tokenizer.get_vocab()) if hasattr(tokenizer, "get_vocab") else vocab_size
    model = model_class(**kwargs, **kwargs["model_conf"], vocab_size=vocab_size)


    

    # init_param
    init_param = kwargs.get("init_param", None)

 funasr/datasets/audio_datasets/datasets.py

@@ -90,7 +90,7 @@

        for key, data_list in outputs.items():
            if isinstance(data_list[0], torch.Tensor):
                if data_list[0].dtype == torch.int64:
                if data_list[0].dtype == torch.int64 or data_list[0].dtype == torch.int32:
    
                    pad_value = self.int_pad_value
                else:
@@ -192,7 +192,7 @@

        for key, data_list in outputs.items():
            if isinstance(data_list[0], torch.Tensor):
                if data_list[0].dtype == torch.int64:
                if data_list[0].dtype == torch.int64 or data_list[0].dtype == torch.int32:
                    pad_value = self.int_pad_value
                else:
                    pad_value = self.float_pad_value

 funasr/datasets/llm_datasets/datasets.py

@@ -5,8 +5,8 @@
from funasr.utils.load_utils import extract_fbank, load_audio_text_image_video


@tables.register("dataset_classes", "AudioLLMDataset")
class AudioLLMDataset(torch.utils.data.Dataset):
@tables.register("dataset_classes", "AudioLLMNARDataset")
class AudioLLMNARDataset(torch.utils.data.Dataset):
    """
    AudioLLMDataset
    """
@@ -38,8 +38,8 @@
        self.tokenizer = tokenizer

        self.float_pad_value = float_pad_value
        self.prompt = kwargs.get("prompt", "Transcribe speech to text.")
        self.prompt_pre = "USER: \nINSTRUCTION: {}\nINPUT: ".format(self.prompt)  # "USER: \nINSTRUCTION: {}\nnINPUT: {}\nASSISTANT: "
        self.prompt = kwargs.get("prompt", "Please copy the following text.")
        self.prompt_pre = "USER: \nINSTRUCTION: {}\nINPUT: ".format(self.prompt)  # "USER: \nINSTRUCTION: {}\nINPUT: {}\nASSISTANT: "
        self.prompt_af = ""
        self.IGNORE_INDEX = kwargs.get("IGNORE_INDEX", -100)
        self.int_pad_value = self.IGNORE_INDEX
@@ -72,29 +72,170 @@
        
        
        prompt_ids_pre = self.tokenizer.encode(self.prompt_pre) # [bos,prompt]
        prompt_pre_length = len(prompt_ids_pre)
        prompt_ids_length = len(prompt_ids_pre)
        
        # bos prompt audio bos target
        # prompt_input = "{}{}".format(self.prompt_pre, target)
        # prompt_input_ids = self.tokenizer.encode(prompt_input) #[bos, prompt, input]
        # audio_length = len(prompt_input_ids) - prompt_ids_length
        target_ids = self.tokenizer.encode(target)
        if target_ids[0] == self.tokenizer.bos_token_id:
            target_ids = target_ids[1:]
        target_ids_length = len(target_ids)
        audio_length = target_ids_length
        input_ids = prompt_ids_pre + target_ids + [self.tokenizer.pad_token_id] + target_ids #[bos, prompt, input, pad, target]
        input_ids = torch.tensor(copy.deepcopy(input_ids), dtype=torch.int64) #[bos, prompt, input, pad, target]
        input_ids[prompt_ids_length:prompt_ids_length+audio_length] = -1  # [bos, prompt,-1, pad, target] # it is no need, only for check
        attention_mask = input_ids.ge(-1) # [true, true, true, true, true], length mask
        
        # bos prompt audio target eos
        # prompt_answer = "{}{}".format(self.prompt_pre, target)
        # prompt_answer_ids = self.tokenizer.encode(prompt_answer) #[bos, prompt, input]
        # answer_length = len(prompt_answer_ids) - prompt_ids_length
        target_ids = self.tokenizer.encode(target)
        if target_ids[0] == self.tokenizer.bos_token_id:
            target_ids = target_ids[1:]
        # target_ids_length = len(target_ids)
        labels_ids = prompt_ids_pre + target_ids + target_ids + [self.tokenizer.eos_token_id] # [bos, prompt, input, target, eos]
        labels_ids = torch.tensor(copy.deepcopy(labels_ids), dtype=torch.int64)  # [bos, prompt, input, target, eos]
        labels_ids[:prompt_ids_length] = -1  # [-1, -1, input, target, eos]
        label_mask = labels_ids.ge(0)  # [false, false, true, true, true], length mask
        labels_ids[~label_mask] = self.IGNORE_INDEX  # [-1, -1, input, target, eos]
        
        audio_mask = [0] * prompt_ids_length + [1] * audio_length + [0] * target_ids_length + [0] # [0, 0, 1, 0, 0]
        audio_mask = torch.tensor(audio_mask, dtype=torch.float32)
        
        ids = target_ids #self.tokenizer.encode(target) # token ids is different from labels_ids
        text = torch.tensor(ids, dtype=torch.int64)
        text_lengths = torch.tensor([len(ids)], dtype=torch.int32)

        prompt_bos_length = torch.tensor([len(prompt_ids_pre)], dtype=torch.int32)
        
        return {"speech": speech,
                "speech_lengths": speech_lengths,
                "text": text,
                "text_lengths": text_lengths,
                "input_ids": input_ids,
                "attention_mask": attention_mask,
                "labels_ids": labels_ids,
                "label_mask": label_mask,
                "audio_mask": audio_mask,
                "prompt_bos_length":  prompt_bos_length,
                }
    
    
    def collator(self, samples: list=None):
        outputs = {}
        for sample in samples:
            for key in sample.keys():
                if key not in outputs:
                    outputs[key] = []
                outputs[key].append(sample[key])

        for key, data_list in outputs.items():
            if isinstance(data_list[0], torch.Tensor):
                if data_list[0].dtype == torch.int64 or data_list[0].dtype == torch.int32:
    
                    pad_value = self.int_pad_value
                else:
                    pad_value = self.float_pad_value
                
                outputs[key] = torch.nn.utils.rnn.pad_sequence(data_list, batch_first=True, padding_value=pad_value)
        return outputs


@tables.register("dataset_classes", "AudioLLMDataset")
class AudioLLMDataset(torch.utils.data.Dataset):
    """
    AudioLLMDataset
    """
    
    def __init__(self,
                 path,
                 index_ds: str = None,
                 frontend=None,
                 tokenizer=None,
                 int_pad_value: int = -1,
                 float_pad_value: float = 0.0,
                 **kwargs):
        super().__init__()
        index_ds_class = tables.index_ds_classes.get(index_ds)
        self.index_ds = index_ds_class(path, **kwargs)
        preprocessor_speech = kwargs.get("preprocessor_speech", None)
        if preprocessor_speech:
            preprocessor_speech_class = tables.preprocessor_classes.get(preprocessor_speech)
            preprocessor_speech = preprocessor_speech_class(**kwargs.get("preprocessor_speech_conf", {}))
        self.preprocessor_speech = preprocessor_speech
        preprocessor_text = kwargs.get("preprocessor_text", None)
        if preprocessor_text:
            preprocessor_text_class = tables.preprocessor_classes.get(preprocessor_text)
            preprocessor_text = preprocessor_text_class(**kwargs.get("preprocessor_text_conf", {}))
        self.preprocessor_text = preprocessor_text
        
        self.frontend = frontend
        self.fs = 16000 if frontend is None else frontend.fs
        self.data_type = "sound"
        self.tokenizer = tokenizer
        
        self.float_pad_value = float_pad_value
        self.prompt = kwargs.get("prompt", "Transcribe speech to text.")
        self.prompt_pre = "USER: \nINSTRUCTION: {}\nINPUT: ".format(
            self.prompt)  # "USER: \nINSTRUCTION: {}\nnINPUT: {}\nASSISTANT: "
        self.prompt_af = ""
        self.IGNORE_INDEX = kwargs.get("IGNORE_INDEX", -100)
        self.int_pad_value = self.IGNORE_INDEX
    
    def get_source_len(self, index):
        item = self.index_ds[index]
        return self.index_ds.get_source_len(item)
    
    def get_target_len(self, index):
        item = self.index_ds[index]
        return self.index_ds.get_target_len(item)
    
    def __len__(self):
        return len(self.index_ds)
    
    def __getitem__(self, index):
        item = self.index_ds[index]
        # import pdb;
        # pdb.set_trace()
        source = item["source"]
        data_src = load_audio_text_image_video(source, fs=self.fs)
        if self.preprocessor_speech:
            data_src = self.preprocessor_speech(data_src, fs=self.fs)
        speech, speech_lengths = extract_fbank(data_src, data_type=self.data_type, frontend=self.frontend,
                                               is_final=True)  # speech: [b, T, d]
        speech = speech.squeeze(0)
        
        target = item["target"]
        if self.preprocessor_text:
            target = self.preprocessor_text(target)
        
        prompt_ids_pre = self.tokenizer.encode(self.prompt_pre)  # [bos,prompt]
        prompt_ids_length = len(prompt_ids_pre)
        
        prompt_input = "{}{}".format(self.prompt_pre, target)
        prompt_input_ids = self.tokenizer.encode(prompt_input)
        audio_length = len(prompt_input_ids) - prompt_pre_length
        audio_length = len(prompt_input_ids) - prompt_ids_length
        input_ids = prompt_input_ids + [self.tokenizer.pad_token_id]
        input_ids = torch.tensor(input_ids, dtype=torch.int64) #[bos, prompt, input, pad]
        input_ids[prompt_pre_length:] = -1  # [bos, prompt,-1,-1]
        attention_mask = input_ids.ge(-1) # [true, true, true, true], length mask

        input_ids = torch.tensor(input_ids, dtype=torch.int64)  # [bos, prompt, input, pad]
        input_ids[prompt_ids_length:] = -1  # [bos, prompt,-1,-1]
        attention_mask = input_ids.ge(-1)  # [true, true, true, true], length mask
        
        prompt_answer = "{}{}".format(self.prompt_pre, target)
        prompt_answer_ids = self.tokenizer.encode(prompt_answer)
        answer_length = len(prompt_answer_ids) - prompt_pre_length
        answer_length = len(prompt_answer_ids) - prompt_ids_length
        labels_ids = copy.deepcopy(prompt_input_ids) + [self.tokenizer.eos_token_id]
        labels_ids = torch.tensor(labels_ids, dtype=torch.int64)  # [bos, prompt, input, eos]
        labels_ids[:prompt_pre_length] = -1  # [-1, -1, input, eos]
        labels_ids[:prompt_ids_length] = -1  # [-1, -1, input, eos]
        label_mask = labels_ids.ge(0)  # [False,False,True,True]
        labels_ids[~label_mask] = self.IGNORE_INDEX  # [-100,-100,input,eos]
        
        audio_mask = [0] * prompt_pre_length + [1] * audio_length + [0]
        audio_mask = [0] * prompt_ids_length + [1] * audio_length + [0]
        audio_mask = torch.tensor(audio_mask, dtype=torch.float32)
        
        ids = self.tokenizer.encode(target) # token ids is different from labels_ids
        ids = self.tokenizer.encode(target)  # token ids is different from labels_ids
        text = torch.tensor(ids, dtype=torch.int64)
        text_lengths = torch.tensor([len(ids)], dtype=torch.int32)
        
@@ -109,19 +250,18 @@
                "audio_mask": audio_mask,
                }
    
    
    def collator(self, samples: list=None):
    def collator(self, samples: list = None):
        outputs = {}
        for sample in samples:
            for key in sample.keys():
                if key not in outputs:
                    outputs[key] = []
                outputs[key].append(sample[key])

        
        for key, data_list in outputs.items():
            if isinstance(data_list[0], torch.Tensor):
                if data_list[0].dtype == torch.int64:
    
                if data_list[0].dtype == torch.int64 or data_list[0].dtype == torch.int32:
                    
                    pad_value = self.int_pad_value
                else:
                    pad_value = self.float_pad_value
@@ -199,26 +339,26 @@
            target = self.preprocessor_text(target)
        
        prompt_ids_pre = self.tokenizer.encode(self.prompt_pre)  # [bos,prompt]
        prompt_pre_length = len(prompt_ids_pre)
        prompt_ids_length = len(prompt_ids_pre)
        
        prompt_input = "{}{}".format(self.prompt_pre, target)
        prompt_input_ids = self.tokenizer.encode(prompt_input)
        audio_length = len(prompt_input_ids) - prompt_pre_length
        audio_length = len(prompt_input_ids) - prompt_ids_length
        input_ids = prompt_input_ids + [self.tokenizer.pad_token_id]
        input_ids = torch.tensor(input_ids, dtype=torch.int64)  # [bos, prompt, input, pad]
        input_ids[prompt_pre_length:] = -1  # [bos, prompt,-1,-1]
        input_ids[prompt_ids_length:] = -1  # [bos, prompt,-1,-1]
        attention_mask = input_ids.ge(-1)  # [true, true, true, true], length mask
        
        prompt_answer = "{}{}".format(self.prompt_pre, target)
        prompt_answer_ids = self.tokenizer.encode(prompt_answer)
        answer_length = len(prompt_answer_ids) - prompt_pre_length
        answer_length = len(prompt_answer_ids) - prompt_ids_length
        labels_ids = copy.deepcopy(prompt_input_ids) + [self.tokenizer.eos_token_id]
        labels_ids = torch.tensor(labels_ids, dtype=torch.int64)  # [bos, prompt, input, eos]
        labels_ids[:prompt_pre_length] = -1  # [-1, -1, input, eos]
        labels_ids[:prompt_ids_length] = -1  # [-1, -1, input, eos]
        label_mask = labels_ids.ge(0)  # [False,False,True,True]
        labels_ids[~label_mask] = self.IGNORE_INDEX  # [-100,-100,input,eos]
        
        audio_mask = [0] * prompt_pre_length + [1] * audio_length + [0]
        audio_mask = [0] * prompt_ids_length + [1] * audio_length + [0]
        audio_mask = torch.tensor(audio_mask, dtype=torch.float32)
        
        ids = self.tokenizer.encode(target)  # token ids is different from labels_ids
@@ -246,7 +386,7 @@
        
        for key, data_list in outputs.items():
            if isinstance(data_list[0], torch.Tensor):
                if data_list[0].dtype == torch.int64:
                if data_list[0].dtype == torch.int64 or data_list[0].dtype == torch.int32:
                    
                    pad_value = self.int_pad_value
                else:

 funasr/datasets/llm_datasets_qwenaudio/datasets.py

@@ -140,7 +140,7 @@
        
        for key, data_list in outputs.items():
            if isinstance(data_list[0], torch.Tensor):
                if data_list[0].dtype == torch.int64:
                if data_list[0].dtype == torch.int64 or data_list[0].dtype == torch.int32:
                    
                    pad_value = self.int_pad_value
                else:

 funasr/datasets/llm_datasets_vicuna/datasets.py

@@ -140,7 +140,7 @@
        
        for key, data_list in outputs.items():
            if isinstance(data_list[0], torch.Tensor):
                if data_list[0].dtype == torch.int64:
                if data_list[0].dtype == torch.int64 or data_list[0].dtype == torch.int32:
                    
                    pad_value = self.int_pad_value
                else:

 funasr/models/llm_asr_nar/model.py

@@ -16,6 +16,7 @@
from funasr.train_utils.device_funcs import force_gatherable
from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank
from funasr.utils import postprocess_utils
from funasr.models.paraformer.cif_predictor import mae_loss
from funasr.utils.datadir_writer import DatadirWriter
from funasr.register import tables

@@ -348,3 +349,335 @@
        
        return results, meta_data


@tables.register("model_classes", "LLMASRNARPrompt")
class LLMASRNARPrompt(nn.Module):
    """ """
    
    def __init__(
        self,
        specaug: str = None,
        specaug_conf: dict = None,
        normalize: str = None,
        normalize_conf: dict = None,
        encoder: str = None,
        encoder_conf: dict = None,
        decoder: str = None,
        decoder_conf: dict = None,
        ctc: str = None,
        ctc_conf: dict = None,
        ctc_weight: float = 0.5,
        llm: str = None,
        llm_conf: dict = None,
        adaptor: str = None,
        adaptor_conf: dict = None,
        input_size: int = 80,
        vocab_size: int = -1,
        ignore_id: int = -1,
        blank_id: int = 0,
        sos: int = 1,
        eos: int = 2,
        lsm_weight: float = 0.0,
        length_normalized_loss: bool = False,
        predictor_weight: int = 1.0,
        report_cer: bool = True,
        report_wer: bool = True,
        sym_space: str = "<space>",
        sym_blank: str = "<blank>",
        # extract_feats_in_collect_stats: bool = True,
        share_embedding: bool = False,
        # preencoder: Optional[AbsPreEncoder] = None,
        # postencoder: Optional[AbsPostEncoder] = None,
        **kwargs,
    ):
        
        super().__init__()
        
        if specaug is not None:
            specaug_class = tables.specaug_classes.get(specaug)
            specaug = specaug_class(**specaug_conf)
        if normalize is not None:
            normalize_class = tables.normalize_classes.get(normalize)
            normalize = normalize_class(**normalize_conf)
        
        # audio encoder
        hub = encoder_conf.get("hub", None)
        if hub == "funasr":
            from funasr import AutoModel
            init_param_path = encoder_conf.get("init_param_path",
                                               "iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch")
            model = AutoModel(model=init_param_path, model_revision="v2.0.4")
            # frontend = model.kwargs.get("frontend")
            model.model.decoder = None
            
            self.audio_encoder = model.model
            # self.frontend = frontend
            self.predictor_weight = predictor_weight
        
        elif hub == "hf":
            pass
        else:
            encoder_class = tables.encoder_classes.get(encoder)
            encoder = encoder_class(input_size=input_size, **encoder_conf)
            encoder_output_size = encoder.output_size()
        
        # llm
        hub = llm_conf.get("hub", "hf")
        self.llm = None
        if hub == "hf":
            from transformers import AutoModelForCausalLM, AutoTokenizer, AutoConfig
            
            init_param_path = llm_conf.get("init_param_path", "vicuna-7b-v1.5")
            model = AutoModelForCausalLM.from_pretrained(
                init_param_path,
                load_in_8bit=None,
                device_map=None,
                use_cache=None,
            )
            freeze = llm_conf.get("freeze", True)
            if freeze:
                for name, param in model.named_parameters():
                    param.requires_grad = False
                model.eval()
            self.llm = model
        
        # adaptor
        adaptor_class = tables.adaptor_classes.get(adaptor)
        adaptor = adaptor_class(**adaptor_conf)
        
        self.adaptor = adaptor
        
        self.blank_id = blank_id
        self.sos = sos if sos is not None else vocab_size - 1
        self.eos = eos if eos is not None else vocab_size - 1
        self.vocab_size = vocab_size
        self.ignore_id = ignore_id
        self.specaug = specaug
        self.normalize = normalize
        self.encoder = encoder
        
        self.criterion_att = LabelSmoothingLoss(
            size=vocab_size,
            padding_idx=ignore_id,
            smoothing=lsm_weight,
            normalize_length=length_normalized_loss,
        )
        self.criterion_pre = mae_loss(normalize_length=length_normalized_loss)
        #
        # if report_cer or report_wer:
        #     self.error_calculator = ErrorCalculator(
        #         token_list, sym_space, sym_blank, report_cer, report_wer
        #     )
        #
        self.error_calculator = None
        
        self.length_normalized_loss = length_normalized_loss
        self.beam_search = None
    
    def forward(
        self,
        speech: torch.Tensor,
        speech_lengths: torch.Tensor,
        text: torch.Tensor,
        text_lengths: torch.Tensor,
        input_ids: torch.Tensor,
        attention_mask: torch.Tensor,
        labels_ids: torch.Tensor,
        label_mask: torch.Tensor,
        audio_mask: torch.Tensor,
        **kwargs,
    ) -> Tuple[torch.Tensor, Dict[str, torch.Tensor], torch.Tensor]:
        """Encoder + Decoder + Calc loss
        Args:
                speech: (Batch, Length, ...)
                speech_lengths: (Batch, )
                text: (Batch, Length)
                text_lengths: (Batch,)
        """
        # import pdb;
        # pdb.set_trace()
        if len(text_lengths.size()) > 1:
            text_lengths = text_lengths[:, 0]
        if len(speech_lengths.size()) > 1:
            speech_lengths = speech_lengths[:, 0]
        
        batch_size = speech.shape[0]
        
        # audio encoder
        encoder_out, encoder_out_lens, loss_pre = self.encode(speech, speech_lengths, audio_mask=audio_mask)
        
        # adaptor
        encoder_out = self.adaptor(encoder_out)
        
        if input_ids is not None:
            input_ids[input_ids == -1] = 0
            input_ids[input_ids == -100] = 0
            if hasattr(self.llm.model, "embed_tokens"):
                inputs_embeds = self.llm.model.embed_tokens(input_ids)
            elif hasattr(self.llm.model.model, "embed_tokens"):
                inputs_embeds = self.llm.model.model.embed_tokens(input_ids)
            else:
                inputs_embeds = self.llm.model.model.model.embed_tokens(input_ids)
            
            if audio_mask is not None:
                # inputs_embeds： [bos, prompt, input, pad, target]
                prompt_bos_length = kwargs.get("prompt_bos_length", None)
                assert prompt_bos_length is not None
                prompt_bos_length = prompt_bos_length[0].item()
                batch_size, token_num, dims = inputs_embeds.shape
                _, l, _ = encoder_out.shape
                encoder_outs_pad = F.pad(encoder_out, (0, 0, prompt_bos_length, token_num - prompt_bos_length - l, 0, 0), value=0.0)
                inputs_embeds = encoder_outs_pad * audio_mask[:, :, None] + inputs_embeds * (1.0 - audio_mask[:, :, None])
                inputs_embeds = F.pad(inputs_embeds[:, 1:, :], (0, 0, 0, 1, 0, 0), value=0.0) # [prompt, input, pad, target, 0.0]
        
        # labels_ids: [bos, prompt, input, target, eos] -> [-1, -1, input, target, eos]
        # loss:
        # inputs_embeds[:-1] -> [prompt, input, pad, target]
        # labels_ids[1:] ->  [prompt, input, target, eos] -> [-1, input, target, eos];
        model_outputs = self.llm(inputs_embeds=inputs_embeds, attention_mask=attention_mask, labels=labels_ids)
        loss_llm = model_outputs.loss
        loss = loss_llm + loss_pre * self.predictor_weight
        stats = {}
        with torch.no_grad():
            preds = torch.argmax(model_outputs.logits, -1)
            acc_att = compute_accuracy(preds[:, :-1], labels_ids[:, 1:], ignore_label=-100)
            stats["acc"] = acc_att
        
        
        stats["loss_pre"] = torch.clone(loss_pre.detach())
        stats["loss_llm"] = torch.clone(loss_llm.detach())
        stats["loss"] = torch.clone(loss.detach())
        
        # force_gatherable: to-device and to-tensor if scalar for DataParallel
        if self.length_normalized_loss:
            batch_size = int((text_lengths + 1).sum())
        loss, stats, weight = force_gatherable((loss, stats, batch_size), loss.device)
        return loss, stats, weight
    
    def encode(
        self, speech: torch.Tensor, speech_lengths: torch.Tensor, **kwargs,
    ):
        
        audio_mask = kwargs.get("audio_mask", None)
        audio_token_lengths = audio_mask.sum(-1) if audio_mask is not None else None
        text_token_int = kwargs.get("text_token_int", None)
        if audio_token_lengths is None:
            audio_token_lengths = torch.tensor([len(text_token_int)], dtype=torch.int64)
        
        batch = {"speech": speech, "speech_lengths": speech_lengths}
        enc, enc_lens = self.audio_encoder.encode(**batch)
        with autocast(False):
            enc_mask = sequence_mask(enc_lens, enc.size(1), device=enc.device)[:, None, :]
            pre_acoustic_embeds, pre_token_length, _, _ = self.audio_encoder.predictor(enc,
                                                                                       mask=enc_mask,
                                                                                       target_label_length=audio_token_lengths,
                                                                                       )
            loss_pre = self.criterion_pre(audio_token_lengths.type_as(pre_token_length), pre_token_length)
        
        return pre_acoustic_embeds, pre_token_length, loss_pre
    
    def inference(self,
                  data_in,
                  data_lengths=None,
                  key: list = None,
                  tokenizer=None,
                  frontend=None,
                  **kwargs,
                  ):
        
        prompt = kwargs.get("prompt", "Transcribe speech to text.")
        
        if kwargs.get("batch_size", 1) > 1:
            raise NotImplementedError("batch decoding is not implemented")
        
        meta_data = {}
        if isinstance(data_in, torch.Tensor) and kwargs.get("data_type", "sound") == "fbank":  # fbank
            speech, speech_lengths = data_in, data_lengths
            if len(speech.shape) < 3:
                speech = speech[None, :, :]
            if speech_lengths is None:
                speech_lengths = speech.shape[1]
        else:
            # extract fbank feats
            time1 = time.perf_counter()
            audio_sample_list = load_audio_text_image_video(data_in, fs=frontend.fs, audio_fs=kwargs.get("fs", 16000),
                                                            data_type=kwargs.get("data_type", "sound"),
                                                            tokenizer=None)
            if len(kwargs.get("data_type")) > 1:
                audio_sample_list, text_token_int_list = audio_sample_list
                text_token_int = text_token_int_list[0].replace(" ", "")
                text_token_int = tokenizer.encode(text_token_int)
            else:
                text_token_int = None
            time2 = time.perf_counter()
            meta_data["load_data"] = f"{time2 - time1:0.3f}"
            speech, speech_lengths = extract_fbank(audio_sample_list, data_type=kwargs.get("data_type", "sound"),
                                                   frontend=frontend)
            time3 = time.perf_counter()
            meta_data["extract_feat"] = f"{time3 - time2:0.3f}"
            meta_data["batch_data_time"] = speech_lengths.sum().item() * frontend.frame_shift * frontend.lfr_n / 1000
        
        speech = speech.to(device=kwargs["device"])
        speech_lengths = speech_lengths.to(device=kwargs["device"])
        
        # Encoder
        encoder_out, encoder_out_lens = self.encode(speech, speech_lengths, text_token_int=text_token_int)
        
        # adaptor
        encoder_out = self.adaptor(encoder_out)
        
        prompt_pre = "USER: \nINSTRUCTION: {}\nINPUT: ".format(prompt)
        prompt_ids = tokenizer.encode(prompt_pre)
        prompt_length = len(prompt_ids)
        prompt_ids = torch.tensor(prompt_ids, dtype=torch.int64).to(kwargs["device"])
        
        if hasattr(self.llm.model, "embed_tokens"):
            inputs_embeds = self.llm.model.embed_tokens(prompt_ids)
        elif hasattr(self.llm.model.model, "embed_tokens"):
            inputs_embeds = self.llm.model.model.embed_tokens(prompt_ids)
        else:
            inputs_embeds = self.llm.model.model.model.embed_tokens(prompt_ids)
        
        inputs_embeds = torch.cat((inputs_embeds[None, :, :], encoder_out), dim=1)  # [prompt, audio]
        attention_mask = torch.ones(inputs_embeds.size()[:-1], dtype=torch.long).to(kwargs["device"])
        
        # model_outputs = self.llm.generate(
        #     inputs_embeds=inputs_embeds,
        #     max_length=kwargs.get("max_length", 200),
        #     max_new_tokens=kwargs.get("max_new_tokens", 200),
        #     num_beams=kwargs.get("num_beams", 4),
        #     do_sample=kwargs.get("do_sample", False),
        #     min_length=kwargs.get("min_length", 1),
        #     top_p=kwargs.get("top_p", 1.0),
        #     repetition_penalty=kwargs.get("repetition_penalty", 1.0),
        #     length_penalty=kwargs.get("length_penalty", 1.0),
        #     temperature=kwargs.get("temperature", 1.0),
        #     attention_mask=attention_mask,
        #     bos_token_id=tokenizer.bos_token_id,
        #     eos_token_id=tokenizer.eos_token_id,
        #     pad_token_id=tokenizer.pad_token_id
        # )
        
        model_outputs = self.llm(inputs_embeds=inputs_embeds, attention_mask=attention_mask, labels=None)
        preds = torch.argmax(model_outputs.logits, -1)
        text = tokenizer.batch_decode(preds, add_special_tokens=False, skip_special_tokens=True)
        
        text = text[0].split(': ')[-1]
        text = text.strip()
        
        # preds = torch.argmax(model_outputs.logits, -1)
        
        ibest_writer = None
        if kwargs.get("output_dir") is not None:
            if not hasattr(self, "writer"):
                self.writer = DatadirWriter(kwargs.get("output_dir"))
            ibest_writer = self.writer[f"{0 + 1}best_recog"]
        
        results = []
        result_i = {"key": key[0], "text": text}
        results.append(result_i)
        
        if ibest_writer is not None:
            ibest_writer["text"][key[0]] = text
        
        return results, meta_data

 funasr/train_utils/trainer.py

@@ -88,6 +88,7 @@
        scaler = GradScaler(enabled=use_fp16) if use_fp16 else None
        scaler = ShardedGradScaler(enabled=use_fp16) if use_ddp else scaler
        self.scaler = scaler
        self.save_checkpoint_interval = kwargs.get("save_checkpoint_interval", 5000)
        
    
        try:
@@ -104,7 +105,7 @@
        self.writer = SummaryWriter(os.path.join(self.output_dir, "tensorboard")) if rank == 0 else None
        
    
    def _save_checkpoint(self, epoch):
    def _save_checkpoint(self, epoch, step=None):
        """
        Saves a checkpoint containing the model's state, the optimizer's state,
        and the scheduler's state at the end of the given epoch. This method is
@@ -123,7 +124,11 @@
            state["scaler_state"] = self.scaler.state_dict()
        # Create output directory if it does not exist
        os.makedirs(self.output_dir, exist_ok=True)
        filename = os.path.join(self.output_dir, f'model.pt.ep{epoch}')
        if step is None:
            filename = os.path.join(self.output_dir, f'model.pt.ep{epoch}')
        else:
            filename = os.path.join(self.output_dir, f'model.pt.ep{epoch}.{step}')
        
        torch.save(state, filename)
        
        print(f'\nCheckpoint saved to {filename}\n')
@@ -337,8 +342,10 @@
                    for key, var in speed_stats.items():
                        self.writer.add_scalar(f'rank{self.local_rank}_{key}/train', eval(var), self.batch_total)


            if (batch_idx+1) % self.save_checkpoint_interval == 0 and self.rank == 0:
                self._save_checkpoint(epoch, step=batch_idx+1)
        pbar.close()
        

    def _validate_epoch(self, epoch):
        """

 funasr/utils/export_utils.py

@@ -4,8 +4,6 @@
def export_onnx(model,
                data_in=None,
                quantize: bool = False,
                fallback_num: int = 5,
                calib_num: int = 100,
                opset_version: int = 14,
                **kwargs):
    model_scripts = model.export(**kwargs)
@@ -19,8 +17,6 @@
        _onnx(m,
              data_in=data_in,
              quantize=quantize,
              fallback_num=fallback_num,
              calib_num=calib_num,
              opset_version=opset_version,
              export_dir=export_dir,
              **kwargs

 runtime/python/onnxruntime/funasr_onnx/paraformer_bin.py

@@ -35,7 +35,8 @@
                 plot_timestamp_to: str = "",
                 quantize: bool = False,
                 intra_op_num_threads: int = 4,
                 cache_dir: str = None
                 cache_dir: str = None,
                 **kwargs
                 ):
        if not Path(model_dir).exists():
            try:
@@ -64,7 +65,7 @@
                      "\npip3 install -U funasr -i https://mirror.sjtu.edu.cn/pypi/web/simple"

            model = AutoModel(model=model_dir)
            model_dir = model.export(type="onnx", quantize=quantize)
            model_dir = model.export(type="onnx", quantize=quantize, **kwargs)
            
        config_file = os.path.join(model_dir, 'config.yaml')
        cmvn_file = os.path.join(model_dir, 'am.mvn')
@@ -243,7 +244,8 @@
                 plot_timestamp_to: str = "",
                 quantize: bool = False,
                 intra_op_num_threads: int = 4,
                 cache_dir: str = None
                 cache_dir: str = None,
                 **kwargs
                 ):

        if not Path(model_dir).exists():
@@ -277,7 +279,7 @@
                      "\npip3 install -U funasr -i https://mirror.sjtu.edu.cn/pypi/web/simple"

            model = AutoModel(model=model_dir)
            model_dir = model.export(type="onnx", quantize=quantize)
            model_dir = model.export(type="onnx", quantize=quantize, **kwargs)
            
        config_file = os.path.join(model_dir, 'config.yaml')
        cmvn_file = os.path.join(model_dir, 'am.mvn')
@@ -403,4 +405,6 @@
    
    
class SeacoParaformer(ContextualParaformer):
    pass # no difference with contextual_paraformer in method of calling onnx models
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        # no difference with contextual_paraformer in method of calling onnx models

 runtime/python/onnxruntime/funasr_onnx/paraformer_online_bin.py

@@ -24,7 +24,8 @@
                 device_id: Union[str, int] = "-1",
                 quantize: bool = False,
                 intra_op_num_threads: int = 4,
                 cache_dir: str = None
                 cache_dir: str = None,
                 **kwargs
                 ):

        if not Path(model_dir).exists():
@@ -56,7 +57,7 @@
                      "\npip3 install -U funasr -i https://mirror.sjtu.edu.cn/pypi/web/simple"

            model = AutoModel(model=model_dir)
            model_dir = model.export(type="onnx", quantize=quantize)
            model_dir = model.export(type="onnx", quantize=quantize, **kwargs)

        config_file = os.path.join(model_dir, 'config.yaml')
        cmvn_file = os.path.join(model_dir, 'am.mvn')

 runtime/python/onnxruntime/funasr_onnx/punc_bin.py

@@ -26,6 +26,7 @@
                 quantize: bool = False,
                 intra_op_num_threads: int = 4,
                 cache_dir: str = None,
                 **kwargs
                 ):
    
        if not Path(model_dir).exists():
@@ -56,7 +57,7 @@
                      "\npip3 install -U funasr -i https://mirror.sjtu.edu.cn/pypi/web/simple"

            model = AutoModel(model=model_dir)
            model_dir = model.export(quantize=quantize)
            model_dir = model.export(type="onnx", quantize=quantize, **kwargs)
            
        config_file = os.path.join(model_dir, 'config.yaml')
        config = read_yaml(config_file)
@@ -168,14 +169,9 @@
    CT-Transformer: Controllable time-delay transformer for real-time punctuation prediction and disfluency detection
    https://arxiv.org/pdf/2003.01309.pdf
    """
    def __init__(self, model_dir: Union[str, Path] = None,
                 batch_size: int = 1,
                 device_id: Union[str, int] = "-1",
                 quantize: bool = False,
                 intra_op_num_threads: int = 4,
                 cache_dir: str = None
    def __init__(self, *args, **kwargs
                 ):
        super().__init__(model_dir, batch_size, device_id, quantize, intra_op_num_threads, cache_dir=cache_dir)
        super().__init__(*args, **kwargs)

    def __call__(self, text: str, param_dict: map, split_size=20):
        cache_key = "cache"

 runtime/python/onnxruntime/funasr_onnx/vad_bin.py

@@ -31,7 +31,8 @@
                 quantize: bool = False,
                 intra_op_num_threads: int = 4,
                 max_end_sil: int = None,
                 cache_dir: str = None
                 cache_dir: str = None,
                 **kwargs
                 ):
        
        if not Path(model_dir).exists():
@@ -62,7 +63,7 @@
                      "\npip3 install -U funasr -i https://mirror.sjtu.edu.cn/pypi/web/simple"
            
            model = AutoModel(model=model_dir)
            model_dir = model.export(type="onnx", quantize=quantize)
            model_dir = model.export(type="onnx", quantize=quantize, **kwargs)
        config_file = os.path.join(model_dir, 'config.yaml')
        cmvn_file = os.path.join(model_dir, 'am.mvn')
        config = read_yaml(config_file)
@@ -196,7 +197,8 @@
                 quantize: bool = False,
                 intra_op_num_threads: int = 4,
                 max_end_sil: int = None,
                 cache_dir: str = None
                 cache_dir: str = None,
                 **kwargs
                 ):
        if not Path(model_dir).exists():
            try:
@@ -226,7 +228,7 @@
                      "\npip3 install -U funasr -i https://mirror.sjtu.edu.cn/pypi/web/simple"
            
            model = AutoModel(model=model_dir)
            model_dir = model.export(type="onnx", quantize=quantize)
            model_dir = model.export(type="onnx", quantize=quantize, **kwargs)
            
        config_file = os.path.join(model_dir, 'config.yaml')
        cmvn_file = os.path.join(model_dir, 'am.mvn')

 runtime/python/onnxruntime/setup.py

@@ -13,7 +13,7 @@


MODULE_NAME = 'funasr_onnx'
VERSION_NUM = '0.3.0'
VERSION_NUM = '0.3.1'

setuptools.setup(
    name=MODULE_NAME,