funasr2

游雁

2023-12-07 fc246ab820cf57ba08afbe3cbeb4d471036eb83c

 funasr/cli/train_cli.py

@@ -37,7 +37,7 @@
@hydra.main()
def main(kwargs: DictConfig):
   # preprocess_config(kwargs)
   import pdb; pdb.set_trace()
   # import pdb; pdb.set_trace()
   # set random seed
   set_all_random_seed(kwargs.get("seed", 0))
   torch.backends.cudnn.enabled = kwargs.get("cudnn_enabled", torch.backends.cudnn.enabled)
@@ -46,11 +46,11 @@
   
   local_rank = int(os.environ.get('LOCAL_RANK', 0))
   # Check if we are using DDP or FSDP
   use_ddp = 'WORLD_SIZE' in os.environ
   use_ddp = 'WORLD_SIZE' in os.environ and int(os.environ["WORLD_SIZE"]) > 1
   use_fsdp = kwargs.get("use_fsdp", None)
   if use_ddp or use_fsdp:
      dist.init_process_group(backend=kwargs.get("backend", "nccl"), init_method='env://')
      device= torch.cuda.set_device(local_rank)
      torch.cuda.set_device(local_rank)
   
   
   # build_tokenizer
@@ -72,9 +72,24 @@
   # model_class = load_class_from_path(kwargs.get("model").split(":"))
   model_class = dynamic_import(kwargs.get("model"))
   model = model_class(**kwargs, **kwargs["model_conf"], vocab_size=len(tokenizer.token_list))
   # model = model.to(device=kwargs.get("device", "cpu"))
	

   frontend = model.frontend
   # init_param
   init_param = kwargs.get("init_param", None)
   if init_param is not None:
      init_param = eval(init_param)
      if isinstance(init_param, Sequence):
         init_param = (init_param,)
      logging.info("init_param is not None: ", init_param)
      for p in init_param:
         logging.info(f"Loading pretrained params from {p}")
         load_pretrained_model(
            model=model,
            init_param=p,
            ignore_init_mismatch=kwargs.get("ignore_init_mismatch", True),
            oss_bucket=kwargs.get("oss_bucket", None),
         )
   else:
      initialize(model, kwargs.get("init", "kaiming_normal"))
   
   # import pdb;
   # pdb.set_trace()
@@ -94,9 +109,12 @@

   if use_ddp:
      model = model.cuda(local_rank)
      model = DDP(model, device_ids=[local_rank])
      model = DDP(model, device_ids=[local_rank],
                  find_unused_parameters=kwargs.get("train_conf", {}).get("find_unused_parameters", False))
   elif use_fsdp:
      model = FSDP(model).cuda(local_rank)
   else:
      model = model.to(device=kwargs.get("device", "cuda"))
      
      
   # optim
@@ -111,27 +129,9 @@
   scheduler_class = scheduler_choices.get(scheduler)
   scheduler = scheduler_class(optim, **kwargs.get("scheduler_conf"))

   # init_param
   init_param = kwargs.get("init_param", None)
   if init_param is not None:
      init_param = eval(init_param)
      if isinstance(init_param, Sequence):
         init_param = (init_param,)
      logging.info("init_param is not None: ", freeze_param)
      for p in init_param:
         logging.info(f"Loading pretrained params from {p}")
         load_pretrained_model(
            model=model,
            init_param=p,
            ignore_init_mismatch=kwargs.get("ignore_init_mismatch", True),
            oss_bucket=kwargs.get("oss_bucket", None),
         )
   else:
      initialize(model, kwargs.get("init", "kaiming_normal"))
	

   # dataset
   dataset_tr = AudioDataset(kwargs.get("train_data_set_list"), frontend=model.frontend, tokenizer=tokenizer, **kwargs.get("dataset_conf"))
   dataset_tr = AudioDataset(kwargs.get("train_data_set_list"), frontend=frontend, tokenizer=tokenizer, **kwargs.get("dataset_conf"))

   # dataloader
   batch_sampler = BatchSampler(dataset_tr, **kwargs.get("dataset_conf"), **kwargs.get("dataset_conf").get("batch_conf"))
@@ -158,13 +158,6 @@
      torch.distributed.destroy_process_group()

   
	
def train(epoch, model, op):
   pass

def val():
   pass


if __name__ == "__main__":
   main()