typo fix

Author: zigzagcai

generate.py

@@ -15,7 +15,7 @@
 from tqdm import tqdm
 
 from internlm.accelerator import get_accelerator
-from internlm.apis.inference import SequenceGenerator
+from internlm.inference.inference import SequenceGenerator
 from internlm.core.context import global_context as gpc
 from internlm.data import build_generation_loader_with_data_type
 from internlm.initialize import initialize_launcher

internlm/core/scheduler/base_scheduler.py

@@ -8,8 +8,8 @@
 
 import torch
 
-from internlm.apis import InferenceParams
 from internlm.core.engine import Engine
+from internlm.inference import InferenceParams
 
 
 class BaseScheduler(ABC):

internlm/data/streaming/dataset.py

@@ -8,10 +8,10 @@
 from datasets.distributed import split_dataset_by_node
 from PIL import Image
 from torch.utils.data import Dataset
+from transformers import AutoTokenizer
 
 from internlm.core.context import ParallelMode
 from internlm.core.context import global_context as gpc
-from transformers import AutoTokenizer
 
 
 class StreamingDataset(Dataset):

internlm/apis/__init__.py internlm/inference/__init__.py

@@ -5,10 +5,10 @@
 import torch.nn.functional as F
 from torch import nn
 
-from internlm.apis import InferenceParams, process_parallel_output
 from internlm.core.context import ParallelMode  # noqa: E402
 from internlm.core.context import global_context as gpc  # noqa: E402
 from internlm.core.trainer import Trainer
+from internlm.inference import InferenceParams, process_parallel_output
 
 __all__ = ["SequenceGenerator"]
 

internlm/apis/inference.py internlm/inference/inference.py

@@ -93,7 +93,7 @@ def set_fp32_attr_for_model(model: Union[nn.Module, nn.ModuleList]):
 
 
 def set_parallel_attr_for_param_groups(model: Union[nn.Module, nn.ModuleList]):
-    def _check_module(name, module):
+    def _check_module(name, module):  # pylint: disable=W0613
         # layer_norm
         if isinstance(module, (RMSNorm, nn.LayerNorm)):
             for param in module.parameters():

internlm/apis/inference_utils.py internlm/inference/inference_utils.py

@@ -4,7 +4,7 @@
 import torch
 from sentencepiece import SentencePieceProcessor
 
-from internlm.apis.inference import SequenceGenerator, batch_tokenize
+from internlm.inference.inference import SequenceGenerator, batch_tokenize
 from internlm.initialize import initialize_launcher  # noqa: E402
 from internlm.initialize.initialize_model import (
     initialize_model_and_parallel_communicator,

internlm/initialize/initialize_model.py

@@ -3,7 +3,7 @@
 import pytest
 from sentencepiece import SentencePieceProcessor
 
-from internlm.apis.inference import SequenceGenerator, batch_tokenize
+from internlm.inference.inference import SequenceGenerator, batch_tokenize
 from internlm.checkpoint import CheckpointManager  # noqa: E402
 from internlm.core.context import global_context as gpc  # noqa: E402
 from internlm.core.trainer import Trainer, TrainState  # noqa: E402

tests/test_infer/test_generate.py

@@ -7,7 +7,7 @@
 
 import torch
 
-from internlm.apis.inference import SequenceGenerator
+from internlm.inference.inference import SequenceGenerator
 from internlm.core.context import ParallelMode
 from internlm.core.context import global_context as gpc
 from internlm.initialize import initialize_launcher