Apply isort and black reformatting

Signed-off-by: nune-tadevosyan <nune-tadevosyan@users.noreply.github.com>

Author: nune-tadevosyan

nemo/collections/asr/models/aed_multitask_models.py

@@ -43,6 +43,7 @@
 from nemo.collections.asr.parts.preprocessing.segment import ChannelSelectorType
 from nemo.collections.asr.parts.submodules.multitask_decoding import MultiTaskDecoding, MultiTaskDecodingConfig
 from nemo.collections.asr.parts.submodules.token_classifier import TokenClassifier
+from nemo.collections.asr.parts.utils.chunking_utils import merge_hypotheses_list, merge_parallel_chunks
 from nemo.collections.asr.parts.utils.rnnt_utils import Hypothesis
 from nemo.collections.asr.parts.utils.timestamp_utils import (
     get_forced_aligned_timestamps_with_external_model,
@@ -70,10 +71,7 @@
 )
 from nemo.utils import logging, model_utils
 from nemo.utils.app_state import AppState
-from nemo.collections.asr.parts.utils.chunking_utils import (
-    merge_parallel_chunks,
-    merge_hypotheses_list
-)
+
 __all__ = ['EncDecMultiTaskModel']
 
 
@@ -119,8 +117,8 @@ class MultiTaskTranscriptionConfig(TranscribeConfig):
     """
     Configuration for Multi Task Transcription
 
-    enable_parallel_chunking: bool = False 
-            Whether to enable parallel processing of audio chunks for long-form audio. 
+    enable_parallel_chunking: bool = False
+            Whether to enable parallel processing of audio chunks for long-form audio.
             It will be automatically enabled for batch size 1.
     """
 
@@ -131,7 +129,7 @@ class MultiTaskTranscriptionConfig(TranscribeConfig):
     _internal: Optional[MultiTaskTranscriptionInternalConfig] = field(
         default_factory=lambda: MultiTaskTranscriptionInternalConfig()
     )
-    enable_parallel_chunking: bool = False 
+    enable_parallel_chunking: bool = False
 
     def __post_init__(self):
         self.prompt = parse_multitask_prompt(self.prompt)
@@ -573,7 +571,7 @@ def transcribe(
                 )
             trcfg = override_config
             trcfg.timestamps = timestamps
-        # Check if only one audio is provided with string 
+        # Check if only one audio is provided with string
         is_one_audio = isinstance(audio, str) and not (audio.endswith("json") or audio.endswith("jsonl"))
         # Check if it is provided as a list of strings
         is_one_audio = is_one_audio or (isinstance(audio, list) and len(audio) == 1)
@@ -1004,8 +1002,6 @@ def _transcribe_forward(
             batch=batch,
         )
 
-    
-
     def _transcribe_output_processing(self, outputs, trcfg: MultiTaskTranscriptionConfig) -> GenericTranscriptionType:
         """
         Internal function to process the model's outputs to return the results to the user. This function is called by
@@ -1058,23 +1054,23 @@ def _transcribe_output_processing(self, outputs, trcfg: MultiTaskTranscriptionCo
             hypotheses = process_aed_timestamp_outputs(
                 hypotheses, self.encoder.subsampling_factor, self.cfg['preprocessor']['window_stride']
             )
-            
-        if merge_to_be_done:            
+
+        if merge_to_be_done:
             merged_hypotheses = merge_parallel_chunks(
                 hypotheses=hypotheses,
                 encoded_len=encoded_len,
                 model=self,
                 subsampling_factor=self.encoder.subsampling_factor,
                 window_stride=self.cfg['preprocessor']['window_stride'],
-                tokenizer=self.tokenizer
+                tokenizer=self.tokenizer,
             )
-            #Inject the id of the cut to hypothese to later be used for separate batches
+            # Inject the id of the cut to hypothese to later be used for separate batches
             setattr(merged_hypotheses, 'id', batch.cuts[0].id.split("-", 1)[0])
             return [merged_hypotheses]
-        
-        if  trcfg.enable_parallel_chunking and  len(hypotheses) == 1:
-            setattr(hypotheses[0], 'id', batch.cuts[0].id.split("-", 1)[0])            
-        
+
+        if trcfg.enable_parallel_chunking and len(hypotheses) == 1:
+            setattr(hypotheses[0], 'id', batch.cuts[0].id.split("-", 1)[0])
+
         return hypotheses
 
     def _setup_transcribe_dataloader(self, config: Dict) -> 'torch.utils.data.DataLoader':

nemo/collections/asr/parts/mixins/transcription.py

@@ -382,7 +382,6 @@ def transcribe_generator(self, audio, override_config: Optional[TranscribeConfig
     """
     Transcribe Execution Flow
     """
-    
 
     def _transcribe_on_begin(self, audio, trcfg: TranscribeConfig):
         """

nemo/collections/asr/parts/utils/aligner_utils.py

@@ -113,8 +113,8 @@ def restore_token_case(word: str, word_tokens: List[str]) -> List[str]:
     while "__" in word:
         word = word.replace("__", "_")
 
-#    while " " in word:
-#        word = word.replace(" ", "")
+    #    while " " in word:
+    #        word = word.replace(" ", "")
 
     word_tokens_cased = []
     word_char_pointer = 0
@@ -135,7 +135,11 @@ def restore_token_case(word: str, word_tokens: List[str]) -> List[str]:
                     word_char_pointer += 1
                 else:
                     if token_char == "▁" or token_char == "_":
-                        if word[word_char_pointer] == "▁" or word[word_char_pointer] == "_" or word[word_char_pointer] == " ":
+                        if (
+                            word[word_char_pointer] == "▁"
+                            or word[word_char_pointer] == "_"
+                            or word[word_char_pointer] == " "
+                        ):
                             token_cased += token_char
                             word_char_pointer += 1
                         elif word_char_pointer == 0: