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

@@ -1001,7 +1001,7 @@ def _join_char_level_timestamps(self, merged_hypothesis, hypotheses, chunk_offse
         # Track time adjustments for filtered tokens within this chunk
         time_adjustment_frames = 0
         for i, h in enumerate(hypotheses):
-            
+
             cumulative_offset += chunk_offsets[i]  # self.chunk_offsets starts with 0,
 
             # update frame numbers
@@ -1011,7 +1011,7 @@ def _join_char_level_timestamps(self, merged_hypothesis, hypotheses, chunk_offse
             for char in h.timestamp['char']:
                 if not char:
                     continue
-                
+
                 # If merged_tokens is provided, filter based on token matching
                 if merged_tokens is not None:
                     # Skip if we've processed all merged tokens or token doesn't match
@@ -1021,12 +1021,12 @@ def _join_char_level_timestamps(self, merged_hypothesis, hypotheses, chunk_offse
                             token_duration_frames = char['end_offset'] - char['start_offset']
                             time_adjustment_frames += token_duration_frames
                         continue
-                
+
                 start = char['start_offset']
                 end = char['end_offset']
 
                 updated_char = dict(char)  # copy all existing keys
-                
+
                 # Apply both the chunk offset and the time adjustment from filtered tokens
                 if start != -1:
                     updated_char['start_offset'] = start + offset - time_adjustment_frames
@@ -1143,7 +1143,9 @@ def _transcribe_output_processing(self, outputs, trcfg: MultiTaskTranscriptionCo
             for i in range(1, len(hypotheses)):
                 merged_tokens = lcs_alignment_merge_buffer(
                     buffer=merged_tokens,
-                    data=hypotheses[i].y_sequence.tolist()[:int(delay * 0.6)],  # only approximately 60% of the tokens are non blank
+                    data=hypotheses[i].y_sequence.tolist()[
+                        : int(delay * 0.6)
+                    ],  # only approximately 60% of the tokens are non blank
                     delay=delay,
                     model=self,
                     max_steps_per_timestep=1,
@@ -1179,7 +1181,7 @@ def _transcribe_output_processing(self, outputs, trcfg: MultiTaskTranscriptionCo
                 },
             )
             chunk_offsets = [0] + [x * self.encoder.subsampling_factor for x in encoded_len.tolist()]
-            
+
             merged_hypotheses = self._join_y_sequence(merged_hypotheses, hypotheses)
             merged_hypotheses.text = final_text
             merged_hypotheses = self._join_timestamp_and_add_word_and_segment_level_timestamps(