add tail padding

egs/librispeech/ASR/conv_emformer_transducer_stateless/decode.py

@@ -59,6 +59,7 @@ Usage:
 
 import argparse
 import logging
+import math
 from collections import defaultdict
 from pathlib import Path
 from typing import Dict, List, Optional, Tuple
@@ -70,7 +71,7 @@ import torch.nn as nn
 from asr_datamodule import LibriSpeechAsrDataModule
 from beam_search import (
     beam_search,
-    fast_beam_search,
+    fast_beam_search_one_best,
     greedy_search,
     greedy_search_batch,
     modified_beam_search,
@@ -91,6 +92,8 @@ from icefall.utils import (
     write_error_stats,
 )
 
+LOG_EPS = math.log(1e-10)
+
 
 def get_parser():
     parser = argparse.ArgumentParser(
@@ -262,13 +265,20 @@ def decode_one_batch(
     supervisions = batch["supervisions"]
     feature_lens = supervisions["num_frames"].to(device)
 
+    feature_lens += params.right_context_length
+    feature = torch.nn.functional.pad(
+        feature,
+        pad=(0, 0, 0, params.right_context_length),
+        value=LOG_EPS,
+    )
+
     encoder_out, encoder_out_lens = model.encoder(
         x=feature, x_lens=feature_lens
     )
     hyps = []
 
     if params.decoding_method == "fast_beam_search":
-        hyp_tokens = fast_beam_search(
+        hyp_tokens = fast_beam_search_one_best(
             model=model,
             decoding_graph=decoding_graph,
             encoder_out=encoder_out,
@@ -294,6 +304,7 @@ def decode_one_batch(
         hyp_tokens = modified_beam_search(
             model=model,
             encoder_out=encoder_out,
+            encoder_out_lens=encoder_out_lens,
             beam=params.beam_size,
         )
         for hyp in sp.decode(hyp_tokens):