add d file

egs/librispeech/ASR/conv_emformer_transducer_stateless2/decode.py

@@ -96,6 +96,7 @@ from icefall.checkpoint import (
     find_checkpoints,
     load_checkpoint,
 )
+from icefall.lexicon import Lexicon
 from icefall.utils import (
     AttributeDict,
     DecodingResults,
@@ -167,6 +168,13 @@ def get_parser():
         help="Path to the BPE model",
     )
 
+    parser.add_argument(
+        "--lang-dir",
+        type=Path,
+        default="data/lang_bpe_500",
+        help="The lang dir containing word table and LG graph",
+    )
+
     parser.add_argument(
         "--decoding-method",
         type=str,
@@ -286,6 +294,8 @@ def decode_one_batch(
     )
 
     encoder_out, encoder_out_lens = model.encoder(x=feature, x_lens=feature_lens)
+    if isinstance(encoder_out, list):
+        encoder_out = encoder_out[-1] # the last item is final output
     hyps = []
 
     if params.decoding_method == "fast_beam_search":
@@ -345,12 +355,10 @@ def decode_one_batch(
         res = DecodingResults(hyps=tokens, timestamps=timestamps)
 
     hyps, timestamps = parse_hyp_and_timestamp(
-        decoding_method=params.decoding_method,
         res=res,
         sp=sp,
         subsampling_factor=params.subsampling_factor,
         frame_shift_ms=params.frame_shift_ms,
-        word_table=word_table,
     )
 
     if params.decoding_method == "greedy_search":
@@ -533,6 +541,7 @@ def main():
     args = parser.parse_args()
     args.exp_dir = Path(args.exp_dir)
 
+    import pdb; pdb.set_trace()
     params = get_params()
     params.update(vars(args))
 
@@ -669,6 +678,9 @@ def main():
     else:
         decoding_graph = None
 
+    lexicon = Lexicon(params.lang_dir)
+    word_table = lexicon.word_table
+
     num_param = sum([p.numel() for p in model.parameters()])
     logging.info(f"Number of model parameters: {num_param}")