fix bugs

egs/librispeech/ASR/codebook_index_extraction.sh

@@ -98,7 +98,7 @@ if [ $stage -eq 4 ]; then
     # Note: order of split manfiests is 1-based, while gpu is 0-based.
     export CUDA_VISIBLE_DEVICES=`(expr $1 + 5)`
     ./vq_pruned_transducer_stateless2/hubert_code_indices.py \
-      --memory-layer=${memory_layer}
+      --memory-layer=${memory_layer} \
       --num-splits $num_jobs \
       --subset=$2 \
       --manifest-idx $1 \
@@ -117,7 +117,7 @@ if [ $stage -eq 4 ]; then
   wait
 fi
 
-cdidx_manifests_dir=`pwd`/data/globalrandom-scaledquantizer-refine_iter-5-${num_utts}-$model_id-${mem_layer}layer-${quantizer_id}-bytes_per_frame-${bytes_per_frame}-enable-refine-True
+cdidx_manifests_dir=`pwd`/data/globalrandom-scaledquantizer-refine_iter-5-${num_utts}-$model_id-${memory_layer}layer-${quantizer_id}-bytes_per_frame-${bytes_per_frame}-enable-refine-True
 if [ $stage -eq 5 ]; then
   for subset in ${train_subsets}; do
     combined_list=`find $cdidx_manifests_dir/splits$num_jobs/ -name cuts_train-${sbuset}*`

egs/librispeech/ASR/vq_pruned_transducer_stateless2/decode.py

@@ -444,6 +444,7 @@ def main():
 
     params = get_params()
     params.update(vars(args))
+    params.extra_output_layer=None
 
     assert params.decoding_method in (
         "greedy_search",

egs/librispeech/ASR/vq_pruned_transducer_stateless2/hubert_decode.py

@@ -159,11 +159,11 @@ def main():
     parser = get_parser()
     LibriSpeechAsrDataModule.add_arguments(parser)
     args = parser.parse_args()
-    args.exp_dir = Path(args.exp_dir)
 
     params = AttributeDict()
     params.update(vars(args))
     params.update(vq_config)
+    params.exp_dir = Path(params.exp_dir)
 
     setup_logger(f"{params.exp_dir}/log-ctc_greedy_search/log-decode")
     logging.info("Decoding started")

egs/librispeech/ASR/vq_pruned_transducer_stateless2/hubert_memory_embeddings.py

@@ -29,15 +29,12 @@ from icefall.utils import (
     setup_logger,
 )
 
-from hubert_utils import extract_layers_result, load_hubert_model, vq_config
-
-
-def get_parser():
-    parser = argparse.ArgumentParser(
-        formatter_class=argparse.ArgumentDefaultsHelpFormatter
-    )
-    return parser
-
+from hubert_utils import (
+    extract_layers_result,
+    load_hubert_model,
+    get_parser,
+    vq_config,
+)
 
 def compute_memory(
     model: torch.nn.Module,

egs/librispeech/ASR/vq_pruned_transducer_stateless2/hubert_utils.py

@@ -33,6 +33,7 @@ from omegaconf import OmegaConf
 vq_config = {
     # TODO: Maybe better to convert this class to yaml driven config.
     # parameters about hubert model inference.
+    "exp_dir": "./vq_pruned_transducer_stateless2/exp/",
     "model_dir": "./vq_pruned_transducer_stateless2/exp/hubert_models/",
     "input_strategy": "AudioSamples",
     "enable_spec_aug": False,