Update hyperparams.yaml

hyperparams.yaml

@@ -6,57 +6,54 @@
 # losses: CTC + KLdiv (Label Smoothing loss)
 # Training: KsponSpeech 965.2h
 # Based on the works of: Jianyuan Zhong, Titouan Parcollet 2021
-# Authors: Dongwon Kim, Dongwoo Kim 2021
+# Authors: Dong Won Kim, Dongwoo Kim 2021, 2024
 # ############################################################################
 # Seed needs to be set at top of yaml, before objects with parameters are made
+# ############################################################################
 
 # Feature parameters
 sample_rate: 16000
-n_fft: 400
+n_fft: 512
 n_mels: 80
 
 ####################### Model parameters ###########################
 # Transformer
-d_model: 256
-nhead: 4
+d_model: 512
+nhead: 8
 num_encoder_layers: 12
 num_decoder_layers: 6
 d_ffn: 2048
-transformer_dropout: 0.0
+transformer_dropout: 0.1
 activation: !name:torch.nn.GELU
 output_neurons: 5000
-vocab_size: 5000
 
 # Outputs
 blank_index: 0
+label_smoothing: 0.1
 pad_index: 0
 bos_index: 1
 eos_index: 2
-unk_index: 0
 
 # Decoding parameters
 min_decode_ratio: 0.0
 max_decode_ratio: 1.0
-test_beam_size: 10
-lm_weight: 0.0
+test_beam_size: 66
+lm_weight: 0.60
 ctc_weight_decode: 0.40
 
 ############################## models ################################
 
-normalizer: !new:speechbrain.processing.features.InputNormalization
-    norm_type: global
-
 CNN: !new:speechbrain.lobes.models.convolution.ConvolutionFrontEnd
     input_shape: (8, 10, 80)
-    num_blocks: 3
+    num_blocks: 2
     num_layers_per_block: 1
-    out_channels: (64, 64, 64)
-    kernel_sizes: (5, 5, 1)
-    strides: (2, 2, 1)
-    residuals: (False, False, True)
-    
+    out_channels: (64, 32)
+    kernel_sizes: (3, 3)
+    strides: (2, 2)
+    residuals: (False, False)
+
 Transformer: !new:speechbrain.lobes.models.transformer.TransformerASR.TransformerASR # yamllint disable-line rule:line-length
-    input_size: 1280
+    input_size: 640
     tgt_vocab: !ref <output_neurons>
     d_model: !ref <d_model>
     nhead: !ref <nhead>
@@ -70,20 +67,6 @@ Transformer: !new:speechbrain.lobes.models.transformer.TransformerASR.Transforme
     normalize_before: True
     causal: False
 
-# NB: It has to match the pre-trained TransformerLM!!
-lm_model: !new:speechbrain.lobes.models.transformer.TransformerLM.TransformerLM # yamllint disable-line rule:line-length
-    vocab: !ref <output_neurons>
-    d_model: 768
-    nhead: 12
-    num_encoder_layers: 12
-    num_decoder_layers: 0
-    d_ffn: 3072
-    dropout: 0.0
-    activation: !name:torch.nn.GELU
-    normalize_before: False
-
-tokenizer: !new:sentencepiece.SentencePieceProcessor
-
 ctc_lin: !new:speechbrain.nnet.linear.Linear
     input_size: !ref <d_model>
     n_neurons: !ref <output_neurons>
@@ -92,25 +75,62 @@ seq_lin: !new:speechbrain.nnet.linear.Linear
     input_size: !ref <d_model>
     n_neurons: !ref <output_neurons>
 
-decoder: !new:speechbrain.decoders.S2STransformerBeamSearch
-    modules: [!ref <Transformer>, !ref <seq_lin>, !ref <ctc_lin>]
-    bos_index: !ref <bos_index>
+transformerlm_scorer: !new:speechbrain.decoders.scorer.TransformerLMScorer
+   language_model: !ref <lm_model>
+   temperature: 1.15
+   
+ctc_scorer: !new:speechbrain.decoders.scorer.CTCScorer
     eos_index: !ref <eos_index>
     blank_index: !ref <blank_index>
+    ctc_fc: !ref <ctc_lin>
+    
+scorer: !new:speechbrain.decoders.scorer.ScorerBuilder
+    full_scorers: [!ref <transformerlm_scorer>, !ref <ctc_scorer>]
+    weights:
+        transformerlm: !ref <lm_weight>
+        ctc: !ref <ctc_weight_decode>
+
+        
+decoder: !new:speechbrain.decoders.S2STransformerBeamSearcher
+    modules: [!ref <Transformer>, !ref <seq_lin>]
+    bos_index: !ref <bos_index>
+    eos_index: !ref <eos_index>
     min_decode_ratio: !ref <min_decode_ratio>
     max_decode_ratio: !ref <max_decode_ratio>
     beam_size: !ref <test_beam_size>
-    ctc_weight: !ref <ctc_weight_decode>
-    lm_weight: !ref <lm_weight>
-    lm_modules: !ref <lm_model>
-    temperature: 1.30
-    temperature_lm: 1.30
+    temperature: 1.15
     using_eos_threshold: False
     length_normalization: True
+    scorer: !ref <scorer>
 
 log_softmax: !new:torch.nn.LogSoftmax
     dim: -1
 
+normalizer: !new:speechbrain.processing.features.InputNormalization
+    norm_type: global
+
+compute_features: !new:speechbrain.lobes.features.Fbank
+    sample_rate: !ref <sample_rate>
+    n_fft: !ref <n_fft>
+    n_mels: !ref <n_mels>
+
+# This is the Transformer LM that is used according to the Huggingface repository
+# Visit the HuggingFace model corresponding to the pretrained_lm_tokenizer_path
+# For more details about the model!
+# NB: It has to match the pre-trained TransformerLM!!
+lm_model: !new:speechbrain.lobes.models.transformer.TransformerLM.TransformerLM
+    vocab: 5000
+    d_model: 768
+    nhead: 12
+    num_encoder_layers: 12
+    num_decoder_layers: 0
+    d_ffn: 3072
+    dropout: 0.0
+    activation: !name:torch.nn.GELU
+    normalize_before: False
+
+tokenizer: !new:sentencepiece.SentencePieceProcessor
+
 Tencoder: !new:speechbrain.lobes.models.transformer.TransformerASR.EncoderWrapper
     transformer: !ref <Transformer>
 
@@ -121,14 +141,10 @@ encoder: !new:speechbrain.nnet.containers.LengthsCapableSequential
     cnn: !ref <CNN>
     transformer_encoder: !ref <Tencoder>
 
+# Models
 asr_model: !new:torch.nn.ModuleList
     - [!ref <CNN>, !ref <Transformer>, !ref <seq_lin>, !ref <ctc_lin>]
 
-compute_features: !new:speechbrain.lobes.features.Fbank
-    sample_rate: !ref <sample_rate>
-    n_fft: !ref <n_fft>
-    n_mels: !ref <n_mels>
-
 modules:
    compute_features: !ref <compute_features>
    normalizer: !ref <normalizer>
@@ -146,4 +162,4 @@ pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
       normalizer: !ref <normalizer>
       asr: !ref <asr_model>
       lm: !ref <lm_model>
-      tokenizer: !ref <tokenizer>
+      tokenizer: !ref <tokenizer>