first push

asr.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b51a1267b9d5f8c770ecb379b30e4f136846ae7ff390cac6036fbe37f947d49f
+size 1943172966

hyperparams.yaml

@@ -0,0 +1,135 @@
+# ############################################################################
+# Model: E2E ASR with Transformer
+# Encoder: Conformer Encoder
+# Decoder: Transformer Decoder + (CTC/att joint)
+# Tokens: bpe
+# losses: CTC + KLdiv (Label Smoothing loss)
+# Authors:  Titouan Parcollet
+# ############################################################################
+
+# Feature parameters
+sample_rate: 16000
+n_fft: 400
+n_mels: 80
+
+
+####################### Model Parameters ###########################
+# Transformer
+d_model: 1024
+nhead: 16
+num_encoder_layers: 18
+num_decoder_layers: 6
+d_ffn: 3072
+transformer_dropout: 0.1
+activation: !name:torch.nn.GELU
+output_neurons: 5120
+
+# Outputs
+blank_index: 3
+label_smoothing: 0.1
+pad_index: 0
+bos_index: 1
+eos_index: 2
+
+# Decoding parameters
+min_decode_ratio: 0.0
+max_decode_ratio: 1.0
+valid_search_interval: 10
+valid_beam_size: 1 # We do greedy here so it's faster to decode ...
+test_beam_size: 80
+ctc_weight_decode: 0.3
+scorer_beam_scale: 0.3
+transformer_beam_search: True
+
+############################## models ################################
+
+tokenizer: !new:sentencepiece.SentencePieceProcessor
+
+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>
+
+CNN: !new:speechbrain.lobes.models.convolution.ConvolutionFrontEnd
+    input_shape: (8, 10, 80)
+    num_blocks: 2
+    num_layers_per_block: 1
+    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: 640
+    tgt_vocab: !ref <output_neurons>
+    d_model: !ref <d_model>
+    nhead: !ref <nhead>
+    num_encoder_layers: !ref <num_encoder_layers>
+    num_decoder_layers: !ref <num_decoder_layers>
+    d_ffn: !ref <d_ffn>
+    dropout: !ref <transformer_dropout>
+    conformer_activation: !ref <activation>
+    activation: !ref <activation>
+    encoder_module: conformer
+    attention_type: RelPosMHAXL
+    normalize_before: True
+    causal: False
+
+ctc_lin: !new:speechbrain.nnet.linear.Linear
+    input_size: !ref <d_model>
+    n_neurons: !ref <output_neurons>
+
+seq_lin: !new:speechbrain.nnet.linear.Linear
+    input_size: !ref <d_model>
+    n_neurons: !ref <output_neurons>
+
+asr_model: !new:torch.nn.ModuleList
+    - [!ref <CNN>, !ref <Transformer>, !ref <seq_lin>, !ref <ctc_lin>]
+
+# Scorer
+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 <ctc_scorer>]
+    weights:
+        ctc: !ref <ctc_weight_decode>
+    scorer_beam_scale: !ref <scorer_beam_scale>
+
+# We compose the inference (encoder) pipeline.
+encoder: !new:speechbrain.nnet.containers.LengthsCapableSequential
+    input_shape: [null, null, !ref <n_mels>]
+    compute_features: !ref <compute_features>
+    normalize: !ref <normalizer>
+    CNN: !ref <CNN>
+
+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: 3
+    temperature: 1.15
+    using_eos_threshold: True
+    scorer: !ref <scorer>
+
+modules:
+    normalizer: !ref <normalizer>
+    encoder: !ref <encoder>
+    transformer: !ref <Transformer>
+    decoder: !ref <decoder>
+
+log_softmax: !new:torch.nn.LogSoftmax
+    dim: -1
+
+pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
+    loadables:
+        normalizer: !ref <normalizer>
+        asr: !ref <asr_model>
+        tokenizer: !ref <tokenizer>

normalizer.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eda13a16647a5d3cad4681ffb158df5795cbd479636f59b98d7b380819db2983
+size 2218

tokenizer.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a16e33c18124234c5c4998f54344c987dbc98736a75a78056d63e58c88da04c
+size 319849