init

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da85c29829d4002daedd676e012936488234d9255e65e86dfab9bec6b1729298
+size 5905440

v3_e2e_rnnt.yaml

@@ -0,0 +1,46 @@
+model_class: rnnt
+sample_rate: 16000
+preprocessor:
+  _target_: gigaam.preprocess.FeatureExtractor
+  sample_rate: 16000
+  features: 64
+  win_length: 320
+  hop_length: 160
+  mel_scale: htk
+  n_fft: 320
+  mel_norm: null
+  center: false
+encoder:
+  _target_: gigaam.encoder.ConformerEncoder
+  feat_in: 64
+  n_layers: 16
+  d_model: 768
+  subsampling_factor: 4
+  ff_expansion_factor: 4
+  self_attention_model: rotary
+  pos_emb_max_len: 5000
+  n_heads: 16
+  conv_kernel_size: 5
+  flash_attn: false
+  subs_kernel_size: 5
+  subsampling: conv1d
+  conv_norm_type: layer_norm
+head:
+  _target_: gigaam.decoder.RNNTHead
+  decoder:
+    pred_hidden: 320
+    pred_rnn_layers: 1
+    num_classes: 1025
+  joint:
+    enc_hidden: 768
+    pred_hidden: 320
+    joint_hidden: 320
+    num_classes: 1025
+decoding:
+  _target_: gigaam.decoding.RNNTGreedyDecoding
+  vocabulary: null
+  model_path: /models/v3_e2e_rnnt_tokenizer.model
+model_name: v3_e2e_rnnt
+hashes:
+  model: 72e2a9b5c7caad963b2bbfd2f298c252
+  tokenizer: 3b3bf8370e882885d79731592fc99f98

v3_e2e_rnnt_decoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b0a16d67fd2cb37061decc93c69e364a9ab27afee3c57495d55b1c974cf7231
+size 4599910

v3_e2e_rnnt_encoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:297c67d9c56bd4e9f34934a5f35ad5d2b49557247e5a69de6bcc4eebfea29988
+size 885084534

v3_e2e_rnnt_joint.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:602ff7017a93311aad34df1437c8d7f49911353c13d6eae7a6ee7b041339465c
+size 2712896

v3_e2e_rnnt_tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:828c12c991019eef952a960661f25a92d6ad279591e2ea466b4aeddf1d20a18a
+size 255336