50 epochs

config.json

@@ -1,6 +1,9 @@
 {
   "_name_or_path": "facebook/wav2vec2-xls-r-300m",
   "activation_dropout": 0.055,
+  "adapter_kernel_size": 3,
+  "adapter_stride": 2,
+  "add_adapter": false,
   "apply_spec_augment": true,
   "architectures": [
     "Wav2Vec2ForCTC"
@@ -58,10 +61,13 @@
   "layer_norm_eps": 1e-05,
   "layerdrop": 0.041,
   "mask_feature_length": 10,
+  "mask_feature_min_masks": 0,
   "mask_feature_prob": 0.0,
   "mask_time_length": 10,
+  "mask_time_min_masks": 2,
   "mask_time_prob": 0.4,
   "model_type": "wav2vec2",
+  "num_adapter_layers": 3,
   "num_attention_heads": 16,
   "num_codevector_groups": 2,
   "num_codevectors_per_group": 320,
@@ -70,10 +76,33 @@
   "num_feat_extract_layers": 7,
   "num_hidden_layers": 24,
   "num_negatives": 100,
+  "output_hidden_size": 1024,
   "pad_token_id": 38,
   "proj_codevector_dim": 768,
+  "tdnn_dilation": [
+    1,
+    2,
+    3,
+    1,
+    1
+  ],
+  "tdnn_dim": [
+    512,
+    512,
+    512,
+    512,
+    1500
+  ],
+  "tdnn_kernel": [
+    5,
+    3,
+    3,
+    1,
+    1
+  ],
   "torch_dtype": "float32",
-  "transformers_version": "4.11.3",
+  "transformers_version": "4.16.1",
   "use_weighted_layer_sum": false,
-  "vocab_size": 41
+  "vocab_size": 41,
+  "xvector_output_dim": 512
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:80fb9a33792275409933c0e0dd58a80a5738a44784b8ad5b2d5d7bdddadd0992
+oid sha256:f0f669e4d738fbb52242e30e2a804595460778b5afd8784b8d8a4876025601e6
 size 1262091761

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2dde1472c46d20edf889a420666ead256e4dfd14ae79e82fb9e458b5122f4ff8
-size 2863
+oid sha256:91fff6e217226fab63b0b117587d18515a1d4e44ab9622a6fb4fa01a4fa1905b
+size 3055