include apostrophes, augment audio, train longer

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
+  "_name_or_path": "/workspace/checkpoints/lg/wav2vec2-large-xlsr-lg-augment/checkpoint-7424",
   "activation_dropout": 0.05,
   "apply_spec_augment": true,
   "architectures": [
@@ -62,7 +62,7 @@
   "mask_time_length": 10,
   "mask_time_min_space": 1,
   "mask_time_other": 0.0,
-  "mask_time_prob": 0.05,
+  "mask_time_prob": 0.1,
   "mask_time_selection": "static",
   "model_type": "wav2vec2",
   "num_attention_heads": 16,
@@ -70,7 +70,7 @@
   "num_conv_pos_embeddings": 128,
   "num_feat_extract_layers": 7,
   "num_hidden_layers": 24,
-  "pad_token_id": 26,
+  "pad_token_id": 27,
   "transformers_version": "4.5.0.dev0",
-  "vocab_size": 27
+  "vocab_size": 28
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:379e362ee3f247a8e610b575c2919fb552226da693b4f8de8aee454f6dd75650
-size 1262044503
+oid sha256:e6a34a3a923d86de6307ba153011d3ec71a4ba609c662bd4997fb69ede168ee1
+size 1262048599

vocab.json

@@ -1 +1 @@
-{"y": 0, "d": 1, "s": 2, "n": 3, "c": 4, "l": 5, "h": 6, "w": 7, "u": 8, "z": 9, "t": 11, "g": 12, "j": 13, "m": 14, "o": 15, "e": 16, "k": 17, "r": 18, "i": 19, "b": 20, "f": 21, "a": 22, "v": 23, "p": 24, "|": 10, "[UNK]": 25, "[PAD]": 26}
+{"y": 0, "k": 1, "o": 2, "e": 3, "u": 4, "p": 5, "c": 6, "t": 7, "b": 8, "m": 9, "d": 10, "f": 11, "h": 12, "s": 13, "v": 14, "r": 15, "w": 17, "l": 18, "j": 19, "g": 20, "i": 21, "z": 22, "'": 23, "n": 24, "a": 25, "|": 16, "[UNK]": 26, "[PAD]": 27}