push stuff

config.json

@@ -1,4 +1,5 @@
 {
+  "_name_or_path": "./",
   "activation_dropout": 0.0,
   "apply_spec_augment": true,
   "architectures": [
@@ -58,8 +59,10 @@
   "mask_channel_length": 10,
   "mask_channel_prob": 0.0,
   "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_min_space": 1,
   "mask_time_prob": 0.075,
   "model_type": "unispeech",
@@ -76,8 +79,8 @@
   "proj_codevector_dim": 768,
   "replace_prob": 0.5,
   "tokenizer_class": "Wav2Vec2CTCTokenizer",
-  "torch_dtype": "float16",
-  "transformers_version": "4.13.0.dev0",
+  "torch_dtype": "float32",
+  "transformers_version": "4.15.0.dev0",
   "use_weighted_layer_sum": false,
   "vocab_size": 65
 }

phonesMatches_reduced.json

@@ -0,0 +1 @@
+{"e": 0, "t\u0283": 1, "r": 2, "k": 3, "\u0254": 4, "d": 5, "\u026a": 6, "a": 7, "j": 8, "\u028a": 9, "t": 10, "i": 11, "l": 12, "p": 13, "\u0261": 14, "t\u02d0": 15, "n": 16, "o": 17, "\u025b": 18, "\u027e": 19, "m": 20, "u": 21, "s": 22, "i\u02d0": 23, "ss": 24, "b": 25, "v": 26, "d\u0292": 27, "f": 28, "w": 29, "z": 30, "\u0283": 31, "\u0272": 32, "\u028e": 33, "e\u026a": 34, "t\u032a": 35, "ts": 36, "p\u02d0": 37, "d\u0292\u02d0": 38, "dz\u02d0": 39, "\u0254\u02d0": 40, "u\u02d0": 41, "k\u02d0": 42, "b\u02d0": 43, "a\u026a": 44, "\u014b": 45, "ts\u02d0": 46, "d\u02d0": 47, "d\u032a": 48, "t\u0283\u02d0": 49, "a\u02d0": 50, "h": 51, "e\u02d0": 52, "\u025b\u02d0": 53, "dz": 54, "o\u02d0": 55, "\u0261\u02d0": 56, "\u0292": 57, "\u026a\u02d0": 58, "o\u026a": 59, "\u028a\u02d0": 60}

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4543945a02416312ce57518f2566291dd3b99afc3a15b303c93907f98ca7ff17
-size 631197591
+oid sha256:cdf09649b51dca05a020d26ec7065fba1b045423739ee13ac58dda26e42b3c2e
+size 1262168511

tokenizer_config.json

@@ -1 +1 @@
-{"unk_token": "<unk>", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "<pad>", "phonemizer_lang": "en-us", "phonemizer_backend": "espeak", "word_delimiter_token": "|", "do_lower_case": false, "tokenizer_class": "Wav2Vec2PhonemeCTCTokenizer"}
+{"unk_token": "<unk>", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "<pad>", "word_delimiter_token": null, "phone_delimiter_token": " ", "do_phonemize": true, "phonemizer_lang": "en-us", "phonemizer_backend": "espeak", "tokenizer_class": "Wav2Vec2PhonemeCTCTokenizer"}

vocab.json

@@ -1 +1 @@
-{"e": 0, "tʃ": 1, "r": 2, "k": 3, "ɔ": 4, "d": 5, "ɪ": 6, "a": 7, "j": 8, "ʊ": 9, "t": 10, "i": 11, "l": 12, "p": 13, "ɡ": 14, "tː": 15, "n": 16, "o": 17, "ɛ": 18, "ɾ": 19, "m": 20, "u": 21, "s": 22, "iː": 23, "ss": 24, "b": 25, "v": 26, "dʒ": 27, "f": 28, "w": 29, "z": 30, "ʃ": 31, "ɲ": 32, "ʎ": 33, "eɪ": 34, "t̪": 35, "ts": 36, "pː": 37, "dʒː": 38, "dzː": 39, "ɔː": 40, "uː": 41, "kː": 42, "bː": 43, "aɪ": 44, "ŋ": 45, "tsː": 46, "dː": 47, "d̪": 48, "tʃː": 49, "aː": 50, "h": 51, "eː": 52, "ɛː": 53, "dz": 54, "oː": 55, "ɡː": 56, "ʒ": 57, "ɪː": 58, "oɪ": 59, "ʊː": 60, "<s>": 43, "<pad>": 42, "</s>": 63, "<unk>": 64}
+{"e": 0, "tʃ": 1, "r": 2, "k": 3, "ɔ": 4, "d": 5, "ɪ": 6, "a": 7, "j": 8, "ʊ": 9, "t": 10, "i": 11, "l": 12, "p": 13, "ɡ": 14, "tː": 15, "n": 16, "o": 17, "ɛ": 18, "ɾ": 19, "m": 20, "u": 21, "s": 22, "iː": 23, "ss": 24, "b": 25, "v": 26, "dʒ": 27, "f": 28, "w": 29, "z": 30, "ʃ": 31, "ɲ": 32, "ʎ": 33, "eɪ": 34, "t̪": 35, "ts": 36, "pː": 37, "dʒː": 38, "dzː": 39, "ɔː": 40, "uː": 41, "kː": 42, "bː": 43, "aɪ": 44, "ŋ": 45, "tsː": 46, "dː": 47, "d̪": 48, "tʃː": 49, "aː": 50, "h": 51, "eː": 52, "ɛː": 53, "dz": 54, "oː": 55, "ɡː": 56, "ʒ": 57, "ɪː": 58, "oɪ": 59, "ʊː": 60, "<pad>": 61, "<s>": 62, "</s>": 63, "<unk>": 64}