WER 51%

config.json

@@ -1,12 +1,14 @@
 {
-  "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
-  "activation_dropout": 0.0,
+  "_name_or_path": "./xlsr",
+  "activation_dropout": 0.1,
   "apply_spec_augment": true,
   "architectures": [
     "Wav2Vec2ForCTC"
   ],
   "attention_dropout": 0.1,
   "bos_token_id": 1,
+  "codevector_dim": 768,
+  "contrastive_logits_temperature": 0.1,
   "conv_bias": true,
   "conv_dim": [
     512,
@@ -36,13 +38,15 @@
     2
   ],
   "ctc_loss_reduction": "mean",
-  "ctc_zero_infinity": false,
+  "ctc_zero_infinity": true,
+  "diversity_loss_weight": 0.1,
   "do_stable_layer_norm": true,
   "eos_token_id": 2,
   "feat_extract_activation": "gelu",
   "feat_extract_dropout": 0.0,
   "feat_extract_norm": "layer",
   "feat_proj_dropout": 0.0,
+  "feat_quantizer_dropout": 0.0,
   "final_dropout": 0.0,
   "gradient_checkpointing": true,
   "hidden_act": "gelu",
@@ -66,11 +70,15 @@
   "mask_time_selection": "static",
   "model_type": "wav2vec2",
   "num_attention_heads": 16,
+  "num_codevector_groups": 2,
+  "num_codevectors_per_group": 320,
   "num_conv_pos_embedding_groups": 16,
   "num_conv_pos_embeddings": 128,
   "num_feat_extract_layers": 7,
   "num_hidden_layers": 24,
-  "pad_token_id": 38,
-  "transformers_version": "4.4.0.dev0",
-  "vocab_size": 39
+  "num_negatives": 100,
+  "pad_token_id": 40,
+  "proj_codevector_dim": 768,
+  "transformers_version": "4.7.0",
+  "vocab_size": 41
 }

preprocessor_config.json

@@ -1,5 +1,6 @@
 {
   "do_normalize": true,
+  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
   "feature_size": 1,
   "padding_side": "right",
   "padding_value": 0.0,

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1cf13d5c227c5f3448e9096bed5bfc3d12f580e4c0059e6120dbbc52627a1483
-size 1262093719
+oid sha256:71c8f04a7a2046ac58b1b41c693b0e330b00a8724a9108487922c1cf6bd006d3
+size 1262091761

vocab.json

@@ -1 +1 @@
-{"б": 0, "а": 1, "н": 2, "р": 3, "т": 4, "ё": 5, "п": 7, "ф": 8, "з": 9, "с": 10, "ш": 11, "д": 12, "г": 13, "ы": 14, "ө": 15, "е": 16, "у": 17, "л": 18, "ч": 19, "ю": 20, "ң": 21, "ә": 22, "ц": 23, "и": 24, "в": 25, "к": 26, "й": 27, "җ": 28, "я": 29, "м": 30, "х": 31, "о": 32, "э": 33, "һ": 34, "ү": 35, "ь": 36, "|": 6, "[UNK]": 37, "[PAD]": 38}
+{"н": 0, "т": 1, "л": 2, "а": 3, "р": 4, "й": 5, "м": 6, "п": 7, "ө": 8, "ф": 9, "һ": 10, "ш": 11, "җ": 12, "о": 13, "в": 14, "ц": 15, "з": 16, "ә": 17, "и": 18, "э": 19, "к": 20, "ж": 21, "д": 22, "ъ": 23, "ь": 24, "е": 25, "ч": 26, "б": 27, "у": 28, "я": 30, "с": 31, "х": 32, "ы": 33, "г": 34, "ё": 35, "ң": 36, "ү": 37, "ю": 38, "|": 29, "[UNK]": 39, "[PAD]": 40}