Training in progress, step 400

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "facebook/wav2vec2-base",
+  "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
   "activation_dropout": 0.0,
   "adapter_kernel_size": 3,
   "adapter_stride": 2,
@@ -11,9 +11,9 @@
   "attention_dropout": 0.1,
   "bos_token_id": 1,
   "classifier_proj_size": 256,
-  "codevector_dim": 256,
+  "codevector_dim": 768,
   "contrastive_logits_temperature": 0.1,
-  "conv_bias": false,
+  "conv_bias": true,
   "conv_dim": [
     512,
     512,
@@ -44,21 +44,22 @@
   "ctc_loss_reduction": "mean",
   "ctc_zero_infinity": false,
   "diversity_loss_weight": 0.1,
-  "do_stable_layer_norm": false,
+  "do_stable_layer_norm": true,
   "eos_token_id": 2,
   "feat_extract_activation": "gelu",
-  "feat_extract_norm": "group",
+  "feat_extract_dropout": 0.0,
+  "feat_extract_norm": "layer",
   "feat_proj_dropout": 0.1,
   "feat_quantizer_dropout": 0.0,
   "final_dropout": 0.0,
-  "freeze_feat_extract_train": true,
+  "gradient_checkpointing": false,
   "hidden_act": "gelu",
   "hidden_dropout": 0.1,
-  "hidden_size": 768,
+  "hidden_size": 1024,
   "initializer_range": 0.02,
-  "intermediate_size": 3072,
+  "intermediate_size": 4096,
   "layer_norm_eps": 1e-05,
-  "layerdrop": 0.0,
+  "layerdrop": 0.1,
   "mask_channel_length": 10,
   "mask_channel_min_space": 1,
   "mask_channel_other": 0.0,
@@ -71,23 +72,21 @@
   "mask_time_min_masks": 2,
   "mask_time_min_space": 1,
   "mask_time_other": 0.0,
-  "mask_time_prob": 0.05,
+  "mask_time_prob": 0.075,
   "mask_time_selection": "static",
   "model_type": "wav2vec2",
-  "no_mask_channel_overlap": false,
-  "no_mask_time_overlap": false,
   "num_adapter_layers": 3,
-  "num_attention_heads": 12,
+  "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": 12,
+  "num_hidden_layers": 24,
   "num_negatives": 100,
-  "output_hidden_size": 768,
+  "output_hidden_size": 1024,
   "pad_token_id": 69,
-  "proj_codevector_dim": 256,
+  "proj_codevector_dim": 768,
   "tdnn_dilation": [
     1,
     2,
@@ -110,8 +109,8 @@
     1
   ],
   "torch_dtype": "float32",
-  "transformers_version": "4.26.0",
+  "transformers_version": "4.26.1",
   "use_weighted_layer_sum": false,
-  "vocab_size": 128,
+  "vocab_size": 70,
   "xvector_output_dim": 512
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a5c130d887c022ccc5b21aca7ea96bbd8c2a2467ed9770bdd62985caddad0a7c
-size 377954081
+oid sha256:2ebfddb0530761bfcbf61046f8cbcc619fbed78e873799bf86a22e740c132273
+size 1262188845

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9ef3ae782dfc7b4eede3b53656d955faf2ed471711d1e8141acbc7ac28f6de28
-size 3515
+oid sha256:ce2c0c66a21281345382990dca0beda63eb3d9666127009f53b51a84eaeef0a7
+size 3579