Training in progress, step 1000

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,
@@ -8,10 +8,10 @@
   "architectures": [
     "Wav2Vec2ForCTC"
   ],
-  "attention_dropout": 0.05,
+  "attention_dropout": 0.08,
   "bos_token_id": 1,
   "classifier_proj_size": 256,
-  "codevector_dim": 256,
+  "codevector_dim": 768,
   "contrastive_logits_temperature": 0.1,
   "conv_bias": true,
   "conv_dim": [
@@ -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": 0.05,
-  "hidden_size": 768,
+  "hidden_dropout": 0.08,
+  "hidden_size": 1024,
   "initializer_range": 0.02,
-  "intermediate_size": 3072,
+  "intermediate_size": 4096,
   "layer_norm_eps": 1e-05,
-  "layerdrop": 0.2,
+  "layerdrop": 0.3,
   "mask_channel_length": 10,
   "mask_channel_min_space": 1,
   "mask_channel_other": 0.0,
@@ -74,8 +75,6 @@
   "mask_time_prob": 0.05,
   "mask_time_selection": "static",
   "model_type": "wav2vec2",
-  "no_mask_channel_overlap": false,
-  "no_mask_time_overlap": false,
   "num_adapter_layers": 3,
   "num_attention_heads": 16,
   "num_codevector_groups": 2,
@@ -83,11 +82,11 @@
   "num_conv_pos_embedding_groups": 16,
   "num_conv_pos_embeddings": 128,
   "num_feat_extract_layers": 7,
-  "num_hidden_layers": 24,
+  "num_hidden_layers": 32,
   "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,

preprocessor_config.json

@@ -3,7 +3,7 @@
   "feature_extractor_type": "Wav2Vec2FeatureExtractor",
   "feature_size": 1,
   "padding_side": "right",
-  "padding_value": 0.0,
-  "return_attention_mask": false,
+  "padding_value": 0,
+  "return_attention_mask": true,
   "sampling_rate": 16000
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d1217d75502a142460de28f023cf7d4fa4bfe30ecab4c614182997eecabcdf2
-size 718083261
+oid sha256:b261e4b4d3846cfbfcb603c8fa38a8e3f88afaf9ac1bd2fdd425bcc6afc5f3e4
+size 1665320813

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b8f8259a4ab001550004b3f231ee778163835ab2721b3c8374fe04a87599f5d6
+oid sha256:594a60f4a455ea0cdc893f22cd0135a5907ce7b74610de32c267fad3518421f7
 size 3579