Training in progress, step 181

config.json

@@ -1,6 +1,6 @@
 {
-  "_name_or_path": "facebook/mms-1b-all",
-  "activation_dropout": 0.05,
+  "_name_or_path": "facebook/wav2vec2-large-960h",
+  "activation_dropout": 0.1,
   "adapter_attn_dim": 16,
   "adapter_kernel_size": 3,
   "adapter_stride": 2,
@@ -9,12 +9,12 @@
   "architectures": [
     "Wav2Vec2ForCTC"
   ],
-  "attention_dropout": 0.05,
+  "attention_dropout": 0.1,
   "bos_token_id": 1,
   "classifier_proj_size": 256,
-  "codevector_dim": 1024,
+  "codevector_dim": 256,
   "contrastive_logits_temperature": 0.1,
-  "conv_bias": true,
+  "conv_bias": false,
   "conv_dim": [
     512,
     512,
@@ -45,19 +45,20 @@
   "ctc_loss_reduction": "mean",
   "ctc_zero_infinity": false,
   "diversity_loss_weight": 0.1,
-  "do_stable_layer_norm": true,
+  "do_stable_layer_norm": false,
   "eos_token_id": 2,
   "feat_extract_activation": "gelu",
   "feat_extract_dropout": 0.0,
-  "feat_extract_norm": "layer",
-  "feat_proj_dropout": 0.05,
+  "feat_extract_norm": "group",
+  "feat_proj_dropout": 0.0,
   "feat_quantizer_dropout": 0.0,
   "final_dropout": 0.0,
   "hidden_act": "gelu",
-  "hidden_dropout": 0.05,
-  "hidden_size": 1280,
+  "hidden_dropout": 0.1,
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
   "initializer_range": 0.02,
-  "intermediate_size": 5120,
+  "intermediate_size": 4096,
   "layer_norm_eps": 1e-05,
   "layerdrop": 0.0,
   "mask_feature_length": 10,
@@ -74,11 +75,11 @@
   "num_conv_pos_embedding_groups": 16,
   "num_conv_pos_embeddings": 128,
   "num_feat_extract_layers": 7,
-  "num_hidden_layers": 48,
+  "num_hidden_layers": 24,
   "num_negatives": 100,
-  "output_hidden_size": 1280,
+  "output_hidden_size": 1024,
   "pad_token_id": 28,
-  "proj_codevector_dim": 1024,
+  "proj_codevector_dim": 256,
   "tdnn_dilation": [
     1,
     2,

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:823a23b3d6d64dd0cff8f9b3c8f3ec93f39133246a98b4b09f3e2626962e66b3
-size 3858890924
+oid sha256:bfaf90cd06605a87819c02506ddef94177262c0279a6ecc725a305ceb6f60ddd
+size 1261893268

preprocessor_config.json

@@ -3,8 +3,8 @@
   "feature_extractor_type": "Wav2Vec2FeatureExtractor",
   "feature_size": 1,
   "padding_side": "right",
-  "padding_value": 0,
+  "padding_value": 0.0,
   "processor_class": "Wav2Vec2Processor",
-  "return_attention_mask": true,
+  "return_attention_mask": false,
   "sampling_rate": 16000
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a1e93d8769125c0125899da107fc172884c089b8493991d9c236433e62d910d5
+oid sha256:29b72b8d3c424e887258d91160b3bfdcf483e9cc9121bae8531fddff1694e827
 size 5432