update to pretraining

config.json

@@ -2,10 +2,12 @@
   "activation_dropout": 0.1,
   "apply_spec_augment": true,
   "architectures": [
-    "Wav2Vec2Model"
+    "Wav2Vec2ForPreTraining"
   ],
   "attention_dropout": 0.1,
   "bos_token_id": 1,
+  "codevector_dim": 256,
+  "contrastive_logits_temperature": 0.1,
   "conv_bias": false,
   "conv_dim": [
     512,
@@ -36,12 +38,14 @@
   ],
   "ctc_loss_reduction": "sum",
   "ctc_zero_infinity": false,
+  "diversity_loss_weight": 0.1,
   "do_stable_layer_norm": false,
   "eos_token_id": 2,
   "feat_extract_activation": "gelu",
   "feat_extract_dropout": 0.0,
   "feat_extract_norm": "group",
   "feat_proj_dropout": 0.1,
+  "feat_quantizer_dropout": 0.0,
   "final_dropout": 0.1,
   "gradient_checkpointing": false,
   "hidden_act": "gelu",
@@ -58,11 +62,15 @@
   "mask_time_prob": 0.05,
   "model_type": "wav2vec2",
   "num_attention_heads": 12,
+  "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_negatives": 100,
   "pad_token_id": 0,
-  "transformers_version": "4.6.0.dev0",
+  "proj_codevector_dim": 256,
+  "transformers_version": "4.7.0.dev0",
   "vocab_size": 32
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aa857b6ddb4b5299ca421cf18667fd8ec94176d326f0d127c31c104684e452d6
-size 377569754
+oid sha256:234e18c3331fa48b2d4c129fb451156e12717899ee5b1f2e78599865fde2e9dd
+size 380267417