hf-tiny-model-private/tiny-random-WavLMForCTC
010
1{2 "activation_dropout": 0.1,3 "adapter_kernel_size": 3,4 "adapter_stride": 2,5 "add_adapter": false,6 "apply_spec_augment": true,7 "architectures": [8 "WavLMForCTC"9 ],10 "attention_dropout": 0.1,11 "bos_token_id": 1,12 "classifier_proj_size": 256,13 "codevector_dim": 256,14 "contrastive_logits_temperature": 0.1,15 "conv_bias": false,16 "conv_dim": [17 32,18 32,19 3220 ],21 "conv_kernel": [22 8,23 8,24 825 ],26 "conv_stride": [27 4,28 4,29 430 ],31 "ctc_loss_reduction": "mean",32 "ctc_zero_infinity": false,33 "diversity_loss_weight": 0.1,34 "do_stable_layer_norm": false,35 "eos_token_id": 2,36 "feat_extract_activation": "gelu",37 "feat_extract_dropout": 0.0,38 "feat_extract_norm": "group",39 "feat_proj_dropout": 0.0,40 "final_dropout": 0.1,41 "hidden_act": "gelu",42 "hidden_dropout": 0.1,43 "hidden_dropout_prob": 0.1,44 "hidden_size": 16,45 "initializer_range": 0.02,46 "intermediate_size": 20,47 "layer_norm_eps": 1e-05,48 "layerdrop": 0.1,49 "mask_feature_length": 10,50 "mask_feature_prob": 0.0,51 "mask_time_length": 10,52 "mask_time_min_masks": 2,53 "mask_time_prob": 0.05,54 "max_bucket_distance": 800,55 "model_type": "wavlm",56 "num_adapter_layers": 3,57 "num_attention_heads": 2,58 "num_buckets": 320,59 "num_codevector_groups": 2,60 "num_codevectors_per_group": 320,61 "num_conv_pos_embedding_groups": 2,62 "num_conv_pos_embeddings": 16,63 "num_ctc_classes": 80,64 "num_feat_extract_layers": 3,65 "num_hidden_layers": 4,66 "num_negatives": 100,67 "output_hidden_size": 16,68 "pad_token_id": 0,69 "proj_codevector_dim": 256,70 "tdnn_dilation": [71 1,72 173 ],74 "tdnn_dim": [75 32,76 3277 ],78 "tdnn_kernel": [79 3,80 381 ],82 "torch_dtype": "float32",83 "transformers_version": "4.28.0.dev0",84 "use_weighted_layer_sum": false,85 "vocab_size": 32,86 "xvector_output_dim": 3287}88 