BuzzASR/catalan
019
1{2 "activation_dropout": 0.0,3 "activation_function": "gelu",4 "apply_spec_augment": false,5 "architectures": [6 "WhisperForConditionalGeneration"7 ],8 "attention_dropout": 0.0,9 "begin_suppress_tokens": null,10 "bos_token_id": 51865,11 "classifier_proj_size": 256,12 "d_model": 1280,13 "decoder_attention_heads": 20,14 "decoder_ffn_dim": 5120,15 "decoder_layerdrop": 0.0,16 "decoder_layers": 32,17 "decoder_start_token_id": 51866,18 "dropout": 0.0,19 "dtype": "float16",20 "encoder_attention_heads": 20,21 "encoder_ffn_dim": 5120,22 "encoder_layerdrop": 0.0,23 "encoder_layers": 32,24 "eos_token_id": 51865,25 "init_std": 0.02,26 "is_encoder_decoder": true,27 "mask_feature_length": 10,28 "mask_feature_min_masks": 0,29 "mask_feature_prob": 0.0,30 "mask_time_length": 10,31 "mask_time_min_masks": 2,32 "mask_time_prob": 0.05,33 "max_length": null,34 "max_source_positions": 1500,35 "max_target_positions": 448,36 "median_filter_width": 7,37 "model_type": "whisper",38 "num_hidden_layers": 32,39 "num_mel_bins": 128,40 "pad_token_id": 51865,41 "scale_embedding": false,42 "transformers_version": "4.57.6",43 "use_cache": true,44 "use_weighted_layer_sum": false,45 "vocab_size": 5347446}47 