config.json
| 1 | { |
| 2 | "activation_dropout": 0.0, |
| 3 | "activation_function": "relu", |
| 4 | "adaptor_dropout": 0.1, |
| 5 | "adaptor_kernel_size": 8, |
| 6 | "adaptor_stride": 8, |
| 7 | "add_adapter": true, |
| 8 | "architectures": [ |
| 9 | "SeamlessM4Tv2Model" |
| 10 | ], |
| 11 | "attention_dropout": 0.1, |
| 12 | "bos_token_id": 2, |
| 13 | "char_vocab_size": 10943, |
| 14 | "conv_depthwise_kernel_size": 31, |
| 15 | "decoder_attention_heads": 16, |
| 16 | "decoder_ffn_dim": 8192, |
| 17 | "decoder_layerdrop": 0.05, |
| 18 | "decoder_layers": 24, |
| 19 | "decoder_start_token_id": 3, |
| 20 | "dropout": 0.1, |
| 21 | "encoder_attention_heads": 16, |
| 22 | "encoder_ffn_dim": 8192, |
| 23 | "encoder_layerdrop": 0.05, |
| 24 | "encoder_layers": 24, |
| 25 | "eos_token_id": 3, |
| 26 | "feature_projection_input_dim": 160, |
| 27 | "hidden_size": 1024, |
| 28 | "initializer_range": 0.02, |
| 29 | "is_encoder_decoder": true, |
| 30 | "lang_embed_dim": 256, |
| 31 | "layer_norm_eps": 1e-05, |
| 32 | "leaky_relu_slope": 0.1, |
| 33 | "left_max_position_embeddings": 64, |
| 34 | "max_new_tokens": 256, |
| 35 | "max_position_embeddings": 4096, |
| 36 | "model_type": "seamless_m4t_v2", |
| 37 | "num_adapter_layers": 1, |
| 38 | "num_attention_heads": 16, |
| 39 | "num_hidden_layers": 24, |
| 40 | "pad_token_id": 0, |
| 41 | "position_embeddings_type": "relative_key", |
| 42 | "resblock_dilation_sizes": [ |
| 43 | [ |
| 44 | 1, |
| 45 | 3, |
| 46 | 5 |
| 47 | ], |
| 48 | [ |
| 49 | 1, |
| 50 | 3, |
| 51 | 5 |
| 52 | ], |
| 53 | [ |
| 54 | 1, |
| 55 | 3, |
| 56 | 5 |
| 57 | ] |
| 58 | ], |
| 59 | "resblock_kernel_sizes": [ |
| 60 | 3, |
| 61 | 7, |
| 62 | 11 |
| 63 | ], |
| 64 | "right_max_position_embeddings": 8, |
| 65 | "sampling_rate": 16000, |
| 66 | "scale_embedding": true, |
| 67 | "speech_encoder_attention_heads": 16, |
| 68 | "speech_encoder_chunk_size": 20000, |
| 69 | "speech_encoder_dropout": 0.0, |
| 70 | "speech_encoder_hidden_act": "swish", |
| 71 | "speech_encoder_intermediate_size": 4096, |
| 72 | "speech_encoder_layerdrop": 0.1, |
| 73 | "speech_encoder_layers": 24, |
| 74 | "speech_encoder_left_chunk_num": 128, |
| 75 | "spkr_embed_dim": 256, |
| 76 | "t2u_bos_token_id": 0, |
| 77 | "t2u_decoder_attention_heads": 16, |
| 78 | "t2u_decoder_ffn_dim": 8192, |
| 79 | "t2u_decoder_layers": 6, |
| 80 | "t2u_encoder_attention_heads": 16, |
| 81 | "t2u_encoder_ffn_dim": 8192, |
| 82 | "t2u_encoder_layers": 6, |
| 83 | "t2u_eos_token_id": 2, |
| 84 | "t2u_max_position_embeddings": 4096, |
| 85 | "t2u_pad_token_id": 1, |
| 86 | "t2u_variance_pred_dropout": 0.5, |
| 87 | "t2u_variance_predictor_embed_dim": 1024, |
| 88 | "t2u_variance_predictor_hidden_dim": 256, |
| 89 | "t2u_variance_predictor_kernel_size": 3, |
| 90 | "t2u_vocab_size": 10082, |
| 91 | "torch_dtype": "float32", |
| 92 | "transformers_version": "4.36.0.dev0", |
| 93 | "unit_embed_dim": 1280, |
| 94 | "unit_hifi_gan_vocab_size": 10000, |
| 95 | "upsample_initial_channel": 512, |
| 96 | "upsample_kernel_sizes": [ |
| 97 | 11, |
| 98 | 8, |
| 99 | 8, |
| 100 | 4, |
| 101 | 4 |
| 102 | ], |
| 103 | "upsample_rates": [ |
| 104 | 5, |
| 105 | 4, |
| 106 | 4, |
| 107 | 2, |
| 108 | 2 |
| 109 | ], |
| 110 | "use_cache": true, |
| 111 | "var_pred_dropout": 0.5, |
| 112 | "variance_predictor_kernel_size": 3, |
| 113 | "vocab_size": 256102, |
| 114 | "vocoder_num_langs": 36, |
| 115 | "vocoder_num_spkrs": 200, |
| 116 | "vocoder_offset": 4 |
| 117 | } |
| 118 | |