godofmining/venus_v1

1{2  "_name_or_path": "RayneAmes-pikachu_v2_venus_v1",3  "architectures": [4    "ParlerTTSForConditionalGeneration"5  ],6  "audio_encoder": {7    "_attn_implementation_autoset": false,8    "_name_or_path": "parler-tts/dac_44khZ_8kbps",9    "add_cross_attention": false,10    "architectures": [11      "DACModel"12    ],13    "bad_words_ids": null,14    "begin_suppress_tokens": null,15    "bos_token_id": null,16    "chunk_size_feed_forward": 0,17    "codebook_size": 1024,18    "cross_attention_hidden_size": null,19    "decoder_start_token_id": null,20    "diversity_penalty": 0.0,21    "do_sample": false,22    "early_stopping": false,23    "encoder_no_repeat_ngram_size": 0,24    "eos_token_id": null,25    "exponential_decay_length_penalty": null,26    "finetuning_task": null,27    "forced_bos_token_id": null,28    "forced_eos_token_id": null,29    "frame_rate": 86,30    "id2label": {31      "0": "LABEL_0",32      "1": "LABEL_1"33    },34    "is_decoder": false,35    "is_encoder_decoder": false,36    "label2id": {37      "LABEL_0": 0,38      "LABEL_1": 139    },40    "latent_dim": 1024,41    "length_penalty": 1.0,42    "max_length": 20,43    "min_length": 0,44    "model_bitrate": 8,45    "model_type": "dac_on_the_hub",46    "no_repeat_ngram_size": 0,47    "num_beam_groups": 1,48    "num_beams": 1,49    "num_codebooks": 9,50    "num_return_sequences": 1,51    "output_attentions": false,52    "output_hidden_states": false,53    "output_scores": false,54    "pad_token_id": null,55    "prefix": null,56    "problem_type": null,57    "pruned_heads": {},58    "remove_invalid_values": false,59    "repetition_penalty": 1.0,60    "return_dict": true,61    "return_dict_in_generate": false,62    "sampling_rate": 44100,63    "sep_token_id": null,64    "suppress_tokens": null,65    "task_specific_params": null,66    "temperature": 1.0,67    "tf_legacy_loss": false,68    "tie_encoder_decoder": false,69    "tie_word_embeddings": true,70    "tokenizer_class": null,71    "top_k": 50,72    "top_p": 1.0,73    "torch_dtype": "float32",74    "torchscript": false,75    "typical_p": 1.0,76    "use_bfloat16": false77  },78  "decoder": {79    "_attn_implementation_autoset": false,80    "_name_or_path": "/fsx/yoach/tmp/artefacts/parler-tts-mini/decoder",81    "activation_dropout": 0.0,82    "activation_function": "gelu",83    "add_cross_attention": true,84    "architectures": [85      "ParlerTTSForCausalLM"86    ],87    "attention_dropout": 0.0,88    "bad_words_ids": null,89    "begin_suppress_tokens": null,90    "bos_token_id": 1025,91    "chunk_size_feed_forward": 0,92    "codebook_weights": null,93    "cross_attention_hidden_size": null,94    "cross_attention_implementation_strategy": null,95    "decoder_start_token_id": null,96    "diversity_penalty": 0.0,97    "do_sample": false,98    "dropout": 0.1,99    "early_stopping": false,100    "encoder_no_repeat_ngram_size": 0,101    "eos_token_id": 1024,102    "exponential_decay_length_penalty": null,103    "ffn_dim": 4096,104    "finetuning_task": null,105    "forced_bos_token_id": null,106    "forced_eos_token_id": null,107    "hidden_size": 1024,108    "id2label": {109      "0": "LABEL_0",110      "1": "LABEL_1"111    },112    "initializer_factor": 0.02,113    "is_decoder": true,114    "is_encoder_decoder": false,115    "label2id": {116      "LABEL_0": 0,117      "LABEL_1": 1118    },119    "layerdrop": 0.0,120    "length_penalty": 1.0,121    "max_length": 20,122    "max_position_embeddings": 4311,123    "min_length": 0,124    "model_type": "parler_tts_decoder",125    "no_repeat_ngram_size": 0,126    "num_attention_heads": 16,127    "num_beam_groups": 1,128    "num_beams": 1,129    "num_codebooks": 9,130    "num_cross_attention_key_value_heads": 16,131    "num_hidden_layers": 24,132    "num_key_value_heads": 16,133    "num_return_sequences": 1,134    "output_attentions": false,135    "output_hidden_states": false,136    "output_scores": false,137    "pad_token_id": 1024,138    "prefix": null,139    "problem_type": null,140    "pruned_heads": {},141    "remove_invalid_values": false,142    "repetition_penalty": 1.0,143    "return_dict": true,144    "return_dict_in_generate": false,145    "rope_embeddings": false,146    "rope_theta": 10000.0,147    "scale_embedding": false,148    "sep_token_id": null,149    "suppress_tokens": null,150    "task_specific_params": null,151    "temperature": 1.0,152    "tf_legacy_loss": false,153    "tie_encoder_decoder": false,154    "tie_word_embeddings": false,155    "tokenizer_class": null,156    "top_k": 50,157    "top_p": 1.0,158    "torch_dtype": "float32",159    "torchscript": false,160    "typical_p": 1.0,161    "use_bfloat16": false,162    "use_cache": true,163    "use_fused_lm_heads": false,164    "vocab_size": 1088165  },166  "decoder_start_token_id": 1025,167  "fine_tuned": true,168  "hidden_dropout_prob": 0.00030000000000000003,169  "is_encoder_decoder": true,170  "model_type": "parler_tts",171  "model_version": "1.1-modified-venus-v1",172  "pad_token_id": 1024,173  "prompt_cross_attention": false,174  "text_encoder": {175    "_attn_implementation_autoset": false,176    "_name_or_path": "google/flan-t5-large",177    "add_cross_attention": false,178    "architectures": [179      "T5ForConditionalGeneration"180    ],181    "bad_words_ids": null,182    "begin_suppress_tokens": null,183    "bos_token_id": null,184    "chunk_size_feed_forward": 0,185    "classifier_dropout": 0.0,186    "cross_attention_hidden_size": null,187    "d_ff": 2816,188    "d_kv": 64,189    "d_model": 1024,190    "decoder_start_token_id": 0,191    "dense_act_fn": "gelu_new",192    "diversity_penalty": 0.0,193    "do_sample": false,194    "dropout_rate": 0.1,195    "early_stopping": false,196    "encoder_no_repeat_ngram_size": 0,197    "eos_token_id": 1,198    "exponential_decay_length_penalty": null,199    "feed_forward_proj": "gated-gelu",200    "finetuning_task": null,201    "forced_bos_token_id": null,202    "forced_eos_token_id": null,203    "id2label": {204      "0": "LABEL_0",205      "1": "LABEL_1"206    },207    "initializer_factor": 1.0,208    "is_decoder": false,209    "is_encoder_decoder": true,210    "is_gated_act": true,211    "label2id": {212      "LABEL_0": 0,213      "LABEL_1": 1214    },215    "layer_norm_epsilon": 1e-06,216    "length_penalty": 1.0,217    "max_length": 20,218    "min_length": 0,219    "model_type": "t5",220    "n_positions": 512,221    "no_repeat_ngram_size": 0,222    "num_beam_groups": 1,223    "num_beams": 1,224    "num_decoder_layers": 24,225    "num_heads": 16,226    "num_layers": 24,227    "num_return_sequences": 1,228    "output_attentions": false,229    "output_hidden_states": false,230    "output_past": true,231    "output_scores": false,232    "pad_token_id": 0,233    "prefix": null,234    "problem_type": null,235    "pruned_heads": {},236    "relative_attention_max_distance": 128,237    "relative_attention_num_buckets": 32,238    "remove_invalid_values": false,239    "repetition_penalty": 1.0,240    "return_dict": true,241    "return_dict_in_generate": false,242    "sep_token_id": null,243    "suppress_tokens": null,244    "task_specific_params": null,245    "temperature": 1.0,246    "tf_legacy_loss": false,247    "tie_encoder_decoder": false,248    "tie_word_embeddings": false,249    "tokenizer_class": null,250    "top_k": 50,251    "top_p": 1.0,252    "torch_dtype": null,253    "torchscript": false,254    "typical_p": 1.0,255    "use_bfloat16": false,256    "use_cache": true,257    "vocab_size": 32128258  },259  "torch_dtype": "float32",260  "transformers_version": "4.46.1",261  "vocab_size": 32128262}263