| { | |
| "_attn_implementation_autoset": true, | |
| "_name_or_path": "efficient-speech/lite-whisper-large-v3-turbo-acc", | |
| "activation_dropout": 0.0, | |
| "activation_function": "gelu", | |
| "apply_spec_augment": false, | |
| "architectures": [ | |
| "LiteWhisperForConditionalGeneration" | |
| ], | |
| "attention_dropout": 0.0, | |
| "begin_suppress_tokens": [ | |
| 220, | |
| 50256 | |
| ], | |
| "bos_token_id": 50257, | |
| "classifier_proj_size": 256, | |
| "d_model": 1280, | |
| "decoder_attention_heads": 20, | |
| "decoder_ffn_dim": 5120, | |
| "decoder_layerdrop": 0.0, | |
| "decoder_layers": 4, | |
| "decoder_start_token_id": 50258, | |
| "dropout": 0.0, | |
| "encoder_attention_heads": 20, | |
| "encoder_ffn_dim": 5120, | |
| "encoder_layerdrop": 0.0, | |
| "encoder_layers": 32, | |
| "eos_token_id": 50257, | |
| "init_std": 0.02, | |
| "is_encoder_decoder": true, | |
| "low_rank_config": [ | |
| { | |
| "fc1": 352, | |
| "fc2": 304, | |
| "k_proj": 48, | |
| "out_proj": 48, | |
| "q_proj": 48, | |
| "v_proj": 64 | |
| }, | |
| { | |
| "fc1": 240, | |
| "fc2": 256, | |
| "k_proj": 80, | |
| "out_proj": 192, | |
| "q_proj": 96, | |
| "v_proj": 192 | |
| }, | |
| { | |
| "fc1": 192, | |
| "fc2": 144, | |
| "k_proj": 64, | |
| "out_proj": 224, | |
| "q_proj": 80, | |
| "v_proj": 240 | |
| }, | |
| { | |
| "fc1": 256, | |
| "fc2": 128, | |
| "k_proj": 80, | |
| "out_proj": 288, | |
| "q_proj": 80, | |
| "v_proj": 336 | |
| }, | |
| { | |
| "fc1": 272, | |
| "fc2": 192, | |
| "k_proj": 96, | |
| "out_proj": 224, | |
| "q_proj": 96, | |
| "v_proj": 368 | |
| }, | |
| { | |
| "fc1": 272, | |
| "fc2": 304, | |
| "k_proj": 80, | |
| "out_proj": 256, | |
| "q_proj": 80, | |
| "v_proj": 352 | |
| }, | |
| { | |
| "fc1": 352, | |
| "fc2": 320, | |
| "k_proj": 128, | |
| "out_proj": 368, | |
| "q_proj": 144, | |
| "v_proj": 576 | |
| }, | |
| { | |
| "fc1": 368, | |
| "fc2": 560, | |
| "k_proj": 128, | |
| "out_proj": 384, | |
| "q_proj": 160, | |
| "v_proj": 512 | |
| }, | |
| { | |
| "fc1": 464, | |
| "fc2": 672, | |
| "k_proj": 144, | |
| "out_proj": 416, | |
| "q_proj": 176, | |
| "v_proj": 544 | |
| }, | |
| { | |
| "fc1": 464, | |
| "fc2": 688, | |
| "k_proj": 160, | |
| "out_proj": 448, | |
| "q_proj": 192, | |
| "v_proj": 608 | |
| }, | |
| { | |
| "fc1": 528, | |
| "fc2": 720, | |
| "k_proj": 96, | |
| "out_proj": 432, | |
| "q_proj": 128, | |
| "v_proj": 576 | |
| }, | |
| { | |
| "fc1": 512, | |
| "fc2": 720, | |
| "k_proj": 128, | |
| "out_proj": 432, | |
| "q_proj": 160, | |
| "v_proj": 608 | |
| }, | |
| { | |
| "fc1": 544, | |
| "fc2": 800, | |
| "k_proj": 128, | |
| "out_proj": 432, | |
| "q_proj": 160, | |
| "v_proj": 640 | |
| }, | |
| { | |
| "fc1": 624, | |
| "fc2": 928, | |
| "k_proj": 112, | |
| "out_proj": 496, | |
| "q_proj": 128, | |
| "v_proj": 624 | |
| }, | |
| { | |
| "fc1": 704, | |
| "fc2": 880, | |
| "k_proj": 144, | |
| "out_proj": 560, | |
| "q_proj": 176 | |
| }, | |
| { | |
| "fc1": 592, | |
| "fc2": 992, | |
| "k_proj": 80, | |
| "out_proj": 560, | |
| "q_proj": 96 | |
| }, | |
| { | |
| "fc1": 640, | |
| "fc2": 1024, | |
| "k_proj": 96, | |
| "out_proj": 432, | |
| "q_proj": 112, | |
| "v_proj": 576 | |
| }, | |
| { | |
| "fc1": 736, | |
| "k_proj": 160, | |
| "out_proj": 544, | |
| "q_proj": 192 | |
| }, | |
| { | |
| "fc1": 800, | |
| "k_proj": 128, | |
| "out_proj": 480, | |
| "q_proj": 160 | |
| }, | |
| { | |
| "fc1": 848, | |
| "k_proj": 144, | |
| "out_proj": 512, | |
| "q_proj": 160 | |
| }, | |
| { | |
| "fc1": 832, | |
| "fc2": 16, | |
| "k_proj": 160, | |
| "out_proj": 608, | |
| "q_proj": 192 | |
| }, | |
| { | |
| "fc1": 976, | |
| "k_proj": 224, | |
| "q_proj": 240 | |
| }, | |
| { | |
| "fc1": 960, | |
| "k_proj": 272, | |
| "q_proj": 288 | |
| }, | |
| { | |
| "k_proj": 272, | |
| "q_proj": 288 | |
| }, | |
| { | |
| "k_proj": 368, | |
| "q_proj": 384 | |
| }, | |
| { | |
| "k_proj": 336, | |
| "q_proj": 384 | |
| }, | |
| { | |
| "k_proj": 400, | |
| "q_proj": 464 | |
| }, | |
| { | |
| "k_proj": 400, | |
| "q_proj": 448 | |
| }, | |
| { | |
| "k_proj": 432, | |
| "q_proj": 496 | |
| }, | |
| { | |
| "k_proj": 560, | |
| "q_proj": 608 | |
| }, | |
| { | |
| "k_proj": 608 | |
| }, | |
| { | |
| "k_proj": 624 | |
| } | |
| ], | |
| "mask_feature_length": 10, | |
| "mask_feature_min_masks": 0, | |
| "mask_feature_prob": 0.0, | |
| "mask_time_length": 10, | |
| "mask_time_min_masks": 2, | |
| "mask_time_prob": 0.05, | |
| "max_source_positions": 1500, | |
| "max_target_positions": 448, | |
| "median_filter_width": 7, | |
| "model_type": "lite-whisper", | |
| "num_hidden_layers": 32, | |
| "num_mel_bins": 128, | |
| "pad_token_id": 50257, | |
| "scale_embedding": false, | |
| "transformers_version": "4.48.3", | |
| "use_cache": true, | |
| "use_weighted_layer_sum": false, | |
| "vocab_size": 51866 | |
| } | |