| { | |
| "_name_or_path": "/net/scratch/hscra/plgrid/plgstefanop/checkpoints/speechlmm-v1/llava-pretrain-audio-seamless-qformer-auto_avsr-mlp-llama_3_1b-speechlmm/v1/s/checkpoint-800", | |
| "add_all_multimodal_tokens": true, | |
| "add_lm_head": true, | |
| "align_text_to_audio": false, | |
| "architectures": [ | |
| "LlavaModel" | |
| ], | |
| "audio_adapter": { | |
| "add_cross_attention": true, | |
| "attention_probs_dropout_prob": 0.1, | |
| "compress_factor": 2, | |
| "cross_attention_every_n_layers": 1, | |
| "hidden_act": "gelu", | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 768, | |
| "hop_size": 0, | |
| "initializer_range": 0.02, | |
| "input_dim": 1024, | |
| "intermediate_size": 3072, | |
| "layer_norm_eps": 1e-12, | |
| "model_type": "qformer", | |
| "num_attention_heads": 12, | |
| "num_hidden_layers": 4, | |
| "num_queries": 1, | |
| "output_dim": 2048, | |
| "transformers_version": "4.45.0", | |
| "triplet_loss": false, | |
| "window_size_in_seconds": 0.3333333333333 | |
| }, | |
| "audio_encoder": { | |
| "_name_or_path": "meetween/seamless-m4t-v2-large-speech-encoder", | |
| "activation_dropout": 0.0, | |
| "activation_function": "relu", | |
| "adaptor_dropout": 0.1, | |
| "adaptor_kernel_size": 8, | |
| "adaptor_stride": 8, | |
| "add_adapter": true, | |
| "architectures": [ | |
| "SeamlessM4Tv2SpeechEncoder" | |
| ], | |
| "attention_dropout": 0.1, | |
| "bos_token_id": 2, | |
| "char_vocab_size": 10943, | |
| "conv_depthwise_kernel_size": 31, | |
| "decoder_attention_heads": 16, | |
| "decoder_ffn_dim": 8192, | |
| "decoder_layerdrop": 0.05, | |
| "decoder_layers": 24, | |
| "decoder_start_token_id": 3, | |
| "dropout": 0.1, | |
| "encoder_attention_heads": 16, | |
| "encoder_ffn_dim": 8192, | |
| "encoder_layerdrop": 0.05, | |
| "encoder_layers": 24, | |
| "eos_token_id": 3, | |
| "feature_projection_input_dim": 160, | |
| "hidden_size": 1024, | |
| "initializer_range": 0.02, | |
| "is_encoder_decoder": true, | |
| "lang_embed_dim": 256, | |
| "layer_norm_eps": 1e-05, | |
| "leaky_relu_slope": 0.1, | |
| "left_max_position_embeddings": 64, | |
| "max_new_tokens": 256, | |
| "max_position_embeddings": 4096, | |
| "model_type": "seamless_m4t_v2", | |
| "num_adapter_layers": 1, | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 24, | |
| "pad_token_id": 0, | |
| "position_embeddings_type": "relative_key", | |
| "resblock_dilation_sizes": [ | |
| [ | |
| 1, | |
| 3, | |
| 5 | |
| ], | |
| [ | |
| 1, | |
| 3, | |
| 5 | |
| ], | |
| [ | |
| 1, | |
| 3, | |
| 5 | |
| ] | |
| ], | |
| "resblock_kernel_sizes": [ | |
| 3, | |
| 7, | |
| 11 | |
| ], | |
| "right_max_position_embeddings": 8, | |
| "sampling_rate": 16000, | |
| "scale_embedding": true, | |
| "speech_encoder_attention_heads": 16, | |
| "speech_encoder_chunk_size": 20000, | |
| "speech_encoder_dropout": 0.0, | |
| "speech_encoder_hidden_act": "swish", | |
| "speech_encoder_intermediate_size": 4096, | |
| "speech_encoder_layerdrop": 0.1, | |
| "speech_encoder_layers": 24, | |
| "speech_encoder_left_chunk_num": 128, | |
| "spkr_embed_dim": 256, | |
| "t2u_bos_token_id": 0, | |
| "t2u_decoder_attention_heads": 16, | |
| "t2u_decoder_ffn_dim": 8192, | |
| "t2u_decoder_layers": 6, | |
| "t2u_encoder_attention_heads": 16, | |
| "t2u_encoder_ffn_dim": 8192, | |
| "t2u_encoder_layers": 6, | |
| "t2u_eos_token_id": 2, | |
| "t2u_max_position_embeddings": 4096, | |
| "t2u_pad_token_id": 1, | |
| "t2u_variance_pred_dropout": 0.5, | |
| "t2u_variance_predictor_embed_dim": 1024, | |
| "t2u_variance_predictor_hidden_dim": 256, | |
| "t2u_variance_predictor_kernel_size": 3, | |
| "t2u_vocab_size": 10082, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.45.0", | |
| "unit_embed_dim": 1280, | |
| "unit_hifi_gan_vocab_size": 10000, | |
| "upsample_initial_channel": 512, | |
| "upsample_kernel_sizes": [ | |
| 11, | |
| 8, | |
| 8, | |
| 4, | |
| 4 | |
| ], | |
| "upsample_rates": [ | |
| 5, | |
| 4, | |
| 4, | |
| 2, | |
| 2 | |
| ], | |
| "use_cache": true, | |
| "var_pred_dropout": 0.5, | |
| "variance_predictor_kernel_size": 3, | |
| "vocab_size": 256102, | |
| "vocoder_num_langs": 36, | |
| "vocoder_num_spkrs": 200, | |
| "vocoder_offset": 4 | |
| }, | |
| "chunk_encoding_strategy": "loop", | |
| "chunk_overlap_in_seconds": 1, | |
| "chunk_size_in_seconds": 15, | |
| "codebook_weights": [ | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1, | |
| 1 | |
| ], | |
| "conversation_version": null, | |
| "hidden_size": 2048, | |
| "mm_use_audio_start_end": false, | |
| "mm_use_im_start_end": false, | |
| "mm_use_video_start_end": false, | |
| "model_type": "llava", | |
| "text_decoder": { | |
| "_name_or_path": "meta-llama/Llama-3.2-1B-Instruct", | |
| "architectures": [ | |
| "LlamaForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 128000, | |
| "conversation_version": "llama_3_1", | |
| "eos_token_id": [ | |
| 128001, | |
| 128008, | |
| 128009 | |
| ], | |
| "head_dim": 64, | |
| "hidden_act": "silu", | |
| "hidden_size": 2048, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 8192, | |
| "max_position_embeddings": 131072, | |
| "mlp_bias": false, | |
| "model_type": "llama", | |
| "num_attention_heads": 32, | |
| "num_hidden_layers": 16, | |
| "num_key_value_heads": 8, | |
| "pretraining_tp": 1, | |
| "rms_norm_eps": 1e-05, | |
| "rope_scaling": { | |
| "factor": 32.0, | |
| "high_freq_factor": 4.0, | |
| "low_freq_factor": 1.0, | |
| "original_max_position_embeddings": 8192, | |
| "rope_type": "llama3" | |
| }, | |
| "rope_theta": 500000.0, | |
| "tie_word_embeddings": true, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.45.0", | |
| "use_cache": true, | |
| "vocab_size": 128256 | |
| }, | |
| "tokenizer_padding_side": "right", | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.45.0", | |
| "use_cache": true, | |
| "video_adapter": { | |
| "force_input_projection": true, | |
| "force_output_projection": true, | |
| "hidden_layers": 4, | |
| "hidden_size": 4096, | |
| "input_dim": 768, | |
| "model_type": "mlp", | |
| "output_dim": 2048, | |
| "residual_type": "interpolation", | |
| "transformers_version": "4.45.0" | |
| }, | |
| "video_encoder": { | |
| "_name_or_path": "/net/scratch/hscra/plgrid/plgstefanop/auto_avsr_ckpt/vsr_trlrs3vox2_base.pth", | |
| "a_upsample_ratio": 1, | |
| "adim": 768, | |
| "aheads": 12, | |
| "cnn_module_kernel": 31, | |
| "ctc_type": "builtin", | |
| "ddim": 768, | |
| "dheads": 12, | |
| "dlayers": 6, | |
| "dropout_rate": 0.1, | |
| "dunits": 3072, | |
| "elayers": 12, | |
| "eunits": 3072, | |
| "hidden_size": 768, | |
| "lsm_weight": 0.1, | |
| "macaron_style": true, | |
| "mtlalpha": 0.1, | |
| "rel_pos_type": "latest", | |
| "relu_type": "swish", | |
| "transformer_attn_dropout_rate": 0.1, | |
| "transformer_encoder_attn_layer_type": "rel_mha", | |
| "transformer_input_layer": "conv3d", | |
| "transformer_length_normalized_loss": false, | |
| "transformers_version": "4.45.0", | |
| "use_cnn_module": true, | |
| "zero_triu": false | |
| }, | |
| "vision_patch_merge_type": "flat", | |
| "vision_select_feature": "patch", | |
| "vision_select_layer": -1, | |
| "vision_use_patch_token": true | |
| } | |