{ "architectures": [ "HyperCLOVAXVisionV2ForCausalLM" ], "audio_config": { "architectures": [ "HyperCLOVAXSeedAudioEncoder" ], "auto_map": { "AutoConfig": "configuration_hyperclovax_seed_audio_encoder.HyperCLOVAXSeedAudioEncoderConfig", "AutoModel": "modeling_hyperclovax_seed_audio_encoder.HyperCLOVAXSeedAudioEncoder" }, "attention_dropout": 0.1, "d_model": 768, "dropout": 0.1, "encoder_attention_heads": 12, "encoder_ffn_dim": 3072, "encoder_layers": 12, "max_source_positions": 1500, "model_type": "hyperclovax_seed_audio_encoder", "num_mel_bins": 128 }, "audio_end_token_id": 128074, "audio_pool_size": 5, "audio_projector_type": "mlp", "audio_start_token_id": 128073, "audio_token_id": 128075, "auto_map": { "AutoConfig": "configuration_hyperclovax_seed_vision_v2.HyperCLOVAXVisionV2Config", "AutoModelForCausalLM": "modeling_hyperclovax_seed_vision_v2.HyperCLOVAXVisionV2ForCausalLM" }, "eos_token_id": 128001, "hidden_size": 3584, "ignore_index": -100, "image_end_token_id": 128057, "image_start_token_id": 128056, "image_token_id": 128060, "vision_projector_type": "patch_merger", "model_type": "hyperclovax_vision_v2", "num_queries_vis_abstractor": -1, "pad_token_id": 128001, "proj_pos_emb": true, "proj_prenorm": false, "text_config": { "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 0, "eos_token_id": 128001, "head_dim": 128, "hidden_act": "silu", "hidden_size": 3584, "initializer_range": 0.02, "intermediate_size": 7168, "logits_scaling": 1.0, "max_position_embeddings": 131072, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 36, "num_key_value_heads": 8, "pretraining_tp": 1, "rms_norm_eps": 1e-06, "rope_scaling": null, "rope_theta": 100000000, "tie_word_embeddings": true, "use_cache": false, "vocab_size": 128256 }, "torch_dtype": "float32", "transformers_version": "4.58.0", "video_audio_token_id": 128072, "video_end_token_id": 128059, "video_start_token_id": 128058, "video_token_id": 128061, "vision_config": { "architectures": [ "HyperCLOVAXSeedVisionEncoder" ], "auto_map": { "AutoConfig": "configuration_hyperclovax_seed_vision_encoder.HyperCLOVAXSeedVisionEncoderConfig", "AutoModel": "modeling_hyperclovax_seed_vision_encoder.HyperCLOVAXSeedVisionEncoder" }, "depth": 32, "disable_merger": true, "fullatt_block_indexes": [ 3, 7, 11, 15, 19, 23, 27, 31 ], "hidden_act": "silu", "hidden_size": 1280, "in_channels": 3, "initializer_range": 0.02, "intermediate_size": 3456, "model_type": "hyperclovax_seed_vision_encoder", "num_heads": 16, "out_hidden_size": 5120, "patch_size": 16, "spatial_merge_size": 2, "temporal_patch_size": 2, "tokens_per_second": 2, "window_size": 128 }, "vision_input_chunk_size": null }