{ "architectures": [ "VoxtralForConditionalGeneration" ], "audio_config": { "activation_dropout": 0.0, "activation_function": "gelu", "attention_dropout": 0.0, "dropout": 0.0, "head_dim": 64, "hidden_size": 1280, "initializer_range": 0.02, "intermediate_size": 5120, "layerdrop": 0.0, "max_source_positions": 1500, "model_type": "voxtral_encoder", "num_attention_heads": 20, "num_hidden_layers": 32, "num_key_value_heads": 20, "num_mel_bins": 128, "scale_embedding": false, "torch_dtype": "bfloat16", "vocab_size": 51866 }, "audio_token_id": 24, "hidden_size": 3072, "model_type": "voxtral", "projector_hidden_act": "gelu", "quantization_config": { "config_groups": { "group_0": { "input_activations": { "actorder": null, "block_structure": null, "dynamic": true, "group_size": null, "num_bits": 8, "observer": null, "observer_kwargs": {}, "strategy": "token", "symmetric": true, "type": "float" }, "output_activations": null, "targets": [ "Linear" ], "weights": { "actorder": null, "block_structure": null, "dynamic": false, "group_size": null, "num_bits": 8, "observer": "minmax", "observer_kwargs": {}, "strategy": "channel", "symmetric": true, "type": "float" } } }, "format": "float-quantized", "global_compression_ratio": null, "ignore": [ "audio_tower.layers.0.self_attn.k_proj", "audio_tower.layers.0.self_attn.v_proj", "audio_tower.layers.0.self_attn.q_proj", "audio_tower.layers.0.self_attn.out_proj", "audio_tower.layers.0.fc1", "audio_tower.layers.0.fc2", "audio_tower.layers.1.self_attn.k_proj", "audio_tower.layers.1.self_attn.v_proj", "audio_tower.layers.1.self_attn.q_proj", "audio_tower.layers.1.self_attn.out_proj", "audio_tower.layers.1.fc1", "audio_tower.layers.1.fc2", "audio_tower.layers.2.self_attn.k_proj", "audio_tower.layers.2.self_attn.v_proj", "audio_tower.layers.2.self_attn.q_proj", "audio_tower.layers.2.self_attn.out_proj", "audio_tower.layers.2.fc1", "audio_tower.layers.2.fc2", "audio_tower.layers.3.self_attn.k_proj", "audio_tower.layers.3.self_attn.v_proj", "audio_tower.layers.3.self_attn.q_proj", "audio_tower.layers.3.self_attn.out_proj", "audio_tower.layers.3.fc1", "audio_tower.layers.3.fc2", "audio_tower.layers.4.self_attn.k_proj", "audio_tower.layers.4.self_attn.v_proj", "audio_tower.layers.4.self_attn.q_proj", "audio_tower.layers.4.self_attn.out_proj", "audio_tower.layers.4.fc1", "audio_tower.layers.4.fc2", "audio_tower.layers.5.self_attn.k_proj", "audio_tower.layers.5.self_attn.v_proj", "audio_tower.layers.5.self_attn.q_proj", "audio_tower.layers.5.self_attn.out_proj", "audio_tower.layers.5.fc1", "audio_tower.layers.5.fc2", "audio_tower.layers.6.self_attn.k_proj", "audio_tower.layers.6.self_attn.v_proj", "audio_tower.layers.6.self_attn.q_proj", "audio_tower.layers.6.self_attn.out_proj", "audio_tower.layers.6.fc1", "audio_tower.layers.6.fc2", "audio_tower.layers.7.self_attn.k_proj", "audio_tower.layers.7.self_attn.v_proj", "audio_tower.layers.7.self_attn.q_proj", "audio_tower.layers.7.self_attn.out_proj", "audio_tower.layers.7.fc1", "audio_tower.layers.7.fc2", "audio_tower.layers.8.self_attn.k_proj", "audio_tower.layers.8.self_attn.v_proj", "audio_tower.layers.8.self_attn.q_proj", "audio_tower.layers.8.self_attn.out_proj", "audio_tower.layers.8.fc1", "audio_tower.layers.8.fc2", "audio_tower.layers.9.self_attn.k_proj", "audio_tower.layers.9.self_attn.v_proj", "audio_tower.layers.9.self_attn.q_proj", "audio_tower.layers.9.self_attn.out_proj", "audio_tower.layers.9.fc1", "audio_tower.layers.9.fc2", "audio_tower.layers.10.self_attn.k_proj", "audio_tower.layers.10.self_attn.v_proj", "audio_tower.layers.10.self_attn.q_proj", "audio_tower.layers.10.self_attn.out_proj", "audio_tower.layers.10.fc1", "audio_tower.layers.10.fc2", "audio_tower.layers.11.self_attn.k_proj", "audio_tower.layers.11.self_attn.v_proj", "audio_tower.layers.11.self_attn.q_proj", "audio_tower.layers.11.self_attn.out_proj", "audio_tower.layers.11.fc1", "audio_tower.layers.11.fc2", "audio_tower.layers.12.self_attn.k_proj", "audio_tower.layers.12.self_attn.v_proj", "audio_tower.layers.12.self_attn.q_proj", "audio_tower.layers.12.self_attn.out_proj", "audio_tower.layers.12.fc1", "audio_tower.layers.12.fc2", "audio_tower.layers.13.self_attn.k_proj", "audio_tower.layers.13.self_attn.v_proj", "audio_tower.layers.13.self_attn.q_proj", "audio_tower.layers.13.self_attn.out_proj", "audio_tower.layers.13.fc1", "audio_tower.layers.13.fc2", "audio_tower.layers.14.self_attn.k_proj", "audio_tower.layers.14.self_attn.v_proj", "audio_tower.layers.14.self_attn.q_proj", "audio_tower.layers.14.self_attn.out_proj", "audio_tower.layers.14.fc1", "audio_tower.layers.14.fc2", "audio_tower.layers.15.self_attn.k_proj", "audio_tower.layers.15.self_attn.v_proj", "audio_tower.layers.15.self_attn.q_proj", "audio_tower.layers.15.self_attn.out_proj", "audio_tower.layers.15.fc1", "audio_tower.layers.15.fc2", "audio_tower.layers.16.self_attn.k_proj", "audio_tower.layers.16.self_attn.v_proj", "audio_tower.layers.16.self_attn.q_proj", "audio_tower.layers.16.self_attn.out_proj", "audio_tower.layers.16.fc1", "audio_tower.layers.16.fc2", "audio_tower.layers.17.self_attn.k_proj", "audio_tower.layers.17.self_attn.v_proj", "audio_tower.layers.17.self_attn.q_proj", "audio_tower.layers.17.self_attn.out_proj", "audio_tower.layers.17.fc1", "audio_tower.layers.17.fc2", "audio_tower.layers.18.self_attn.k_proj", "audio_tower.layers.18.self_attn.v_proj", "audio_tower.layers.18.self_attn.q_proj", "audio_tower.layers.18.self_attn.out_proj", "audio_tower.layers.18.fc1", "audio_tower.layers.18.fc2", "audio_tower.layers.19.self_attn.k_proj", "audio_tower.layers.19.self_attn.v_proj", "audio_tower.layers.19.self_attn.q_proj", "audio_tower.layers.19.self_attn.out_proj", "audio_tower.layers.19.fc1", "audio_tower.layers.19.fc2", "audio_tower.layers.20.self_attn.k_proj", "audio_tower.layers.20.self_attn.v_proj", "audio_tower.layers.20.self_attn.q_proj", "audio_tower.layers.20.self_attn.out_proj", "audio_tower.layers.20.fc1", "audio_tower.layers.20.fc2", "audio_tower.layers.21.self_attn.k_proj", "audio_tower.layers.21.self_attn.v_proj", "audio_tower.layers.21.self_attn.q_proj", "audio_tower.layers.21.self_attn.out_proj", "audio_tower.layers.21.fc1", "audio_tower.layers.21.fc2", "audio_tower.layers.22.self_attn.k_proj", "audio_tower.layers.22.self_attn.v_proj", "audio_tower.layers.22.self_attn.q_proj", "audio_tower.layers.22.self_attn.out_proj", "audio_tower.layers.22.fc1", "audio_tower.layers.22.fc2", "audio_tower.layers.23.self_attn.k_proj", "audio_tower.layers.23.self_attn.v_proj", "audio_tower.layers.23.self_attn.q_proj", "audio_tower.layers.23.self_attn.out_proj", "audio_tower.layers.23.fc1", "audio_tower.layers.23.fc2", "audio_tower.layers.24.self_attn.k_proj", "audio_tower.layers.24.self_attn.v_proj", "audio_tower.layers.24.self_attn.q_proj", "audio_tower.layers.24.self_attn.out_proj", "audio_tower.layers.24.fc1", "audio_tower.layers.24.fc2", "audio_tower.layers.25.self_attn.k_proj", "audio_tower.layers.25.self_attn.v_proj", "audio_tower.layers.25.self_attn.q_proj", "audio_tower.layers.25.self_attn.out_proj", "audio_tower.layers.25.fc1", "audio_tower.layers.25.fc2", "audio_tower.layers.26.self_attn.k_proj", "audio_tower.layers.26.self_attn.v_proj", "audio_tower.layers.26.self_attn.q_proj", "audio_tower.layers.26.self_attn.out_proj", "audio_tower.layers.26.fc1", "audio_tower.layers.26.fc2", "audio_tower.layers.27.self_attn.k_proj", "audio_tower.layers.27.self_attn.v_proj", "audio_tower.layers.27.self_attn.q_proj", "audio_tower.layers.27.self_attn.out_proj", "audio_tower.layers.27.fc1", "audio_tower.layers.27.fc2", "audio_tower.layers.28.self_attn.k_proj", "audio_tower.layers.28.self_attn.v_proj", "audio_tower.layers.28.self_attn.q_proj", "audio_tower.layers.28.self_attn.out_proj", "audio_tower.layers.28.fc1", "audio_tower.layers.28.fc2", "audio_tower.layers.29.self_attn.k_proj", "audio_tower.layers.29.self_attn.v_proj", "audio_tower.layers.29.self_attn.q_proj", "audio_tower.layers.29.self_attn.out_proj", "audio_tower.layers.29.fc1", "audio_tower.layers.29.fc2", "audio_tower.layers.30.self_attn.k_proj", "audio_tower.layers.30.self_attn.v_proj", "audio_tower.layers.30.self_attn.q_proj", "audio_tower.layers.30.self_attn.out_proj", "audio_tower.layers.30.fc1", "audio_tower.layers.30.fc2", "audio_tower.layers.31.self_attn.k_proj", "audio_tower.layers.31.self_attn.v_proj", "audio_tower.layers.31.self_attn.q_proj", "audio_tower.layers.31.self_attn.out_proj", "audio_tower.layers.31.fc1", "audio_tower.layers.31.fc2", "language_model.lm_head", "multi_modal_projector.linear_1", "multi_modal_projector.linear_2" ], "kv_cache_scheme": null, "quant_method": "compressed-tensors", "quantization_status": "compressed" }, "text_config": { "attention_bias": false, "attention_dropout": 0.0, "head_dim": 128, "hidden_act": "silu", "hidden_size": 3072, "initializer_range": 0.02, "intermediate_size": 8192, "max_position_embeddings": 131072, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 32, "num_hidden_layers": 30, "num_key_value_heads": 8, "pretraining_tp": 1, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 100000000.0, "sliding_window": null, "torch_dtype": "bfloat16", "use_cache": true, "vocab_size": 131072 }, "torch_dtype": "bfloat16", "transformers_version": "4.54.0", "vocab_size": 131072 }