{ "architectures": [ "MuseGlimmerForConditionalGeneration" ], "dtype": "bfloat16", "image_token_id": 200092, "model_type": "muse_glimmer", "out_hidden_size": 6144, "projector_hidden_act": "gelu", "projector_hidden_size": 4096, "quantization_config": { "config_groups": { "group_0": { "format": "pack-quantized", "input_activations": null, "output_activations": null, "targets": [ "Linear" ], "weights": { "actorder": "static", "block_structure": null, "dynamic": false, "group_size": 128, "num_bits": 4, "observer": "memoryless_minmax", "observer_kwargs": {}, "scale_dtype": null, "strategy": "group", "symmetric": true, "type": "int", "zp_dtype": null } } }, "format": "pack-quantized", "global_compression_ratio": null, "ignore": [ "model.vision_tower.patch_embedder.patch_embedding", "model.vision_tower.layers.0.attn.proj", "model.vision_tower.layers.0.attn.q_proj", "model.vision_tower.layers.0.attn.k_proj", "model.vision_tower.layers.0.attn.v_proj", "model.vision_tower.layers.0.mlp.fc1", "model.vision_tower.layers.0.mlp.fc2", "model.vision_tower.layers.1.attn.proj", "model.vision_tower.layers.1.attn.q_proj", "model.vision_tower.layers.1.attn.k_proj", "model.vision_tower.layers.1.attn.v_proj", "model.vision_tower.layers.1.mlp.fc1", "model.vision_tower.layers.1.mlp.fc2", "model.vision_tower.layers.2.attn.proj", "model.vision_tower.layers.2.attn.q_proj", "model.vision_tower.layers.2.attn.k_proj", "model.vision_tower.layers.2.attn.v_proj", "model.vision_tower.layers.2.mlp.fc1", "model.vision_tower.layers.2.mlp.fc2", "model.vision_tower.layers.3.attn.proj", "model.vision_tower.layers.3.attn.q_proj", "model.vision_tower.layers.3.attn.k_proj", "model.vision_tower.layers.3.attn.v_proj", "model.vision_tower.layers.3.mlp.fc1", "model.vision_tower.layers.3.mlp.fc2", "model.vision_tower.layers.4.attn.proj", "model.vision_tower.layers.4.attn.q_proj", "model.vision_tower.layers.4.attn.k_proj", "model.vision_tower.layers.4.attn.v_proj", "model.vision_tower.layers.4.mlp.fc1", "model.vision_tower.layers.4.mlp.fc2", "model.vision_tower.layers.5.attn.proj", "model.vision_tower.layers.5.attn.q_proj", "model.vision_tower.layers.5.attn.k_proj", "model.vision_tower.layers.5.attn.v_proj", "model.vision_tower.layers.5.mlp.fc1", "model.vision_tower.layers.5.mlp.fc2", "model.vision_tower.layers.6.attn.proj", "model.vision_tower.layers.6.attn.q_proj", "model.vision_tower.layers.6.attn.k_proj", "model.vision_tower.layers.6.attn.v_proj", "model.vision_tower.layers.6.mlp.fc1", "model.vision_tower.layers.6.mlp.fc2", "model.vision_tower.layers.7.attn.proj", "model.vision_tower.layers.7.attn.q_proj", "model.vision_tower.layers.7.attn.k_proj", "model.vision_tower.layers.7.attn.v_proj", "model.vision_tower.layers.7.mlp.fc1", "model.vision_tower.layers.7.mlp.fc2", "model.vision_tower.layers.8.attn.proj", "model.vision_tower.layers.8.attn.q_proj", "model.vision_tower.layers.8.attn.k_proj", "model.vision_tower.layers.8.attn.v_proj", "model.vision_tower.layers.8.mlp.fc1", "model.vision_tower.layers.8.mlp.fc2", "model.vision_tower.layers.9.attn.proj", "model.vision_tower.layers.9.attn.q_proj", "model.vision_tower.layers.9.attn.k_proj", "model.vision_tower.layers.9.attn.v_proj", "model.vision_tower.layers.9.mlp.fc1", "model.vision_tower.layers.9.mlp.fc2", "model.vision_tower.layers.10.attn.proj", "model.vision_tower.layers.10.attn.q_proj", "model.vision_tower.layers.10.attn.k_proj", "model.vision_tower.layers.10.attn.v_proj", "model.vision_tower.layers.10.mlp.fc1", "model.vision_tower.layers.10.mlp.fc2", "model.vision_tower.layers.11.attn.proj", "model.vision_tower.layers.11.attn.q_proj", "model.vision_tower.layers.11.attn.k_proj", "model.vision_tower.layers.11.attn.v_proj", "model.vision_tower.layers.11.mlp.fc1", "model.vision_tower.layers.11.mlp.fc2", "model.vision_tower.layers.12.attn.proj", "model.vision_tower.layers.12.attn.q_proj", "model.vision_tower.layers.12.attn.k_proj", "model.vision_tower.layers.12.attn.v_proj", "model.vision_tower.layers.12.mlp.fc1", "model.vision_tower.layers.12.mlp.fc2", "model.vision_tower.layers.13.attn.proj", "model.vision_tower.layers.13.attn.q_proj", "model.vision_tower.layers.13.attn.k_proj", "model.vision_tower.layers.13.attn.v_proj", "model.vision_tower.layers.13.mlp.fc1", "model.vision_tower.layers.13.mlp.fc2", "model.vision_tower.layers.14.attn.proj", "model.vision_tower.layers.14.attn.q_proj", "model.vision_tower.layers.14.attn.k_proj", "model.vision_tower.layers.14.attn.v_proj", "model.vision_tower.layers.14.mlp.fc1", "model.vision_tower.layers.14.mlp.fc2", "model.vision_tower.layers.15.attn.proj", "model.vision_tower.layers.15.attn.q_proj", "model.vision_tower.layers.15.attn.k_proj", "model.vision_tower.layers.15.attn.v_proj", "model.vision_tower.layers.15.mlp.fc1", "model.vision_tower.layers.15.mlp.fc2", "model.vision_tower.layers.16.attn.proj", "model.vision_tower.layers.16.attn.q_proj", "model.vision_tower.layers.16.attn.k_proj", "model.vision_tower.layers.16.attn.v_proj", "model.vision_tower.layers.16.mlp.fc1", "model.vision_tower.layers.16.mlp.fc2", "model.vision_tower.layers.17.attn.proj", "model.vision_tower.layers.17.attn.q_proj", "model.vision_tower.layers.17.attn.k_proj", "model.vision_tower.layers.17.attn.v_proj", "model.vision_tower.layers.17.mlp.fc1", "model.vision_tower.layers.17.mlp.fc2", "model.vision_tower.layers.18.attn.proj", "model.vision_tower.layers.18.attn.q_proj", "model.vision_tower.layers.18.attn.k_proj", "model.vision_tower.layers.18.attn.v_proj", "model.vision_tower.layers.18.mlp.fc1", "model.vision_tower.layers.18.mlp.fc2", "model.vision_tower.layers.19.attn.proj", "model.vision_tower.layers.19.attn.q_proj", "model.vision_tower.layers.19.attn.k_proj", "model.vision_tower.layers.19.attn.v_proj", "model.vision_tower.layers.19.mlp.fc1", "model.vision_tower.layers.19.mlp.fc2", "model.vision_tower.layers.20.attn.proj", "model.vision_tower.layers.20.attn.q_proj", "model.vision_tower.layers.20.attn.k_proj", "model.vision_tower.layers.20.attn.v_proj", "model.vision_tower.layers.20.mlp.fc1", "model.vision_tower.layers.20.mlp.fc2", "model.vision_tower.layers.21.attn.proj", "model.vision_tower.layers.21.attn.q_proj", "model.vision_tower.layers.21.attn.k_proj", "model.vision_tower.layers.21.attn.v_proj", "model.vision_tower.layers.21.mlp.fc1", "model.vision_tower.layers.21.mlp.fc2", "model.vision_tower.layers.22.attn.proj", "model.vision_tower.layers.22.attn.q_proj", "model.vision_tower.layers.22.attn.k_proj", "model.vision_tower.layers.22.attn.v_proj", "model.vision_tower.layers.22.mlp.fc1", "model.vision_tower.layers.22.mlp.fc2", "model.vision_tower.layers.23.attn.proj", "model.vision_tower.layers.23.attn.q_proj", "model.vision_tower.layers.23.attn.k_proj", "model.vision_tower.layers.23.attn.v_proj", "model.vision_tower.layers.23.mlp.fc1", "model.vision_tower.layers.23.mlp.fc2", "model.vision_tower.layers.24.attn.proj", "model.vision_tower.layers.24.attn.q_proj", "model.vision_tower.layers.24.attn.k_proj", "model.vision_tower.layers.24.attn.v_proj", "model.vision_tower.layers.24.mlp.fc1", "model.vision_tower.layers.24.mlp.fc2", "model.vision_tower.layers.25.attn.proj", "model.vision_tower.layers.25.attn.q_proj", "model.vision_tower.layers.25.attn.k_proj", "model.vision_tower.layers.25.attn.v_proj", "model.vision_tower.layers.25.mlp.fc1", "model.vision_tower.layers.25.mlp.fc2", "model.vision_tower.layers.26.attn.proj", "model.vision_tower.layers.26.attn.q_proj", "model.vision_tower.layers.26.attn.k_proj", "model.vision_tower.layers.26.attn.v_proj", "model.vision_tower.layers.26.mlp.fc1", "model.vision_tower.layers.26.mlp.fc2", "model.vision_tower.layers.27.attn.proj", "model.vision_tower.layers.27.attn.q_proj", "model.vision_tower.layers.27.attn.k_proj", "model.vision_tower.layers.27.attn.v_proj", "model.vision_tower.layers.27.mlp.fc1", "model.vision_tower.layers.27.mlp.fc2", "model.vision_tower.layers.28.attn.proj", "model.vision_tower.layers.28.attn.q_proj", "model.vision_tower.layers.28.attn.k_proj", "model.vision_tower.layers.28.attn.v_proj", "model.vision_tower.layers.28.mlp.fc1", "model.vision_tower.layers.28.mlp.fc2", "model.vision_tower.layers.29.attn.proj", "model.vision_tower.layers.29.attn.q_proj", "model.vision_tower.layers.29.attn.k_proj", "model.vision_tower.layers.29.attn.v_proj", "model.vision_tower.layers.29.mlp.fc1", "model.vision_tower.layers.29.mlp.fc2", "model.vision_tower.layers.30.attn.proj", "model.vision_tower.layers.30.attn.q_proj", "model.vision_tower.layers.30.attn.k_proj", "model.vision_tower.layers.30.attn.v_proj", "model.vision_tower.layers.30.mlp.fc1", "model.vision_tower.layers.30.mlp.fc2", "model.vision_tower.layers.31.attn.proj", "model.vision_tower.layers.31.attn.q_proj", "model.vision_tower.layers.31.attn.k_proj", "model.vision_tower.layers.31.attn.v_proj", "model.vision_tower.layers.31.mlp.fc1", "model.vision_tower.layers.31.mlp.fc2", "model.vision_tower.layers.32.attn.proj", "model.vision_tower.layers.32.attn.q_proj", "model.vision_tower.layers.32.attn.k_proj", "model.vision_tower.layers.32.attn.v_proj", "model.vision_tower.layers.32.mlp.fc1", "model.vision_tower.layers.32.mlp.fc2", "model.vision_tower.layers.33.attn.proj", "model.vision_tower.layers.33.attn.q_proj", "model.vision_tower.layers.33.attn.k_proj", "model.vision_tower.layers.33.attn.v_proj", "model.vision_tower.layers.33.mlp.fc1", "model.vision_tower.layers.33.mlp.fc2", "model.vision_tower.layers.34.attn.proj", "model.vision_tower.layers.34.attn.q_proj", "model.vision_tower.layers.34.attn.k_proj", "model.vision_tower.layers.34.attn.v_proj", "model.vision_tower.layers.34.mlp.fc1", "model.vision_tower.layers.34.mlp.fc2", "model.vision_tower.layers.35.attn.proj", "model.vision_tower.layers.35.attn.q_proj", "model.vision_tower.layers.35.attn.k_proj", "model.vision_tower.layers.35.attn.v_proj", "model.vision_tower.layers.35.mlp.fc1", "model.vision_tower.layers.35.mlp.fc2", "model.vision_tower.layers.36.attn.proj", "model.vision_tower.layers.36.attn.q_proj", "model.vision_tower.layers.36.attn.k_proj", "model.vision_tower.layers.36.attn.v_proj", "model.vision_tower.layers.36.mlp.fc1", "model.vision_tower.layers.36.mlp.fc2", "model.vision_tower.layers.37.attn.proj", "model.vision_tower.layers.37.attn.q_proj", "model.vision_tower.layers.37.attn.k_proj", "model.vision_tower.layers.37.attn.v_proj", "model.vision_tower.layers.37.mlp.fc1", "model.vision_tower.layers.37.mlp.fc2", "model.vision_tower.layers.38.attn.proj", "model.vision_tower.layers.38.attn.q_proj", "model.vision_tower.layers.38.attn.k_proj", "model.vision_tower.layers.38.attn.v_proj", "model.vision_tower.layers.38.mlp.fc1", "model.vision_tower.layers.38.mlp.fc2", "model.vision_tower.layers.39.attn.proj", "model.vision_tower.layers.39.attn.q_proj", "model.vision_tower.layers.39.attn.k_proj", "model.vision_tower.layers.39.attn.v_proj", "model.vision_tower.layers.39.mlp.fc1", "model.vision_tower.layers.39.mlp.fc2", "model.vision_tower.layers.40.attn.proj", "model.vision_tower.layers.40.attn.q_proj", "model.vision_tower.layers.40.attn.k_proj", "model.vision_tower.layers.40.attn.v_proj", "model.vision_tower.layers.40.mlp.fc1", "model.vision_tower.layers.40.mlp.fc2", "model.vision_tower.layers.41.attn.proj", "model.vision_tower.layers.41.attn.q_proj", "model.vision_tower.layers.41.attn.k_proj", "model.vision_tower.layers.41.attn.v_proj", "model.vision_tower.layers.41.mlp.fc1", "model.vision_tower.layers.41.mlp.fc2", "model.vision_tower.layers.42.attn.proj", "model.vision_tower.layers.42.attn.q_proj", "model.vision_tower.layers.42.attn.k_proj", "model.vision_tower.layers.42.attn.v_proj", "model.vision_tower.layers.42.mlp.fc1", "model.vision_tower.layers.42.mlp.fc2", "model.vision_tower.layers.43.attn.proj", "model.vision_tower.layers.43.attn.q_proj", "model.vision_tower.layers.43.attn.k_proj", "model.vision_tower.layers.43.attn.v_proj", "model.vision_tower.layers.43.mlp.fc1", "model.vision_tower.layers.43.mlp.fc2", "model.vision_tower.layers.44.attn.proj", "model.vision_tower.layers.44.attn.q_proj", "model.vision_tower.layers.44.attn.k_proj", "model.vision_tower.layers.44.attn.v_proj", "model.vision_tower.layers.44.mlp.fc1", "model.vision_tower.layers.44.mlp.fc2", "model.vision_tower.layers.45.attn.proj", "model.vision_tower.layers.45.attn.q_proj", "model.vision_tower.layers.45.attn.k_proj", "model.vision_tower.layers.45.attn.v_proj", "model.vision_tower.layers.45.mlp.fc1", "model.vision_tower.layers.45.mlp.fc2", "model.vision_tower.layers.46.attn.proj", "model.vision_tower.layers.46.attn.q_proj", "model.vision_tower.layers.46.attn.k_proj", "model.vision_tower.layers.46.attn.v_proj", "model.vision_tower.layers.46.mlp.fc1", "model.vision_tower.layers.46.mlp.fc2", "model.vision_tower.layers.47.attn.proj", "model.vision_tower.layers.47.attn.q_proj", "model.vision_tower.layers.47.attn.k_proj", "model.vision_tower.layers.47.attn.v_proj", "model.vision_tower.layers.47.mlp.fc1", "model.vision_tower.layers.47.mlp.fc2", "model.vision_tower.layers.48.attn.proj", "model.vision_tower.layers.48.attn.q_proj", "model.vision_tower.layers.48.attn.k_proj", "model.vision_tower.layers.48.attn.v_proj", "model.vision_tower.layers.48.mlp.fc1", "model.vision_tower.layers.48.mlp.fc2", "model.vision_tower.layers.49.attn.proj", "model.vision_tower.layers.49.attn.q_proj", "model.vision_tower.layers.49.attn.k_proj", "model.vision_tower.layers.49.attn.v_proj", "model.vision_tower.layers.49.mlp.fc1", "model.vision_tower.layers.49.mlp.fc2", "model.vision_adapter.fc1", "model.vision_adapter.fc2", "model.vision_projection", "lm_head" ], "kv_cache_scheme": null, "quant_method": "compressed-tensors", "quantization_status": "compressed", "sparsity_config": {}, "transform_config": {}, "version": "0.18.1.a20260806" }, "text_config": { "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 200000, "dtype": "bfloat16", "eos_token_id": 200001, "final_logit_softcapping": 20.0, "head_dim": 128, "hidden_activation": "silu", "hidden_size": 6656, "initializer_range": 0.02, "intermediate_size": 19968, "layer_rope_theta": [ 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0, 500000.0, 500000.0, 500000.0, 0 ], "layer_types": [ "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention", "sliding_attention", "sliding_attention", "sliding_attention", "full_attention" ], "max_position_embeddings": 131072, "model_type": "muse_glimmer_text", "num_attention_heads": 32, "num_hidden_layers": 52, "num_key_value_heads": 2, "output_multiplier": 0.19611613513818404, "pad_token_id": null, "post_norm_eps": 1e-08, "qk_scale_factor": 3.87, "rms_norm_eps": 1e-05, "rope_parameters": { "rope_theta": 500000.0, "rope_type": "default" }, "sliding_window": 2048, "tie_word_embeddings": false, "use_cache": true, "vocab_size": 202048 }, "transformers_version": "5.15.0", "video_token_id": 200091, "vision_config": { "dtype": "bfloat16", "hidden_act": "gelu", "hidden_size": 1536, "intermediate_size": 8960, "layer_norm_eps": 1e-05, "layer_types": [ "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "window_attention", "window_attention", "full_attention", "window_attention", "full_attention" ], "max_position_embeddings": 1024, "merge_size": 2, "model_type": "muse_glimmer_vision", "num_attention_heads": 16, "num_hidden_layers": 50, "patch_size": 14, "patch_temporal": 2, "pos_emb_height": 32, "pos_emb_width": 32, "rope_parameters": { "rope_theta": 10000.0, "rope_type": "default" } } }