Muse-Glimmer-30B text decoder -> Core AI: decode int4hu (--head-sym), token-exact vs fp16 oracle, 26.69 tok/s on M4 Max
d284183 verified | { | |
| "architectures": [ | |
| "MuseGlimmerForConditionalGeneration" | |
| ], | |
| "dtype": "bfloat16", | |
| "image_token_id": 200092, | |
| "model_type": "muse_glimmer", | |
| "out_hidden_size": 6144, | |
| "projector_hidden_act": "gelu", | |
| "projector_hidden_size": 4096, | |
| "text_config": { | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 200000, | |
| "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.dev0", | |
| "video_token_id": 200091, | |
| "vision_config": { | |
| "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" | |
| } | |
| } | |
| } | |