Muse-Spark-Open-DSpark / config.json
ZixiQi's picture
Upload verified Muse Spark DSpark step-900 checkpoint
1018a20 verified
Raw History Blame Contribute Delete
2.44 kB
{
"attention_bias": false,
"attention_dropout": 0.0,
"bos_token_id": 200000,
"draft_position_mode": "rope",
"eos_token_id": [
200001,
200008
],
"ffn_dim_multiplier": 2.0,
"head_dim": 64,
"hidden_act": "silu",
"hidden_size": 8192,
"initializer_range": 0.02,
"intermediate_size": 16384,
"intermediate_size_mlp": 16384,
"layer_types": [
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention"
],
"max_position_embeddings": 1048576,
"metap": {
"base_width": 256.0,
"m_emb": 1.0,
"metap_mode": "sp",
"router_has_one_inf_dim": true,
"use_metap": true,
"use_metap_output_multiplier_without_metap": false
},
"model_type": "muse_spark_dspark",
"non_linearity": "swiglu",
"norm_affine": true,
"norm_zero_centered_gamma": true,
"num_attention_heads": 128,
"num_hidden_layers": 5,
"num_key_value_heads": 16,
"output_norm_gain_center_type": "zero",
"output_soft_cap_temp": 20.0,
"pad_token_id": 200018,
"post_norm_affine": false,
"post_norm_gain_center_type": "one",
"residual_dtype": "fp32",
"rms_norm_eps": 1e-05,
"rms_post_norm_eps": 1e-08,
"rope_theta": 500000.0,
"rope_use_fp32_in_outer_product": true,
"sliding_window": 2048,
"tie_word_embeddings": false,
"use_attn_o_gate": true,
"use_attn_o_norm": true,
"use_ffn_learnable_scales": true,
"use_qk_norm": true,
"use_residual_gate": true,
"use_rope": true,
"use_sliding_window": true,
"use_sqrt_attn_scaling": false,
"vocab_size": 202048,
"architectures": [
"MuseSparkDSparkModel"
],
"moe": {
"is_enabled": false
},
"global_attn_cfg": [
2048
],
"every_n_layers_nope": null,
"num_target_layers": 5,
"target_hidden_size": 8192,
"target_num_hidden_layers": 62,
"target_layer_ids": [
13,
29,
45,
57,
61
],
"draft_vocab_size": 202048,
"mask_token_id": 200018,
"markov_rank": 512,
"markov_head_type": "vanilla",
"enable_confidence_head": true,
"confidence_head_with_markov": true,
"sample_from_anchor": true,
"block_size": 7,
"sliding_window_non_causal": false,
"full_attention_causal": false,
"use_aux_hidden_state": true,
"dflash_config": {
"causal": true,
"mask_token_id": 200018,
"target_layer_ids": [
13,
29,
45,
57,
61
],
"use_aux_hidden_state": true
},
"dtype": "bfloat16"
}