MirecX's picture
L3H5 v12 step_026000: m_accept 1.38, k=1 63.8%, 2211-shard training
f9f4303 verified
{
"architectures": [
"DFlashDraftModel"
],
"attention_bias": false,
"attention_dropout": 0.0,
"auto_map": {
"AutoModel": "dflash.DFlashDraftModel"
},
"block_size": 16,
"bos_token_id": 200034,
"dflash_config": {
"mask_token_id": 200063,
"target_layer_ids": [
2,
16,
30,
43,
57
]
},
"dtype": "bfloat16",
"eos_token_id": 200033,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 3072,
"initializer_range": 0.02,
"intermediate_size": 9216,
"layer_types": [
"full_attention",
"full_attention",
"full_attention"
],
"max_position_embeddings": 196608,
"max_window_layers": 3,
"model_type": "qwen3",
"num_attention_heads": 24,
"num_hidden_layers": 3,
"num_key_value_heads": 8,
"num_target_layers": 62,
"pad_token_id": 248044,
"rms_norm_eps": 1e-06,
"rope_parameters": {
"rope_theta": 10000000,
"rope_type": "default"
},
"sliding_window": null,
"tie_word_embeddings": false,
"transformers_version": "5.6.2",
"use_cache": false,
"use_sliding_window": false,
"vocab_size": 200064
}