Text Generation
Transformers
Safetensors
qwen3_next
conversational
Entity-27th commited on
Commit
e46f2e8
·
1 Parent(s): 31402d0

Upload Qwen3NextForCausalLM

Browse files
Files changed (1) hide show
  1. config.json +63 -26
config.json CHANGED
@@ -1,41 +1,78 @@
1
  {
2
  "architectures": [
3
- "CustomTransformerForCausalLM"
4
  ],
 
5
  "attention_dropout": 0.0,
 
6
  "dtype": "float32",
7
- "enable_checkpointing": false,
8
  "gate_alpha": 0.8,
9
  "gate_normalize": true,
10
- "gate_temperature": 1.5,
11
- "gated_attention_min_tokens": 0.05,
12
- "gated_attention_reduction": 16,
13
- "gated_attention_threshold": 0.25,
14
- "gated_delta_net_bias": true,
15
- "gated_delta_net_min_tokens": 0.05,
16
- "gated_delta_net_reduction": 16,
17
- "gated_delta_net_threshold": 0.25,
18
- "hidden_size": 2560,
19
- "intermediate_size": 6144,
20
- "kv_cast_dtype": "bfloat16",
21
- "max_position_embeddings": 128000,
22
- "model_type": "ISAC-V0",
23
- "num_attention_heads": 40,
24
- "num_hidden_layers": 40,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25
  "num_key_value_heads": 10,
 
 
26
  "pr_chunk_size": 2048,
27
  "pr_degree": 2,
28
- "rms_norm_eps": 1e-05,
29
  "rope_scaling": {
30
- "factor": 31.25,
31
- "original_max_position_embeddings": 4096,
32
- "type": "ntk"
33
  },
34
  "rope_theta": 10000.0,
35
- "sliding_window": null,
 
 
36
  "transformers_version": "4.57.1",
37
- "use_cache": false,
38
- "use_gated_attention": true,
39
- "use_gated_delta_net": true,
40
- "vocab_size": 200019
41
  }
 
1
  {
2
  "architectures": [
3
+ "Qwen3NextForCausalLM"
4
  ],
5
+ "attention_bias": false,
6
  "attention_dropout": 0.0,
7
+ "decoder_sparse_step": 0,
8
  "dtype": "float32",
 
9
  "gate_alpha": 0.8,
10
  "gate_normalize": true,
11
+ "gate_temperature": 1.0,
12
+ "head_dim": 64,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 1280,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 3328,
17
+ "layer_types": [
18
+ "linear_attention",
19
+ "linear_attention",
20
+ "linear_attention",
21
+ "full_attention",
22
+ "linear_attention",
23
+ "linear_attention",
24
+ "linear_attention",
25
+ "full_attention",
26
+ "linear_attention",
27
+ "linear_attention",
28
+ "linear_attention",
29
+ "full_attention",
30
+ "linear_attention",
31
+ "linear_attention",
32
+ "linear_attention",
33
+ "full_attention",
34
+ "linear_attention",
35
+ "linear_attention",
36
+ "linear_attention",
37
+ "full_attention",
38
+ "linear_attention",
39
+ "linear_attention",
40
+ "linear_attention",
41
+ "full_attention",
42
+ "linear_attention",
43
+ "linear_attention"
44
+ ],
45
+ "linear_conv_kernel_dim": 4,
46
+ "linear_key_head_dim": 128,
47
+ "linear_num_key_heads": 16,
48
+ "linear_num_value_heads": 32,
49
+ "linear_value_head_dim": 128,
50
+ "max_position_embeddings": 1048576,
51
+ "mlp_only_layers": [],
52
+ "model_type": "qwen3_next",
53
+ "moe_intermediate_size": 0,
54
+ "norm_topk_prob": true,
55
+ "num_attention_heads": 20,
56
+ "num_experts": 0,
57
+ "num_experts_per_tok": 0,
58
+ "num_hidden_layers": 26,
59
  "num_key_value_heads": 10,
60
+ "output_router_logits": false,
61
+ "partial_rotary_factor": 0.25,
62
  "pr_chunk_size": 2048,
63
  "pr_degree": 2,
64
+ "rms_norm_eps": 1e-06,
65
  "rope_scaling": {
66
+ "factor": 32.0,
67
+ "original_max_position_embeddings": 32768,
68
+ "rope_type": "yarn"
69
  },
70
  "rope_theta": 10000.0,
71
+ "router_aux_loss_coef": 0.001,
72
+ "shared_expert_intermediate_size": 0,
73
+ "token_route_threshold": 0.2,
74
  "transformers_version": "4.57.1",
75
+ "use_cache": true,
76
+ "use_power_retention": true,
77
+ "vocab_size": 151936
 
78
  }