Hugging Face's logo

bacchist
/
pokerbench_Qwen3-4B-unsloth-bnb-4bit

Text Generation
Transformers
Safetensors
English
qwen3
text-generation-inference
unsloth
conversational
4-bit precision
bitsandbytes
Model card Files
xet
 Community
bacchist commited on
Commit
74537a3
·
verified ·
1 Parent(s): b6528a7

(Trained with Unsloth)

Browse files
Files changed (2) hide show
  1. config.json +124 -68
  2. model.safetensors +3 -0
config.json CHANGED
@@ -1,70 +1,126 @@
1
  {
2
- "architectures": [
3
- "Qwen3ForCausalLM"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4
  ],
5
- "attention_bias": false,
6
- "attention_dropout": 0.0,
7
- "eos_token_id": 151645,
8
- "head_dim": 128,
9
- "hidden_act": "silu",
10
- "hidden_size": 2560,
11
- "initializer_range": 0.02,
12
- "intermediate_size": 9728,
13
- "layer_types": [
14
- "full_attention",
15
- "full_attention",
16
- "full_attention",
17
- "full_attention",
18
- "full_attention",
19
- "full_attention",
20
- "full_attention",
21
- "full_attention",
22
- "full_attention",
23
- "full_attention",
24
- "full_attention",
25
- "full_attention",
26
- "full_attention",
27
- "full_attention",
28
- "full_attention",
29
- "full_attention",
30
- "full_attention",
31
- "full_attention",
32
- "full_attention",
33
- "full_attention",
34
- "full_attention",
35
- "full_attention",
36
- "full_attention",
37
- "full_attention",
38
- "full_attention",
39
- "full_attention",
40
- "full_attention",
41
- "full_attention",
42
- "full_attention",
43
- "full_attention",
44
- "full_attention",
45
- "full_attention",
46
- "full_attention",
47
- "full_attention",
48
- "full_attention",
49
- "full_attention"
50
- ],
51
- "max_position_embeddings": 40960,
52
- "max_window_layers": 36,
53
- "model_type": "qwen3",
54
- "num_attention_heads": 32,
55
- "num_hidden_layers": 36,
56
- "num_key_value_heads": 8,
57
- "pad_token_id": 151654,
58
- "rms_norm_eps": 1e-06,
59
- "rope_scaling": null,
60
- "rope_theta": 1000000,
61
- "sliding_window": null,
62
- "tie_word_embeddings": true,
63
- "torch_dtype": "float16",
64
- "transformers_version": "4.54.0",
65
- "unsloth_fixed": true,
66
- "unsloth_version": "2025.7.11",
67
- "use_cache": true,
68
- "use_sliding_window": false,
69
- "vocab_size": 151936
70
- }
 
1
  {
2
+ "architectures": [
3
+ "Qwen3ForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "eos_token_id": 151645,
8
+ "head_dim": 128,
9
+ "hidden_act": "silu",
10
+ "hidden_size": 2560,
11
+ "initializer_range": 0.02,
12
+ "intermediate_size": 9728,
13
+ "layer_types": [
14
+ "full_attention",
15
+ "full_attention",
16
+ "full_attention",
17
+ "full_attention",
18
+ "full_attention",
19
+ "full_attention",
20
+ "full_attention",
21
+ "full_attention",
22
+ "full_attention",
23
+ "full_attention",
24
+ "full_attention",
25
+ "full_attention",
26
+ "full_attention",
27
+ "full_attention",
28
+ "full_attention",
29
+ "full_attention",
30
+ "full_attention",
31
+ "full_attention",
32
+ "full_attention",
33
+ "full_attention",
34
+ "full_attention",
35
+ "full_attention",
36
+ "full_attention",
37
+ "full_attention",
38
+ "full_attention",
39
+ "full_attention",
40
+ "full_attention",
41
+ "full_attention",
42
+ "full_attention",
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention",
49
+ "full_attention"
50
+ ],
51
+ "max_position_embeddings": 40960,
52
+ "max_window_layers": 36,
53
+ "model_type": "qwen3",
54
+ "num_attention_heads": 32,
55
+ "num_hidden_layers": 36,
56
+ "num_key_value_heads": 8,
57
+ "pad_token_id": 151654,
58
+ "quantization_config": {
59
+ "bnb_4bit_compute_dtype": "float16",
60
+ "bnb_4bit_quant_type": "nf4",
61
+ "bnb_4bit_use_double_quant": true,
62
+ "llm_int8_enable_fp32_cpu_offload": false,
63
+ "llm_int8_has_fp16_weight": false,
64
+ "llm_int8_skip_modules": [
65
+ "model.layers.0.self_attn.q_proj",
66
+ "model.layers.0.self_attn.k_proj",
67
+ "model.layers.0.self_attn.v_proj",
68
+ "model.layers.0.self_attn.o_proj",
69
+ "model.layers.0.mlp.gate_proj",
70
+ "model.layers.0.mlp.up_proj",
71
+ "model.layers.0.mlp.down_proj",
72
+ "model.layers.1.self_attn.q_proj",
73
+ "model.layers.1.self_attn.k_proj",
74
+ "model.layers.1.self_attn.v_proj",
75
+ "model.layers.1.self_attn.o_proj",
76
+ "model.layers.1.mlp.gate_proj",
77
+ "model.layers.1.mlp.up_proj",
78
+ "model.layers.1.mlp.down_proj",
79
+ "model.layers.3.mlp.gate_proj",
80
+ "model.layers.3.mlp.up_proj",
81
+ "model.layers.3.mlp.down_proj",
82
+ "model.layers.4.self_attn.q_proj",
83
+ "model.layers.4.self_attn.k_proj",
84
+ "model.layers.4.self_attn.v_proj",
85
+ "model.layers.4.self_attn.o_proj",
86
+ "model.layers.4.mlp.gate_proj",
87
+ "model.layers.4.mlp.up_proj",
88
+ "model.layers.4.mlp.down_proj",
89
+ "model.layers.6.self_attn.q_proj",
90
+ "model.layers.6.self_attn.k_proj",
91
+ "model.layers.6.self_attn.v_proj",
92
+ "model.layers.6.self_attn.o_proj",
93
+ "model.layers.6.mlp.gate_proj",
94
+ "model.layers.6.mlp.up_proj",
95
+ "model.layers.6.mlp.down_proj",
96
+ "model.layers.33.self_attn.q_proj",
97
+ "model.layers.33.self_attn.k_proj",
98
+ "model.layers.33.self_attn.v_proj",
99
+ "model.layers.33.self_attn.o_proj",
100
+ "model.layers.34.self_attn.q_proj",
101
+ "model.layers.34.self_attn.k_proj",
102
+ "model.layers.34.self_attn.v_proj",
103
+ "model.layers.34.self_attn.o_proj",
104
+ "model.layers.34.mlp.gate_proj",
105
+ "model.layers.34.mlp.up_proj",
106
+ "model.layers.34.mlp.down_proj",
107
+ "lm_head"
108
  ],
109
+ "llm_int8_threshold": 6.0,
110
+ "load_in_4bit": true,
111
+ "load_in_8bit": false,
112
+ "quant_method": "bitsandbytes"
113
+ },
114
+ "rms_norm_eps": 1e-06,
115
+ "rope_scaling": null,
116
+ "rope_theta": 1000000,
117
+ "sliding_window": null,
118
+ "tie_word_embeddings": true,
119
+ "torch_dtype": "float16",
120
+ "transformers_version": "4.54.0",
121
+ "unsloth_fixed": true,
122
+ "unsloth_version": "2025.7.11",
123
+ "use_cache": true,
124
+ "use_sliding_window": false,
125
+ "vocab_size": 151936
126
+ }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad3cbe20d021a8ba54059f5744dcbfacfe28c633f00e55f874d9d9558b7bd711
3
+ size 3551775126