Hugging Face's logo

inferencerlabs
/
Qwen3.8-Flash-Next-MLX-Q9

Image-Text-to-Text
MLX
English
qwen4_exp
quantized
conversational
Model card Files
xet
 Community
1
inferencerlabs commited on
Commit
e0e70ef
·
verified ·
1 Parent(s): da9fdb8

Upload model file

Browse files
Files changed (1) hide show
  1. config.json +178 -0
config.json ADDED
@@ -0,0 +1,178 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen4ExpForConditionalGeneration"
4
+ ],
5
+ "audio_config": {},
6
+ "do_sample": true,
7
+ "eos_token_id": [
8
+ 248046,
9
+ 248044
10
+ ],
11
+ "generation_config": {
12
+ "bos_token_id": 248044,
13
+ "do_sample": true,
14
+ "eos_token_id": [
15
+ 248046,
16
+ 248044
17
+ ],
18
+ "pad_token_id": 248044,
19
+ "temperature": 1.0,
20
+ "top_k": 20,
21
+ "top_p": 0.95
22
+ },
23
+ "image_token_id": 248056,
24
+ "language_model_only": false,
25
+ "model_type": "qwen4_exp",
26
+ "quantization_config": {
27
+ "group_size": 32
28
+ },
29
+ "temperature": 1.0,
30
+ "text_config": {
31
+ "attention_bias": false,
32
+ "attention_dropout": 0.0,
33
+ "bos_token_id": 248044,
34
+ "dtype": "bfloat16",
35
+ "eos_token_id": 248044,
36
+ "full_attention_interval": 4,
37
+ "hc_count": 4,
38
+ "hc_lowrank": 320,
39
+ "head_dim": 256,
40
+ "heads_per_ngram": 8,
41
+ "hidden_act": "silu",
42
+ "hidden_size": 2560,
43
+ "indexer_budget": 2048,
44
+ "indexer_compress_ratio": 4,
45
+ "indexer_head_dim": 128,
46
+ "indexer_kv_heads": 1,
47
+ "indexer_n_heads": 4,
48
+ "initializer_range": 0.02,
49
+ "layer_types": [
50
+ "linear_attention",
51
+ "linear_attention",
52
+ "linear_attention",
53
+ "full_attention",
54
+ "linear_attention",
55
+ "linear_attention",
56
+ "linear_attention",
57
+ "full_attention",
58
+ "linear_attention",
59
+ "linear_attention",
60
+ "linear_attention",
61
+ "full_attention",
62
+ "linear_attention",
63
+ "linear_attention",
64
+ "linear_attention",
65
+ "full_attention",
66
+ "linear_attention",
67
+ "linear_attention",
68
+ "linear_attention",
69
+ "full_attention",
70
+ "linear_attention",
71
+ "linear_attention",
72
+ "linear_attention",
73
+ "full_attention",
74
+ "linear_attention",
75
+ "linear_attention",
76
+ "linear_attention",
77
+ "full_attention",
78
+ "linear_attention",
79
+ "linear_attention",
80
+ "linear_attention",
81
+ "full_attention",
82
+ "linear_attention",
83
+ "linear_attention",
84
+ "linear_attention",
85
+ "full_attention",
86
+ "linear_attention",
87
+ "linear_attention",
88
+ "linear_attention",
89
+ "full_attention",
90
+ "linear_attention",
91
+ "linear_attention",
92
+ "linear_attention",
93
+ "full_attention",
94
+ "linear_attention",
95
+ "linear_attention",
96
+ "linear_attention",
97
+ "full_attention"
98
+ ],
99
+ "linear_conv_kernel_dim": 4,
100
+ "linear_key_head_dim": 128,
101
+ "linear_num_key_heads": 16,
102
+ "linear_num_value_heads": 48,
103
+ "linear_value_head_dim": 128,
104
+ "make_ngram_vocab_size_divisible_by": 128,
105
+ "mamba_ssm_dtype": "float32",
106
+ "max_position_embeddings": 262144,
107
+ "model_type": "qwen4_exp_text",
108
+ "moe_intermediate_size": 640,
109
+ "mtp": {
110
+ "hybrid": true,
111
+ "layer_types": [
112
+ "full_attention"
113
+ ],
114
+ "mtp_use_hidden_state_from_layer": null,
115
+ "num_hidden_layers": 1,
116
+ "rope_theta": 10000000
117
+ },
118
+ "mtp_num_hidden_layers": 1,
119
+ "mtp_use_dedicated_embeddings": false,
120
+ "ngram_size": 3,
121
+ "ngram_vocab_size_base": 20000000,
122
+ "num_attention_heads": 24,
123
+ "num_experts": 512,
124
+ "num_experts_per_tok": 10,
125
+ "num_hidden_layers": 48,
126
+ "num_key_value_heads": 2,
127
+ "output_gate_type": "sigmoid",
128
+ "output_router_logits": false,
129
+ "pad_token_id": null,
130
+ "partial_rotary_factor": 0.25,
131
+ "ple_conv_kernel_size": 4,
132
+ "ple_embed_dim": 2560,
133
+ "ple_layer_ids": [
134
+ 2
135
+ ],
136
+ "rms_norm_eps": 1e-06,
137
+ "rope_parameters": {
138
+ "mrope_interleaved": true,
139
+ "mrope_section": [
140
+ 11,
141
+ 11,
142
+ 10
143
+ ],
144
+ "partial_rotary_factor": 0.25,
145
+ "rope_theta": 10000000,
146
+ "type": "default"
147
+ },
148
+ "router_aux_loss_coef": 0.001,
149
+ "shared_expert_intermediate_size": 640,
150
+ "split_ngram_parts": 128,
151
+ "tie_word_embeddings": false,
152
+ "use_cache": true,
153
+ "vocab_size": 248320
154
+ },
155
+ "tie_word_embeddings": false,
156
+ "top_k": 20,
157
+ "top_p": 0.95,
158
+ "transformers_version": "5.8.0.dev0",
159
+ "video_token_id": 248057,
160
+ "vision_config": {
161
+ "deepstack_visual_indexes": [],
162
+ "depth": 27,
163
+ "hidden_act": "gelu_pytorch_tanh",
164
+ "hidden_size": 1152,
165
+ "in_channels": 3,
166
+ "initializer_range": 0.02,
167
+ "intermediate_size": 4304,
168
+ "model_type": "qwen4_exp",
169
+ "num_heads": 16,
170
+ "num_position_embeddings": 2304,
171
+ "out_hidden_size": 2560,
172
+ "patch_size": 16,
173
+ "spatial_merge_size": 2,
174
+ "temporal_patch_size": 2
175
+ },
176
+ "vision_end_token_id": 248054,
177
+ "vision_start_token_id": 248053
178
+ }