Hugging Face's logo

FunAudioLLM
/
Fun-ASR-Nano-2512-hf

Automatic Speech Recognition
Transformers
Safetensors
fun_asr_nano
text-generation
speech-recognition
asr
end-to-end
multilingual
streaming
Model card Files
xet
 Community
2
langgz commited on
Commit
35bf550
·
verified ·
1 Parent(s): 2aa316d

Re-convert Fun-ASR-Nano to dtype-consistent bf16 (modular split layout); shrink 2.75GB->1.97GB

Browse files
Files changed (1) hide show
  1. config.json +55 -39
config.json CHANGED
@@ -1,33 +1,30 @@
1
  {
2
- "model_type": "fun_asr_nano",
 
 
 
 
3
  "architectures": [
4
  "FunAsrNanoForConditionalGeneration"
5
  ],
6
  "audio_encoder_config": {
7
- "model_type": "fun_asr_nano_encoder",
8
- "input_size": 560,
9
- "output_size": 512,
10
  "attention_heads": 4,
11
- "linear_units": 2048,
12
- "num_blocks": 50,
13
- "tp_blocks": 20,
14
  "dropout_rate": 0.1,
15
- "attention_dropout_rate": 0.0,
 
16
  "kernel_size": 11,
 
 
 
 
17
  "sanm_shift": 0,
18
- "initializer_range": 0.02
19
- },
20
- "adaptor_config": {
21
- "model_type": "fun_asr_nano_adaptor",
22
- "downsample_rate": 1,
23
- "encoder_dim": 512,
24
- "llm_dim": 1024,
25
- "ffn_dim": 2048,
26
- "num_layers": 2,
27
- "attention_heads": 8,
28
- "dropout_rate": 0.0,
29
- "use_low_frame_rate": true
30
  },
 
 
 
 
31
  "text_config": {
32
  "architectures": [
33
  "Qwen3ForCausalLM"
@@ -35,41 +32,60 @@
35
  "attention_bias": false,
36
  "attention_dropout": 0.0,
37
  "bos_token_id": 151643,
 
38
  "eos_token_id": 151645,
39
  "head_dim": 128,
40
  "hidden_act": "silu",
41
  "hidden_size": 1024,
42
  "initializer_range": 0.02,
43
  "intermediate_size": 3072,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
44
  "max_position_embeddings": 40960,
45
  "max_window_layers": 28,
46
  "model_type": "qwen3",
47
  "num_attention_heads": 16,
48
  "num_hidden_layers": 28,
49
  "num_key_value_heads": 8,
 
50
  "rms_norm_eps": 1e-06,
51
- "rope_scaling": null,
52
- "rope_theta": 1000000,
 
 
53
  "sliding_window": null,
54
  "tie_word_embeddings": true,
55
- "torch_dtype": "bfloat16",
56
- "transformers_version": "4.51.0",
57
  "use_cache": true,
58
  "use_sliding_window": false,
59
  "vocab_size": 151936
60
  },
61
- "ctc_config": {
62
- "model_type": "fun_asr_nano_ctc",
63
- "vocab_size": 60515,
64
- "encoder_dim": 512,
65
- "decoder_dim": 512,
66
- "ffn_dim": 2048,
67
- "num_layers": 5,
68
- "downsample_rate": 1,
69
- "blank_id": 60514,
70
- "dropout_rate": 0.0
71
- },
72
- "audio_token_index": 151646,
73
- "initializer_range": 0.02,
74
- "torch_dtype": "bfloat16"
75
- }
 
1
  {
2
+ "adaptor_attention_heads": 8,
3
+ "adaptor_downsample_rate": 1,
4
+ "adaptor_dropout_rate": 0.0,
5
+ "adaptor_ffn_dim": 2048,
6
+ "adaptor_num_layers": 2,
7
  "architectures": [
8
  "FunAsrNanoForConditionalGeneration"
9
  ],
10
  "audio_encoder_config": {
11
+ "attention_dropout_rate": 0.1,
 
 
12
  "attention_heads": 4,
 
 
 
13
  "dropout_rate": 0.1,
14
+ "initializer_range": 0.02,
15
+ "input_size": 560,
16
  "kernel_size": 11,
17
+ "linear_units": 2048,
18
+ "model_type": "fun_asr_nano_encoder",
19
+ "num_blocks": 50,
20
+ "output_size": 512,
21
  "sanm_shift": 0,
22
+ "tp_blocks": 20
 
 
 
 
 
 
 
 
 
 
 
23
  },
24
+ "audio_token_index": 151646,
25
+ "dtype": "bfloat16",
26
+ "initializer_range": 0.02,
27
+ "model_type": "fun_asr_nano",
28
  "text_config": {
29
  "architectures": [
30
  "Qwen3ForCausalLM"
 
32
  "attention_bias": false,
33
  "attention_dropout": 0.0,
34
  "bos_token_id": 151643,
35
+ "dtype": "bfloat16",
36
  "eos_token_id": 151645,
37
  "head_dim": 128,
38
  "hidden_act": "silu",
39
  "hidden_size": 1024,
40
  "initializer_range": 0.02,
41
  "intermediate_size": 3072,
42
+ "layer_types": [
43
+ "full_attention",
44
+ "full_attention",
45
+ "full_attention",
46
+ "full_attention",
47
+ "full_attention",
48
+ "full_attention",
49
+ "full_attention",
50
+ "full_attention",
51
+ "full_attention",
52
+ "full_attention",
53
+ "full_attention",
54
+ "full_attention",
55
+ "full_attention",
56
+ "full_attention",
57
+ "full_attention",
58
+ "full_attention",
59
+ "full_attention",
60
+ "full_attention",
61
+ "full_attention",
62
+ "full_attention",
63
+ "full_attention",
64
+ "full_attention",
65
+ "full_attention",
66
+ "full_attention",
67
+ "full_attention",
68
+ "full_attention",
69
+ "full_attention",
70
+ "full_attention"
71
+ ],
72
  "max_position_embeddings": 40960,
73
  "max_window_layers": 28,
74
  "model_type": "qwen3",
75
  "num_attention_heads": 16,
76
  "num_hidden_layers": 28,
77
  "num_key_value_heads": 8,
78
+ "pad_token_id": null,
79
  "rms_norm_eps": 1e-06,
80
+ "rope_parameters": {
81
+ "rope_theta": 1000000,
82
+ "rope_type": "default"
83
+ },
84
  "sliding_window": null,
85
  "tie_word_embeddings": true,
 
 
86
  "use_cache": true,
87
  "use_sliding_window": false,
88
  "vocab_size": 151936
89
  },
90
+ "transformers_version": "5.13.0.dev0"
91
+ }