Hugging Face's logo

mazesmazes
/
tiny-audio

Automatic Speech Recognition
Transformers
Safetensors
English
asr_model
feature-extraction
asr
speech-recognition
audio
qwen
glm-asr
custom_code
Model card Files
xet
Community
mazesmazes commited on
Commit
cbe595b
·
verified ·
1 Parent(s): 69c7ff3

Training in progress, step 1000

Browse files
Files changed (4) hide show
  1. config.json +199 -31
  2. generation_config.json +7 -4
  3. model.safetensors +2 -2
  4. training_args.bin +2 -2
config.json CHANGED
@@ -1,38 +1,206 @@
1
  {
2
  "architectures": [
3
- "LlamaForCausalLM"
4
  ],
5
- "attention_bias": false,
6
- "attention_dropout": 0.0,
7
- "bos_token_id": 1,
8
- "dtype": "bfloat16",
9
- "eos_token_id": 2,
10
- "head_dim": 64,
11
- "hidden_act": "silu",
12
- "hidden_size": 576,
13
- "initializer_range": 0.041666666666666664,
14
- "intermediate_size": 1536,
15
- "is_llama_config": true,
16
- "max_position_embeddings": 8192,
17
- "mlp_bias": false,
18
- "model_type": "llama",
19
- "num_attention_heads": 9,
20
- "num_hidden_layers": 30,
21
- "num_key_value_heads": 3,
22
- "pad_token_id": 2,
23
- "pretraining_tp": 1,
24
- "rms_norm_eps": 1e-05,
25
- "rope_interleaved": false,
26
- "rope_scaling": null,
27
- "rope_theta": 100000,
28
- "tie_word_embeddings": true,
29
- "transformers.js_config": {
30
- "kv_cache_dtype": {
31
- "fp16": "float16",
32
- "q4f16": "float16"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
33
  }
34
  },
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
35
  "transformers_version": "4.57.3",
36
- "use_cache": true,
37
- "vocab_size": 114690
 
 
38
  }
 
1
  {
2
  "architectures": [
3
+ "ASRModel"
4
  ],
5
+ "attn_implementation": "flash_attention_2",
6
+ "audio_config": {
7
+ "_name_or_path": "openai/whisper-large-v3-turbo",
8
+ "activation_dropout": 0.0,
9
+ "activation_function": "gelu",
10
+ "apply_spec_augment": false,
11
+ "architectures": [
12
+ "WhisperForConditionalGeneration"
13
+ ],
14
+ "attention_dropout": 0.0,
15
+ "bos_token_id": 50257,
16
+ "classifier_proj_size": 256,
17
+ "d_model": 1280,
18
+ "decoder_attention_heads": 20,
19
+ "decoder_ffn_dim": 5120,
20
+ "decoder_layerdrop": 0.0,
21
+ "decoder_layers": 4,
22
+ "decoder_start_token_id": 50258,
23
+ "dropout": 0.0,
24
+ "dtype": "bfloat16",
25
+ "encoder_attention_heads": 20,
26
+ "encoder_ffn_dim": 5120,
27
+ "encoder_layerdrop": 0.0,
28
+ "encoder_layers": 32,
29
+ "eos_token_id": 50257,
30
+ "init_std": 0.02,
31
+ "mask_feature_length": 10,
32
+ "mask_feature_min_masks": 0,
33
+ "mask_feature_prob": 0.0,
34
+ "mask_time_length": 10,
35
+ "mask_time_min_masks": 2,
36
+ "mask_time_prob": 0.05,
37
+ "max_source_positions": 1500,
38
+ "max_target_positions": 448,
39
+ "median_filter_width": 7,
40
+ "model_type": "whisper",
41
+ "num_hidden_layers": 32,
42
+ "num_mel_bins": 128,
43
+ "pad_token_id": 50257,
44
+ "scale_embedding": false,
45
+ "use_cache": true,
46
+ "use_weighted_layer_sum": false,
47
+ "vocab_size": 51866
48
+ },
49
+ "audio_model_id": "openai/whisper-large-v3-turbo",
50
+ "audio_sample_rate": 16000,
51
+ "auto_map": {
52
+ "AutoConfig": "asr_config.ASRConfig",
53
+ "AutoModel": "asr_modeling.ASRModel",
54
+ "AutoModelForSpeechSeq2Seq": "asr_modeling.ASRModel",
55
+ "AutoProcessor": "asr_processing.ASRProcessor"
56
+ },
57
+ "custom_pipelines": {
58
+ "automatic-speech-recognition": {
59
+ "impl": "asr_pipeline.ASRPipeline",
60
+ "pt": [
61
+ "AutoModelForSpeechSeq2Seq"
62
+ ],
63
+ "tf": [],
64
+ "type": "audio"
65
  }
66
  },
67
+ "decoder_model_name": "HuggingFaceTB/SmolLM3-3B",
68
+ "downsample_rate": 16,
69
+ "dtype": "bfloat16",
70
+ "encoder_dim": 1280,
71
+ "encoder_model_name": "openai/whisper-large-v3-turbo",
72
+ "inference_diversity_penalty": 0.0,
73
+ "inference_warmup_tokens": 10,
74
+ "label_smoothing": 0.0,
75
+ "llm_dim": 2048,
76
+ "max_new_tokens": 96,
77
+ "min_new_tokens": 0,
78
+ "model_dtype": "bfloat16",
79
+ "model_type": "asr_model",
80
+ "num_experts": 4,
81
+ "num_experts_per_tok": 2,
82
+ "pipeline_tag": "automatic-speech-recognition",
83
+ "projector_dropout": 0.0,
84
+ "projector_hidden_dim": null,
85
+ "projector_init_std": 0.02,
86
+ "projector_input_noise": 0.02,
87
+ "projector_num_layers": 2,
88
+ "projector_pool_stride": 2,
89
+ "projector_type": "mlp",
90
+ "router_aux_loss_coef": 0.01,
91
+ "system_prompt": "/no_think /system_override",
92
+ "temperature": 0.1,
93
+ "text_config": {
94
+ "_name_or_path": "HuggingFaceTB/SmolLM3-3B",
95
+ "architectures": [
96
+ "SmolLM3ForCausalLM"
97
+ ],
98
+ "attention_bias": false,
99
+ "attention_dropout": 0.0,
100
+ "bos_token_id": null,
101
+ "dtype": "bfloat16",
102
+ "eos_token_id": 128012,
103
+ "hidden_act": "silu",
104
+ "hidden_size": 2048,
105
+ "initializer_range": 0.02,
106
+ "intermediate_size": 11008,
107
+ "layer_types": [
108
+ "full_attention",
109
+ "full_attention",
110
+ "full_attention",
111
+ "full_attention",
112
+ "full_attention",
113
+ "full_attention",
114
+ "full_attention",
115
+ "full_attention",
116
+ "full_attention",
117
+ "full_attention",
118
+ "full_attention",
119
+ "full_attention",
120
+ "full_attention",
121
+ "full_attention",
122
+ "full_attention",
123
+ "full_attention",
124
+ "full_attention",
125
+ "full_attention",
126
+ "full_attention",
127
+ "full_attention",
128
+ "full_attention",
129
+ "full_attention",
130
+ "full_attention",
131
+ "full_attention",
132
+ "full_attention",
133
+ "full_attention",
134
+ "full_attention",
135
+ "full_attention",
136
+ "full_attention",
137
+ "full_attention",
138
+ "full_attention",
139
+ "full_attention",
140
+ "full_attention",
141
+ "full_attention",
142
+ "full_attention",
143
+ "full_attention"
144
+ ],
145
+ "max_position_embeddings": 65536,
146
+ "max_window_layers": 28,
147
+ "mlp_bias": false,
148
+ "model_type": "smollm3",
149
+ "no_rope_layer_interval": 4,
150
+ "no_rope_layers": [
151
+ 1,
152
+ 1,
153
+ 1,
154
+ 0,
155
+ 1,
156
+ 1,
157
+ 1,
158
+ 0,
159
+ 1,
160
+ 1,
161
+ 1,
162
+ 0,
163
+ 1,
164
+ 1,
165
+ 1,
166
+ 0,
167
+ 1,
168
+ 1,
169
+ 1,
170
+ 0,
171
+ 1,
172
+ 1,
173
+ 1,
174
+ 0,
175
+ 1,
176
+ 1,
177
+ 1,
178
+ 0,
179
+ 1,
180
+ 1,
181
+ 1,
182
+ 0,
183
+ 1,
184
+ 1,
185
+ 1,
186
+ 0
187
+ ],
188
+ "num_attention_heads": 16,
189
+ "num_hidden_layers": 36,
190
+ "num_key_value_heads": 4,
191
+ "pretraining_tp": 2,
192
+ "rms_norm_eps": 1e-06,
193
+ "rope_scaling": null,
194
+ "rope_theta": 5000000.0,
195
+ "sliding_window": null,
196
+ "use_cache": false,
197
+ "use_sliding_window": false,
198
+ "vocab_size": 128257
199
+ },
200
+ "text_model_id": "HuggingFaceTB/SmolLM3-3B",
201
  "transformers_version": "4.57.3",
202
+ "use_cache": false,
203
+ "use_specaugment": true,
204
+ "user_prompt": "Transcribe: <audio>",
205
+ "vocab_size": 128257
206
  }
generation_config.json CHANGED
@@ -1,7 +1,10 @@
1
  {
2
- "_from_model_config": true,
3
- "bos_token_id": 1,
4
- "eos_token_id": 2,
5
- "pad_token_id": 2,
 
 
 
6
  "transformers_version": "4.57.3"
7
  }
 
1
  {
2
+ "bos_token_id": 128000,
3
+ "eos_token_id": 128012,
4
+ "max_new_tokens": 96,
5
+ "pad_token_id": 128004,
6
+ "temperature": null,
7
+ "top_k": null,
8
+ "top_p": null,
9
  "transformers_version": "4.57.3"
10
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c131771ff86c5ee00452268a20a97c4cdb23e4563a91f72aa8e0a365b8af82fa
3
- size 344560440
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7588b3b3105bbd102ccd6f2e62cb24546c35457f3f6d9cdbe80dc674b662a922
3
+ size 23462224
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:18f97c11d99501fa1943ef96a3096ee882804d0458ce256f09af8f7f9757f1ac
3
- size 5841
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a614500a8c2b6ea97976e7dbd2e49f64b5adef85d1fffcffceebe71b6fd4ce58
3
+ size 5905