mazesmazes commited on
Commit
2f24ce4
·
verified ·
1 Parent(s): 44ecbc9

Training in progress, step 5000

Browse files
Files changed (4) hide show
  1. config.json +349 -0
  2. generation_config.json +17 -0
  3. model.safetensors +3 -0
  4. training_args.bin +3 -0
config.json ADDED
@@ -0,0 +1,349 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "ASRModel"
4
+ ],
5
+ "attn_implementation": "flash_attention_2",
6
+ "audio_config": {
7
+ "_name_or_path": "zai-org/GLM-ASR-Nano-2512",
8
+ "architectures": [
9
+ "GlmAsrForConditionalGeneration"
10
+ ],
11
+ "audio_config": {
12
+ "_name_or_path": "",
13
+ "architectures": null,
14
+ "attention_dropout": 0.0,
15
+ "chunk_size_feed_forward": 0,
16
+ "dtype": null,
17
+ "head_dim": 64,
18
+ "hidden_act": "gelu",
19
+ "hidden_size": 1280,
20
+ "id2label": {
21
+ "0": "LABEL_0",
22
+ "1": "LABEL_1"
23
+ },
24
+ "initializer_range": 0.02,
25
+ "intermediate_size": 5120,
26
+ "is_encoder_decoder": false,
27
+ "label2id": {
28
+ "LABEL_0": 0,
29
+ "LABEL_1": 1
30
+ },
31
+ "max_position_embeddings": 1500,
32
+ "model_type": "glmasr_encoder",
33
+ "num_attention_heads": 20,
34
+ "num_hidden_layers": 32,
35
+ "num_key_value_heads": 20,
36
+ "num_mel_bins": 128,
37
+ "output_attentions": false,
38
+ "output_hidden_states": false,
39
+ "partial_rotary_factor": 0.5,
40
+ "problem_type": null,
41
+ "return_dict": true,
42
+ "rope_parameters": {
43
+ "partial_rotary_factor": 0.5,
44
+ "rope_theta": 10000.0,
45
+ "rope_type": "default"
46
+ }
47
+ },
48
+ "audio_token_id": 59260,
49
+ "dtype": "bfloat16",
50
+ "hidden_size": 2048,
51
+ "model_type": "glmasr",
52
+ "num_mel_bins": 128,
53
+ "projector_hidden_act": "gelu",
54
+ "text_config": {
55
+ "_name_or_path": "",
56
+ "architectures": null,
57
+ "attention_bias": false,
58
+ "attention_dropout": 0.0,
59
+ "bos_token_id": 1,
60
+ "chunk_size_feed_forward": 0,
61
+ "dtype": null,
62
+ "eos_token_id": [
63
+ 59246,
64
+ 59253,
65
+ 59255
66
+ ],
67
+ "head_dim": 128,
68
+ "hidden_act": "silu",
69
+ "hidden_size": 2048,
70
+ "id2label": {
71
+ "0": "LABEL_0",
72
+ "1": "LABEL_1"
73
+ },
74
+ "initializer_range": 0.02,
75
+ "intermediate_size": 6144,
76
+ "is_encoder_decoder": false,
77
+ "label2id": {
78
+ "LABEL_0": 0,
79
+ "LABEL_1": 1
80
+ },
81
+ "max_position_embeddings": 8192,
82
+ "mlp_bias": false,
83
+ "model_type": "llama",
84
+ "num_attention_heads": 16,
85
+ "num_hidden_layers": 28,
86
+ "num_key_value_heads": 4,
87
+ "output_attentions": false,
88
+ "output_hidden_states": false,
89
+ "pad_token_id": null,
90
+ "pretraining_tp": 1,
91
+ "problem_type": null,
92
+ "return_dict": true,
93
+ "rms_norm_eps": 1e-05,
94
+ "rope_parameters": {
95
+ "rope_theta": 10000.0,
96
+ "rope_type": "default"
97
+ },
98
+ "tie_word_embeddings": false,
99
+ "use_cache": true,
100
+ "vocab_size": 59264
101
+ },
102
+ "vocab_size": 59264
103
+ },
104
+ "audio_model_id": "zai-org/GLM-ASR-Nano-2512",
105
+ "audio_sample_rate": 16000,
106
+ "auto_map": {
107
+ "AutoConfig": "asr_config.ASRConfig",
108
+ "AutoModel": "asr_modeling.ASRModel",
109
+ "AutoModelForSpeechSeq2Seq": "asr_modeling.ASRModel",
110
+ "AutoProcessor": "asr_processing.ASRProcessor"
111
+ },
112
+ "custom_pipelines": {
113
+ "automatic-speech-recognition": {
114
+ "impl": "asr_pipeline.ASRPipeline",
115
+ "pt": [
116
+ "AutoModelForSpeechSeq2Seq"
117
+ ],
118
+ "tf": [],
119
+ "type": "audio"
120
+ }
121
+ },
122
+ "downsample_rate": 5,
123
+ "dtype": "bfloat16",
124
+ "encoder": {
125
+ "_name_or_path": "zai-org/GLM-ASR-Nano-2512",
126
+ "architectures": [
127
+ "GlmAsrForConditionalGeneration"
128
+ ],
129
+ "audio_config": {
130
+ "_name_or_path": "",
131
+ "architectures": null,
132
+ "attention_dropout": 0.0,
133
+ "chunk_size_feed_forward": 0,
134
+ "dtype": null,
135
+ "head_dim": 64,
136
+ "hidden_act": "gelu",
137
+ "hidden_size": 1280,
138
+ "id2label": {
139
+ "0": "LABEL_0",
140
+ "1": "LABEL_1"
141
+ },
142
+ "initializer_range": 0.02,
143
+ "intermediate_size": 5120,
144
+ "is_encoder_decoder": false,
145
+ "label2id": {
146
+ "LABEL_0": 0,
147
+ "LABEL_1": 1
148
+ },
149
+ "max_position_embeddings": 1500,
150
+ "model_type": "glmasr_encoder",
151
+ "num_attention_heads": 20,
152
+ "num_hidden_layers": 32,
153
+ "num_key_value_heads": 20,
154
+ "num_mel_bins": 128,
155
+ "output_attentions": false,
156
+ "output_hidden_states": false,
157
+ "partial_rotary_factor": 0.5,
158
+ "problem_type": null,
159
+ "return_dict": true,
160
+ "rope_parameters": {
161
+ "partial_rotary_factor": 0.5,
162
+ "rope_theta": 10000.0,
163
+ "rope_type": "default"
164
+ }
165
+ },
166
+ "audio_token_id": 59260,
167
+ "dtype": "bfloat16",
168
+ "hidden_size": 2048,
169
+ "model_type": "glmasr",
170
+ "num_mel_bins": 128,
171
+ "projector_hidden_act": "gelu",
172
+ "text_config": {
173
+ "_name_or_path": "",
174
+ "architectures": null,
175
+ "attention_bias": false,
176
+ "attention_dropout": 0.0,
177
+ "bos_token_id": 1,
178
+ "chunk_size_feed_forward": 0,
179
+ "dtype": null,
180
+ "eos_token_id": [
181
+ 59246,
182
+ 59253,
183
+ 59255
184
+ ],
185
+ "head_dim": 128,
186
+ "hidden_act": "silu",
187
+ "hidden_size": 2048,
188
+ "id2label": {
189
+ "0": "LABEL_0",
190
+ "1": "LABEL_1"
191
+ },
192
+ "initializer_range": 0.02,
193
+ "intermediate_size": 6144,
194
+ "is_encoder_decoder": false,
195
+ "label2id": {
196
+ "LABEL_0": 0,
197
+ "LABEL_1": 1
198
+ },
199
+ "max_position_embeddings": 8192,
200
+ "mlp_bias": false,
201
+ "model_type": "llama",
202
+ "num_attention_heads": 16,
203
+ "num_hidden_layers": 28,
204
+ "num_key_value_heads": 4,
205
+ "output_attentions": false,
206
+ "output_hidden_states": false,
207
+ "pad_token_id": null,
208
+ "pretraining_tp": 1,
209
+ "problem_type": null,
210
+ "return_dict": true,
211
+ "rms_norm_eps": 1e-05,
212
+ "rope_parameters": {
213
+ "rope_theta": 10000.0,
214
+ "rope_type": "default"
215
+ },
216
+ "tie_word_embeddings": false,
217
+ "use_cache": true,
218
+ "vocab_size": 59264
219
+ },
220
+ "vocab_size": 59264
221
+ },
222
+ "encoder_conv_layers": [
223
+ [
224
+ 1,
225
+ 3,
226
+ 1
227
+ ],
228
+ [
229
+ 1,
230
+ 3,
231
+ 2
232
+ ]
233
+ ],
234
+ "encoder_dim": 1280,
235
+ "freeze_projector": false,
236
+ "freq_mask_length": 27,
237
+ "inference_warmup_tokens": 10,
238
+ "label_smoothing": 0.0,
239
+ "length_penalty": 1.0,
240
+ "llm_dim": 2048,
241
+ "lora_alpha": 32,
242
+ "lora_dropout": 0.0,
243
+ "lora_rank": 8,
244
+ "lora_target_modules": [
245
+ "q_proj",
246
+ "k_proj",
247
+ "v_proj",
248
+ "o_proj",
249
+ "gate_proj",
250
+ "up_proj",
251
+ "down_proj"
252
+ ],
253
+ "max_new_tokens": 128,
254
+ "min_new_tokens": 0,
255
+ "model_dtype": "bfloat16",
256
+ "model_type": "asr_model",
257
+ "no_repeat_ngram_size": 0,
258
+ "num_beams": 1,
259
+ "num_experts": 4,
260
+ "num_experts_per_tok": 2,
261
+ "num_freq_masks": 1,
262
+ "num_time_masks": 1,
263
+ "pipeline_tag": "automatic-speech-recognition",
264
+ "pretrained_model_path": "mazesmazes/tiny-audio-multitask",
265
+ "projector_dropout": 0.0,
266
+ "projector_hidden_dim": null,
267
+ "projector_init_std": 0.02,
268
+ "projector_num_layers": 2,
269
+ "projector_pool_stride": 4,
270
+ "projector_type": "mlp",
271
+ "qformer_hidden_size": null,
272
+ "qformer_intermediate_size": null,
273
+ "qformer_num_heads": 16,
274
+ "qformer_num_layers": 2,
275
+ "qformer_window_size": 15,
276
+ "repetition_penalty": 1.0,
277
+ "router_aux_loss_coef": 0.01,
278
+ "system_prompt": "",
279
+ "text_config": {
280
+ "_name_or_path": "Qwen/Qwen3-1.7B",
281
+ "architectures": [
282
+ "Qwen3ForCausalLM"
283
+ ],
284
+ "attention_bias": false,
285
+ "attention_dropout": 0.0,
286
+ "bos_token_id": null,
287
+ "dtype": "bfloat16",
288
+ "eos_token_id": 151645,
289
+ "head_dim": 128,
290
+ "hidden_act": "silu",
291
+ "hidden_size": 2048,
292
+ "initializer_range": 0.02,
293
+ "intermediate_size": 6144,
294
+ "layer_types": [
295
+ "full_attention",
296
+ "full_attention",
297
+ "full_attention",
298
+ "full_attention",
299
+ "full_attention",
300
+ "full_attention",
301
+ "full_attention",
302
+ "full_attention",
303
+ "full_attention",
304
+ "full_attention",
305
+ "full_attention",
306
+ "full_attention",
307
+ "full_attention",
308
+ "full_attention",
309
+ "full_attention",
310
+ "full_attention",
311
+ "full_attention",
312
+ "full_attention",
313
+ "full_attention",
314
+ "full_attention",
315
+ "full_attention",
316
+ "full_attention",
317
+ "full_attention",
318
+ "full_attention",
319
+ "full_attention",
320
+ "full_attention",
321
+ "full_attention",
322
+ "full_attention"
323
+ ],
324
+ "max_position_embeddings": 40960,
325
+ "max_window_layers": 28,
326
+ "model_type": "qwen3",
327
+ "num_attention_heads": 16,
328
+ "num_hidden_layers": 28,
329
+ "num_key_value_heads": 8,
330
+ "pad_token_id": 151643,
331
+ "rms_norm_eps": 1e-06,
332
+ "rope_parameters": {
333
+ "rope_theta": 1000000,
334
+ "rope_type": "default"
335
+ },
336
+ "sliding_window": null,
337
+ "tie_word_embeddings": true,
338
+ "use_cache": true,
339
+ "use_sliding_window": false,
340
+ "vocab_size": 151670
341
+ },
342
+ "text_model_id": "Qwen/Qwen3-1.7B",
343
+ "time_mask_length": 100,
344
+ "transformers_version": "5.0.0.dev0",
345
+ "use_cache": false,
346
+ "use_lora": false,
347
+ "use_specaugment": true,
348
+ "vocab_size": 151670
349
+ }
generation_config.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "do_sample": false,
4
+ "eos_token_id": [
5
+ 151645,
6
+ 151643
7
+ ],
8
+ "length_penalty": 1.0,
9
+ "max_new_tokens": 128,
10
+ "min_new_tokens": 0,
11
+ "no_repeat_ngram_size": 0,
12
+ "num_beams": 1,
13
+ "pad_token_id": 151643,
14
+ "repetition_penalty": 1.0,
15
+ "transformers_version": "5.0.0.dev0",
16
+ "use_cache": true
17
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f600f7687be1302c86a8bdcb62a32a2d210f560e8147e977729825b383d2d15a
3
+ size 58732960
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:060e9c0aa3404e278bdec0b14f0a68b938b1ec109a6bbd119f1c4dfcbb908299
3
+ size 5265