mazesmazes commited on
Commit
9809418
·
verified ·
1 Parent(s): 70f3fe7

Training in progress, step 20000

Browse files
Files changed (4) hide show
  1. config.json +392 -0
  2. generation_config.json +15 -0
  3. model.safetensors +3 -0
  4. training_args.bin +3 -0
config.json ADDED
@@ -0,0 +1,392 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "add_cross_attention": false,
14
+ "architectures": null,
15
+ "attention_dropout": 0.0,
16
+ "bos_token_id": null,
17
+ "chunk_size_feed_forward": 0,
18
+ "cross_attention_hidden_size": null,
19
+ "decoder_start_token_id": null,
20
+ "dtype": null,
21
+ "eos_token_id": null,
22
+ "finetuning_task": null,
23
+ "head_dim": 64,
24
+ "hidden_act": "gelu",
25
+ "hidden_size": 1280,
26
+ "id2label": {
27
+ "0": "LABEL_0",
28
+ "1": "LABEL_1"
29
+ },
30
+ "initializer_range": 0.02,
31
+ "intermediate_size": 5120,
32
+ "is_decoder": false,
33
+ "is_encoder_decoder": false,
34
+ "label2id": {
35
+ "LABEL_0": 0,
36
+ "LABEL_1": 1
37
+ },
38
+ "max_position_embeddings": 1500,
39
+ "model_type": "glmasr_encoder",
40
+ "num_attention_heads": 20,
41
+ "num_hidden_layers": 32,
42
+ "num_key_value_heads": 20,
43
+ "num_mel_bins": 128,
44
+ "output_attentions": false,
45
+ "output_hidden_states": false,
46
+ "pad_token_id": null,
47
+ "partial_rotary_factor": 0.5,
48
+ "prefix": null,
49
+ "problem_type": null,
50
+ "return_dict": true,
51
+ "rope_parameters": {
52
+ "partial_rotary_factor": 0.5,
53
+ "rope_theta": 10000.0,
54
+ "rope_type": "default"
55
+ },
56
+ "sep_token_id": null,
57
+ "task_specific_params": null,
58
+ "tie_word_embeddings": true,
59
+ "tokenizer_class": null
60
+ },
61
+ "audio_token_id": 59260,
62
+ "dtype": "bfloat16",
63
+ "hidden_size": 2048,
64
+ "model_type": "glmasr",
65
+ "num_mel_bins": 128,
66
+ "projector_hidden_act": "gelu",
67
+ "text_config": {
68
+ "_name_or_path": "",
69
+ "add_cross_attention": false,
70
+ "architectures": null,
71
+ "attention_bias": false,
72
+ "attention_dropout": 0.0,
73
+ "bos_token_id": 1,
74
+ "chunk_size_feed_forward": 0,
75
+ "cross_attention_hidden_size": null,
76
+ "decoder_start_token_id": null,
77
+ "dtype": null,
78
+ "eos_token_id": [
79
+ 59246,
80
+ 59253,
81
+ 59255
82
+ ],
83
+ "finetuning_task": null,
84
+ "head_dim": 128,
85
+ "hidden_act": "silu",
86
+ "hidden_size": 2048,
87
+ "id2label": {
88
+ "0": "LABEL_0",
89
+ "1": "LABEL_1"
90
+ },
91
+ "initializer_range": 0.02,
92
+ "intermediate_size": 6144,
93
+ "is_decoder": false,
94
+ "is_encoder_decoder": false,
95
+ "label2id": {
96
+ "LABEL_0": 0,
97
+ "LABEL_1": 1
98
+ },
99
+ "max_position_embeddings": 8192,
100
+ "mlp_bias": false,
101
+ "model_type": "llama",
102
+ "num_attention_heads": 16,
103
+ "num_hidden_layers": 28,
104
+ "num_key_value_heads": 4,
105
+ "output_attentions": false,
106
+ "output_hidden_states": false,
107
+ "pad_token_id": null,
108
+ "prefix": null,
109
+ "pretraining_tp": 1,
110
+ "problem_type": null,
111
+ "return_dict": true,
112
+ "rms_norm_eps": 1e-05,
113
+ "rope_parameters": {
114
+ "rope_theta": 10000.0,
115
+ "rope_type": "default"
116
+ },
117
+ "sep_token_id": null,
118
+ "task_specific_params": null,
119
+ "tie_word_embeddings": false,
120
+ "tokenizer_class": null,
121
+ "use_cache": true,
122
+ "vocab_size": 59264
123
+ },
124
+ "vocab_size": 59264
125
+ },
126
+ "audio_model_id": "zai-org/GLM-ASR-Nano-2512",
127
+ "audio_sample_rate": 16000,
128
+ "auto_map": {
129
+ "AutoConfig": "asr_config.ASRConfig",
130
+ "AutoModel": "asr_modeling.ASRModel",
131
+ "AutoModelForSpeechSeq2Seq": "asr_modeling.ASRModel",
132
+ "AutoProcessor": "asr_processing.ASRProcessor"
133
+ },
134
+ "custom_pipelines": {
135
+ "automatic-speech-recognition": {
136
+ "impl": "asr_pipeline.ASRPipeline",
137
+ "pt": [
138
+ "AutoModelForSpeechSeq2Seq"
139
+ ],
140
+ "tf": [],
141
+ "type": "audio"
142
+ }
143
+ },
144
+ "downsample_rate": 5,
145
+ "dtype": "bfloat16",
146
+ "encoder": {
147
+ "_name_or_path": "zai-org/GLM-ASR-Nano-2512",
148
+ "architectures": [
149
+ "GlmAsrForConditionalGeneration"
150
+ ],
151
+ "audio_config": {
152
+ "_name_or_path": "",
153
+ "add_cross_attention": false,
154
+ "architectures": null,
155
+ "attention_dropout": 0.0,
156
+ "bos_token_id": null,
157
+ "chunk_size_feed_forward": 0,
158
+ "cross_attention_hidden_size": null,
159
+ "decoder_start_token_id": null,
160
+ "dtype": null,
161
+ "eos_token_id": null,
162
+ "finetuning_task": null,
163
+ "head_dim": 64,
164
+ "hidden_act": "gelu",
165
+ "hidden_size": 1280,
166
+ "id2label": {
167
+ "0": "LABEL_0",
168
+ "1": "LABEL_1"
169
+ },
170
+ "initializer_range": 0.02,
171
+ "intermediate_size": 5120,
172
+ "is_decoder": false,
173
+ "is_encoder_decoder": false,
174
+ "label2id": {
175
+ "LABEL_0": 0,
176
+ "LABEL_1": 1
177
+ },
178
+ "max_position_embeddings": 1500,
179
+ "model_type": "glmasr_encoder",
180
+ "num_attention_heads": 20,
181
+ "num_hidden_layers": 32,
182
+ "num_key_value_heads": 20,
183
+ "num_mel_bins": 128,
184
+ "output_attentions": false,
185
+ "output_hidden_states": false,
186
+ "pad_token_id": null,
187
+ "partial_rotary_factor": 0.5,
188
+ "prefix": null,
189
+ "problem_type": null,
190
+ "return_dict": true,
191
+ "rope_parameters": {
192
+ "partial_rotary_factor": 0.5,
193
+ "rope_theta": 10000.0,
194
+ "rope_type": "default"
195
+ },
196
+ "sep_token_id": null,
197
+ "task_specific_params": null,
198
+ "tie_word_embeddings": true,
199
+ "tokenizer_class": null
200
+ },
201
+ "audio_token_id": 59260,
202
+ "dtype": "bfloat16",
203
+ "hidden_size": 2048,
204
+ "model_type": "glmasr",
205
+ "num_mel_bins": 128,
206
+ "projector_hidden_act": "gelu",
207
+ "text_config": {
208
+ "_name_or_path": "",
209
+ "add_cross_attention": false,
210
+ "architectures": null,
211
+ "attention_bias": false,
212
+ "attention_dropout": 0.0,
213
+ "bos_token_id": 1,
214
+ "chunk_size_feed_forward": 0,
215
+ "cross_attention_hidden_size": null,
216
+ "decoder_start_token_id": null,
217
+ "dtype": null,
218
+ "eos_token_id": [
219
+ 59246,
220
+ 59253,
221
+ 59255
222
+ ],
223
+ "finetuning_task": null,
224
+ "head_dim": 128,
225
+ "hidden_act": "silu",
226
+ "hidden_size": 2048,
227
+ "id2label": {
228
+ "0": "LABEL_0",
229
+ "1": "LABEL_1"
230
+ },
231
+ "initializer_range": 0.02,
232
+ "intermediate_size": 6144,
233
+ "is_decoder": false,
234
+ "is_encoder_decoder": false,
235
+ "label2id": {
236
+ "LABEL_0": 0,
237
+ "LABEL_1": 1
238
+ },
239
+ "max_position_embeddings": 8192,
240
+ "mlp_bias": false,
241
+ "model_type": "llama",
242
+ "num_attention_heads": 16,
243
+ "num_hidden_layers": 28,
244
+ "num_key_value_heads": 4,
245
+ "output_attentions": false,
246
+ "output_hidden_states": false,
247
+ "pad_token_id": null,
248
+ "prefix": null,
249
+ "pretraining_tp": 1,
250
+ "problem_type": null,
251
+ "return_dict": true,
252
+ "rms_norm_eps": 1e-05,
253
+ "rope_parameters": {
254
+ "rope_theta": 10000.0,
255
+ "rope_type": "default"
256
+ },
257
+ "sep_token_id": null,
258
+ "task_specific_params": null,
259
+ "tie_word_embeddings": false,
260
+ "tokenizer_class": null,
261
+ "use_cache": true,
262
+ "vocab_size": 59264
263
+ },
264
+ "vocab_size": 59264
265
+ },
266
+ "encoder_conv_layers": [
267
+ [
268
+ 1,
269
+ 3,
270
+ 1
271
+ ],
272
+ [
273
+ 1,
274
+ 3,
275
+ 2
276
+ ]
277
+ ],
278
+ "encoder_dim": 1280,
279
+ "freeze_projector": false,
280
+ "freq_mask_length": 27,
281
+ "inference_warmup_tokens": 10,
282
+ "label_smoothing": 0.0,
283
+ "length_penalty": 1.0,
284
+ "llm_dim": 1024,
285
+ "lora_alpha": 32,
286
+ "lora_dropout": 0.0,
287
+ "lora_rank": 8,
288
+ "lora_target_modules": [
289
+ "q_proj",
290
+ "k_proj",
291
+ "v_proj",
292
+ "o_proj",
293
+ "gate_proj",
294
+ "up_proj",
295
+ "down_proj"
296
+ ],
297
+ "max_new_tokens": 128,
298
+ "min_new_tokens": 0,
299
+ "model_dtype": "bfloat16",
300
+ "model_type": "asr_model",
301
+ "no_repeat_ngram_size": 0,
302
+ "num_beams": 1,
303
+ "num_experts": 4,
304
+ "num_experts_per_tok": 2,
305
+ "num_freq_masks": 2,
306
+ "num_time_masks": 10,
307
+ "pipeline_tag": "automatic-speech-recognition",
308
+ "pretrained_model_path": "mazesmazes/tiny-audio",
309
+ "projector_dropout": 0.0,
310
+ "projector_hidden_dim": null,
311
+ "projector_init_std": 0.02,
312
+ "projector_num_layers": 2,
313
+ "projector_pool_stride": 4,
314
+ "projector_type": "mlp",
315
+ "qformer_hidden_size": null,
316
+ "qformer_intermediate_size": null,
317
+ "qformer_num_heads": 16,
318
+ "qformer_num_layers": 2,
319
+ "qformer_window_size": 15,
320
+ "repetition_penalty": 1.0,
321
+ "router_aux_loss_coef": 0.01,
322
+ "system_prompt": "",
323
+ "text_config": {
324
+ "_name_or_path": "Qwen/Qwen3-0.6B",
325
+ "architectures": [
326
+ "Qwen3ForCausalLM"
327
+ ],
328
+ "attention_bias": false,
329
+ "attention_dropout": 0.0,
330
+ "dtype": "bfloat16",
331
+ "eos_token_id": 151645,
332
+ "head_dim": 128,
333
+ "hidden_act": "silu",
334
+ "hidden_size": 1024,
335
+ "initializer_range": 0.02,
336
+ "intermediate_size": 3072,
337
+ "layer_types": [
338
+ "full_attention",
339
+ "full_attention",
340
+ "full_attention",
341
+ "full_attention",
342
+ "full_attention",
343
+ "full_attention",
344
+ "full_attention",
345
+ "full_attention",
346
+ "full_attention",
347
+ "full_attention",
348
+ "full_attention",
349
+ "full_attention",
350
+ "full_attention",
351
+ "full_attention",
352
+ "full_attention",
353
+ "full_attention",
354
+ "full_attention",
355
+ "full_attention",
356
+ "full_attention",
357
+ "full_attention",
358
+ "full_attention",
359
+ "full_attention",
360
+ "full_attention",
361
+ "full_attention",
362
+ "full_attention",
363
+ "full_attention",
364
+ "full_attention",
365
+ "full_attention"
366
+ ],
367
+ "max_position_embeddings": 40960,
368
+ "max_window_layers": 28,
369
+ "model_type": "qwen3",
370
+ "num_attention_heads": 16,
371
+ "num_hidden_layers": 28,
372
+ "num_key_value_heads": 8,
373
+ "pad_token_id": 151643,
374
+ "rms_norm_eps": 1e-06,
375
+ "rope_parameters": {
376
+ "rope_theta": 1000000,
377
+ "rope_type": "default"
378
+ },
379
+ "sliding_window": null,
380
+ "tie_word_embeddings": true,
381
+ "use_cache": true,
382
+ "use_sliding_window": false,
383
+ "vocab_size": 151670
384
+ },
385
+ "text_model_id": "Qwen/Qwen3-0.6B",
386
+ "time_mask_length": 100,
387
+ "transformers_version": "5.0.0.dev0",
388
+ "use_cache": false,
389
+ "use_lora": false,
390
+ "use_specaugment": true,
391
+ "vocab_size": 151670
392
+ }
generation_config.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "eos_token_id": [
4
+ 151645,
5
+ 151643
6
+ ],
7
+ "length_penalty": 1.0,
8
+ "max_new_tokens": 128,
9
+ "min_new_tokens": 0,
10
+ "no_repeat_ngram_size": 0,
11
+ "num_beams": 1,
12
+ "pad_token_id": 151643,
13
+ "repetition_penalty": 1.0,
14
+ "transformers_version": "5.0.0.dev0"
15
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:464c5a90f99e6f5810b0eaad1a331af966925588be6a6611ddf49220caac7a8c
3
+ size 12587424
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18572cd996e881f07a0575e01bb0e0151fd4feb2cf38350b9c02d1ee7f580ec9
3
+ size 5265