Quentin Meeus
commited on
Commit
•
e508ee7
1
Parent(s):
1e16abf
add logs
Browse files
logs/whisper-spoken-ner-small-pipe-lora.err
ADDED
@@ -0,0 +1,659 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Process rank: 0, device: cuda:0, n_gpu: 1distributed training: True, 16-bits training: True
|
2 |
+
/users/spraak/qmeeus/micromamba/envs/torch-cu121/lib/python3.10/site-packages/transformers/configuration_utils.py:508: FutureWarning: The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.
|
3 |
+
warnings.warn(
|
4 |
+
[INFO|configuration_utils.py:737] 2024-01-08 18:44:38,532 >> loading configuration file configs/whisper_small_ner_mtl.json
|
5 |
+
[WARNING|configuration_utils.py:617] 2024-01-08 18:44:38,532 >> You are using a model of type whisper to instantiate a model of type whisper_for_slu. This is not supported for all configurations of models and can yield errors.
|
6 |
+
[INFO|configuration_utils.py:802] 2024-01-08 18:44:38,535 >> Model config WhisperSLUConfig {
|
7 |
+
"_name_or_path": "openai/whisper-small",
|
8 |
+
"activation_dropout": 0.0,
|
9 |
+
"activation_function": "gelu",
|
10 |
+
"adaptor_activation": "relu",
|
11 |
+
"adaptor_init": "constant",
|
12 |
+
"adaptor_layernorm": true,
|
13 |
+
"apply_spec_augment": false,
|
14 |
+
"architectures": [
|
15 |
+
"WhisperForConditionalGeneration"
|
16 |
+
],
|
17 |
+
"attention_dropout": 0.0,
|
18 |
+
"begin_suppress_tokens": [
|
19 |
+
220,
|
20 |
+
50257
|
21 |
+
],
|
22 |
+
"bos_token_id": 50257,
|
23 |
+
"classifier_proj_size": 256,
|
24 |
+
"crf_transition_matrix": null,
|
25 |
+
"d_model": 768,
|
26 |
+
"decoder_attention_heads": 12,
|
27 |
+
"decoder_ffn_dim": 3072,
|
28 |
+
"decoder_layerdrop": 0.0,
|
29 |
+
"decoder_layers": 12,
|
30 |
+
"decoder_start_token_id": 50258,
|
31 |
+
"dropout": 0.0,
|
32 |
+
"encoder_attention_heads": 12,
|
33 |
+
"encoder_ffn_dim": 3072,
|
34 |
+
"encoder_layerdrop": 0.0,
|
35 |
+
"encoder_layers": 12,
|
36 |
+
"eos_token_id": 50257,
|
37 |
+
"forced_decoder_ids": [
|
38 |
+
[
|
39 |
+
1,
|
40 |
+
50259
|
41 |
+
],
|
42 |
+
[
|
43 |
+
2,
|
44 |
+
50359
|
45 |
+
],
|
46 |
+
[
|
47 |
+
3,
|
48 |
+
50363
|
49 |
+
]
|
50 |
+
],
|
51 |
+
"init_std": 0.02,
|
52 |
+
"is_encoder_decoder": true,
|
53 |
+
"mask_feature_length": 10,
|
54 |
+
"mask_feature_min_masks": 0,
|
55 |
+
"mask_feature_prob": 0.0,
|
56 |
+
"mask_time_length": 10,
|
57 |
+
"mask_time_min_masks": 2,
|
58 |
+
"mask_time_prob": 0.05,
|
59 |
+
"max_length": 448,
|
60 |
+
"max_source_positions": 1500,
|
61 |
+
"max_target_positions": 448,
|
62 |
+
"median_filter_width": 7,
|
63 |
+
"model_type": "whisper_for_slu",
|
64 |
+
"num_hidden_layers": 12,
|
65 |
+
"num_mel_bins": 80,
|
66 |
+
"pad_token_id": 50257,
|
67 |
+
"scale_embedding": false,
|
68 |
+
"slu_attention_heads": 12,
|
69 |
+
"slu_dropout": 0.3,
|
70 |
+
"slu_embed_dim": 768,
|
71 |
+
"slu_ffn_dim": 2048,
|
72 |
+
"slu_focus": 1.0,
|
73 |
+
"slu_input_from": "decoder",
|
74 |
+
"slu_input_layers": [
|
75 |
+
11
|
76 |
+
],
|
77 |
+
"slu_labels": null,
|
78 |
+
"slu_layers": 2,
|
79 |
+
"slu_max_positions": null,
|
80 |
+
"slu_output_dim": 37,
|
81 |
+
"slu_pad_token_id": 1,
|
82 |
+
"slu_start_token_id": 36,
|
83 |
+
"slu_task": "named_entity_recognition",
|
84 |
+
"slu_weight": 0.2,
|
85 |
+
"suppress_tokens": [
|
86 |
+
1,
|
87 |
+
2,
|
88 |
+
7,
|
89 |
+
8,
|
90 |
+
9,
|
91 |
+
10,
|
92 |
+
14,
|
93 |
+
25,
|
94 |
+
26,
|
95 |
+
27,
|
96 |
+
28,
|
97 |
+
29,
|
98 |
+
31,
|
99 |
+
58,
|
100 |
+
59,
|
101 |
+
60,
|
102 |
+
61,
|
103 |
+
62,
|
104 |
+
63,
|
105 |
+
90,
|
106 |
+
91,
|
107 |
+
92,
|
108 |
+
93,
|
109 |
+
359,
|
110 |
+
503,
|
111 |
+
522,
|
112 |
+
542,
|
113 |
+
873,
|
114 |
+
893,
|
115 |
+
902,
|
116 |
+
918,
|
117 |
+
922,
|
118 |
+
931,
|
119 |
+
1350,
|
120 |
+
1853,
|
121 |
+
1982,
|
122 |
+
2460,
|
123 |
+
2627,
|
124 |
+
3246,
|
125 |
+
3253,
|
126 |
+
3268,
|
127 |
+
3536,
|
128 |
+
3846,
|
129 |
+
3961,
|
130 |
+
4183,
|
131 |
+
4667,
|
132 |
+
6585,
|
133 |
+
6647,
|
134 |
+
7273,
|
135 |
+
9061,
|
136 |
+
9383,
|
137 |
+
10428,
|
138 |
+
10929,
|
139 |
+
11938,
|
140 |
+
12033,
|
141 |
+
12331,
|
142 |
+
12562,
|
143 |
+
13793,
|
144 |
+
14157,
|
145 |
+
14635,
|
146 |
+
15265,
|
147 |
+
15618,
|
148 |
+
16553,
|
149 |
+
16604,
|
150 |
+
18362,
|
151 |
+
18956,
|
152 |
+
20075,
|
153 |
+
21675,
|
154 |
+
22520,
|
155 |
+
26130,
|
156 |
+
26161,
|
157 |
+
26435,
|
158 |
+
28279,
|
159 |
+
29464,
|
160 |
+
31650,
|
161 |
+
32302,
|
162 |
+
32470,
|
163 |
+
36865,
|
164 |
+
42863,
|
165 |
+
47425,
|
166 |
+
49870,
|
167 |
+
50254,
|
168 |
+
50258,
|
169 |
+
50360,
|
170 |
+
50361,
|
171 |
+
50362
|
172 |
+
],
|
173 |
+
"task": "token_classification",
|
174 |
+
"teacher": null,
|
175 |
+
"torch_dtype": "float32",
|
176 |
+
"transformers_version": "4.37.0.dev0",
|
177 |
+
"use_cache": true,
|
178 |
+
"use_crf": false,
|
179 |
+
"use_weighted_layer_sum": false,
|
180 |
+
"vocab_size": 51865
|
181 |
+
}
|
182 |
+
|
183 |
+
/users/spraak/qmeeus/micromamba/envs/torch-cu121/lib/python3.10/site-packages/transformers/models/auto/feature_extraction_auto.py:328: FutureWarning: The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.
|
184 |
+
warnings.warn(
|
185 |
+
[INFO|feature_extraction_utils.py:535] 2024-01-08 18:44:38,556 >> loading configuration file /esat/audioslave/qmeeus/exp/whisper_slu/train/whisper-small-spoken-ner/preprocessor_config.json
|
186 |
+
[INFO|feature_extraction_utils.py:579] 2024-01-08 18:44:38,563 >> Feature extractor WhisperFeatureExtractor {
|
187 |
+
"chunk_length": 30,
|
188 |
+
"feature_extractor_type": "WhisperFeatureExtractor",
|
189 |
+
"feature_size": 80,
|
190 |
+
"hop_length": 160,
|
191 |
+
"n_fft": 400,
|
192 |
+
"n_samples": 480000,
|
193 |
+
"nb_max_frames": 3000,
|
194 |
+
"padding_side": "right",
|
195 |
+
"padding_value": 0.0,
|
196 |
+
"processor_class": "WhisperProcessor",
|
197 |
+
"return_attention_mask": false,
|
198 |
+
"sampling_rate": 16000
|
199 |
+
}
|
200 |
+
|
201 |
+
/users/spraak/qmeeus/micromamba/envs/torch-cu121/lib/python3.10/site-packages/transformers/models/auto/tokenization_auto.py:691: FutureWarning: The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.
|
202 |
+
warnings.warn(
|
203 |
+
[INFO|tokenization_utils_base.py:2024] 2024-01-08 18:44:38,630 >> loading file vocab.json
|
204 |
+
[INFO|tokenization_utils_base.py:2024] 2024-01-08 18:44:38,630 >> loading file tokenizer.json
|
205 |
+
[INFO|tokenization_utils_base.py:2024] 2024-01-08 18:44:38,630 >> loading file merges.txt
|
206 |
+
[INFO|tokenization_utils_base.py:2024] 2024-01-08 18:44:38,630 >> loading file normalizer.json
|
207 |
+
[INFO|tokenization_utils_base.py:2024] 2024-01-08 18:44:38,630 >> loading file added_tokens.json
|
208 |
+
[INFO|tokenization_utils_base.py:2024] 2024-01-08 18:44:38,631 >> loading file special_tokens_map.json
|
209 |
+
[INFO|tokenization_utils_base.py:2024] 2024-01-08 18:44:38,631 >> loading file tokenizer_config.json
|
210 |
+
[WARNING|logging.py:314] 2024-01-08 18:44:39,435 >> Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
|
211 |
+
/users/spraak/qmeeus/micromamba/envs/torch-cu121/lib/python3.10/site-packages/transformers/modeling_utils.py:2790: FutureWarning: The `use_auth_token` argument is deprecated and will be removed in v5 of Transformers. Please use `token` instead.
|
212 |
+
warnings.warn(
|
213 |
+
[INFO|modeling_utils.py:3373] 2024-01-08 18:44:39,454 >> loading weights file /esat/audioslave/qmeeus/exp/whisper_slu/train/whisper-small-spoken-ner/model.safetensors
|
214 |
+
[INFO|configuration_utils.py:826] 2024-01-08 18:44:41,796 >> Generate config GenerationConfig {
|
215 |
+
"begin_suppress_tokens": [
|
216 |
+
220,
|
217 |
+
50257
|
218 |
+
],
|
219 |
+
"bos_token_id": 50257,
|
220 |
+
"decoder_start_token_id": 50258,
|
221 |
+
"eos_token_id": 50257,
|
222 |
+
"forced_decoder_ids": [
|
223 |
+
[
|
224 |
+
1,
|
225 |
+
50259
|
226 |
+
],
|
227 |
+
[
|
228 |
+
2,
|
229 |
+
50359
|
230 |
+
],
|
231 |
+
[
|
232 |
+
3,
|
233 |
+
50363
|
234 |
+
]
|
235 |
+
],
|
236 |
+
"max_length": 448,
|
237 |
+
"pad_token_id": 50257
|
238 |
+
}
|
239 |
+
|
240 |
+
[INFO|modeling_utils.py:4227] 2024-01-08 18:44:42,780 >> All model checkpoint weights were used when initializing WhisperSLU.
|
241 |
+
|
242 |
+
[INFO|modeling_utils.py:4235] 2024-01-08 18:44:42,780 >> All the weights of WhisperSLU were initialized from the model checkpoint at /esat/audioslave/qmeeus/exp/whisper_slu/train/whisper-small-spoken-ner.
|
243 |
+
If your task is similar to the task the model of the checkpoint was trained on, you can already use WhisperSLU for predictions without further training.
|
244 |
+
[INFO|configuration_utils.py:779] 2024-01-08 18:44:42,795 >> loading configuration file /esat/audioslave/qmeeus/exp/whisper_slu/train/whisper-small-spoken-ner/generation_config.json
|
245 |
+
[INFO|configuration_utils.py:826] 2024-01-08 18:44:42,796 >> Generate config GenerationConfig {
|
246 |
+
"alignment_heads": [
|
247 |
+
[
|
248 |
+
5,
|
249 |
+
3
|
250 |
+
],
|
251 |
+
[
|
252 |
+
5,
|
253 |
+
9
|
254 |
+
],
|
255 |
+
[
|
256 |
+
8,
|
257 |
+
0
|
258 |
+
],
|
259 |
+
[
|
260 |
+
8,
|
261 |
+
4
|
262 |
+
],
|
263 |
+
[
|
264 |
+
8,
|
265 |
+
7
|
266 |
+
],
|
267 |
+
[
|
268 |
+
8,
|
269 |
+
8
|
270 |
+
],
|
271 |
+
[
|
272 |
+
9,
|
273 |
+
0
|
274 |
+
],
|
275 |
+
[
|
276 |
+
9,
|
277 |
+
7
|
278 |
+
],
|
279 |
+
[
|
280 |
+
9,
|
281 |
+
9
|
282 |
+
],
|
283 |
+
[
|
284 |
+
10,
|
285 |
+
5
|
286 |
+
]
|
287 |
+
],
|
288 |
+
"begin_suppress_tokens": [
|
289 |
+
220,
|
290 |
+
50257
|
291 |
+
],
|
292 |
+
"bos_token_id": 50257,
|
293 |
+
"decoder_start_token_id": 50258,
|
294 |
+
"eos_token_id": 50257,
|
295 |
+
"forced_decoder_ids": [
|
296 |
+
[
|
297 |
+
1,
|
298 |
+
null
|
299 |
+
],
|
300 |
+
[
|
301 |
+
2,
|
302 |
+
50359
|
303 |
+
]
|
304 |
+
],
|
305 |
+
"is_multilingual": true,
|
306 |
+
"lang_to_id": {
|
307 |
+
"<|af|>": 50327,
|
308 |
+
"<|am|>": 50334,
|
309 |
+
"<|ar|>": 50272,
|
310 |
+
"<|as|>": 50350,
|
311 |
+
"<|az|>": 50304,
|
312 |
+
"<|ba|>": 50355,
|
313 |
+
"<|be|>": 50330,
|
314 |
+
"<|bg|>": 50292,
|
315 |
+
"<|bn|>": 50302,
|
316 |
+
"<|bo|>": 50347,
|
317 |
+
"<|br|>": 50309,
|
318 |
+
"<|bs|>": 50315,
|
319 |
+
"<|ca|>": 50270,
|
320 |
+
"<|cs|>": 50283,
|
321 |
+
"<|cy|>": 50297,
|
322 |
+
"<|da|>": 50285,
|
323 |
+
"<|de|>": 50261,
|
324 |
+
"<|el|>": 50281,
|
325 |
+
"<|en|>": 50259,
|
326 |
+
"<|es|>": 50262,
|
327 |
+
"<|et|>": 50307,
|
328 |
+
"<|eu|>": 50310,
|
329 |
+
"<|fa|>": 50300,
|
330 |
+
"<|fi|>": 50277,
|
331 |
+
"<|fo|>": 50338,
|
332 |
+
"<|fr|>": 50265,
|
333 |
+
"<|gl|>": 50319,
|
334 |
+
"<|gu|>": 50333,
|
335 |
+
"<|haw|>": 50352,
|
336 |
+
"<|ha|>": 50354,
|
337 |
+
"<|he|>": 50279,
|
338 |
+
"<|hi|>": 50276,
|
339 |
+
"<|hr|>": 50291,
|
340 |
+
"<|ht|>": 50339,
|
341 |
+
"<|hu|>": 50286,
|
342 |
+
"<|hy|>": 50312,
|
343 |
+
"<|id|>": 50275,
|
344 |
+
"<|is|>": 50311,
|
345 |
+
"<|it|>": 50274,
|
346 |
+
"<|ja|>": 50266,
|
347 |
+
"<|jw|>": 50356,
|
348 |
+
"<|ka|>": 50329,
|
349 |
+
"<|kk|>": 50316,
|
350 |
+
"<|km|>": 50323,
|
351 |
+
"<|kn|>": 50306,
|
352 |
+
"<|ko|>": 50264,
|
353 |
+
"<|la|>": 50294,
|
354 |
+
"<|lb|>": 50345,
|
355 |
+
"<|ln|>": 50353,
|
356 |
+
"<|lo|>": 50336,
|
357 |
+
"<|lt|>": 50293,
|
358 |
+
"<|lv|>": 50301,
|
359 |
+
"<|mg|>": 50349,
|
360 |
+
"<|mi|>": 50295,
|
361 |
+
"<|mk|>": 50308,
|
362 |
+
"<|ml|>": 50296,
|
363 |
+
"<|mn|>": 50314,
|
364 |
+
"<|mr|>": 50320,
|
365 |
+
"<|ms|>": 50282,
|
366 |
+
"<|mt|>": 50343,
|
367 |
+
"<|my|>": 50346,
|
368 |
+
"<|ne|>": 50313,
|
369 |
+
"<|nl|>": 50271,
|
370 |
+
"<|nn|>": 50342,
|
371 |
+
"<|no|>": 50288,
|
372 |
+
"<|oc|>": 50328,
|
373 |
+
"<|pa|>": 50321,
|
374 |
+
"<|pl|>": 50269,
|
375 |
+
"<|ps|>": 50340,
|
376 |
+
"<|pt|>": 50267,
|
377 |
+
"<|ro|>": 50284,
|
378 |
+
"<|ru|>": 50263,
|
379 |
+
"<|sa|>": 50344,
|
380 |
+
"<|sd|>": 50332,
|
381 |
+
"<|si|>": 50322,
|
382 |
+
"<|sk|>": 50298,
|
383 |
+
"<|sl|>": 50305,
|
384 |
+
"<|sn|>": 50324,
|
385 |
+
"<|so|>": 50326,
|
386 |
+
"<|sq|>": 50317,
|
387 |
+
"<|sr|>": 50303,
|
388 |
+
"<|su|>": 50357,
|
389 |
+
"<|sv|>": 50273,
|
390 |
+
"<|sw|>": 50318,
|
391 |
+
"<|ta|>": 50287,
|
392 |
+
"<|te|>": 50299,
|
393 |
+
"<|tg|>": 50331,
|
394 |
+
"<|th|>": 50289,
|
395 |
+
"<|tk|>": 50341,
|
396 |
+
"<|tl|>": 50348,
|
397 |
+
"<|tr|>": 50268,
|
398 |
+
"<|tt|>": 50351,
|
399 |
+
"<|uk|>": 50280,
|
400 |
+
"<|ur|>": 50290,
|
401 |
+
"<|uz|>": 50337,
|
402 |
+
"<|vi|>": 50278,
|
403 |
+
"<|yi|>": 50335,
|
404 |
+
"<|yo|>": 50325,
|
405 |
+
"<|zh|>": 50260
|
406 |
+
},
|
407 |
+
"max_initial_timestamp_index": 1,
|
408 |
+
"max_length": 448,
|
409 |
+
"no_timestamps_token_id": 50363,
|
410 |
+
"pad_token_id": 50257,
|
411 |
+
"return_timestamps": false,
|
412 |
+
"suppress_tokens": [
|
413 |
+
1,
|
414 |
+
2,
|
415 |
+
7,
|
416 |
+
8,
|
417 |
+
9,
|
418 |
+
10,
|
419 |
+
14,
|
420 |
+
25,
|
421 |
+
26,
|
422 |
+
27,
|
423 |
+
28,
|
424 |
+
29,
|
425 |
+
31,
|
426 |
+
58,
|
427 |
+
59,
|
428 |
+
60,
|
429 |
+
61,
|
430 |
+
62,
|
431 |
+
63,
|
432 |
+
90,
|
433 |
+
91,
|
434 |
+
92,
|
435 |
+
93,
|
436 |
+
359,
|
437 |
+
503,
|
438 |
+
522,
|
439 |
+
542,
|
440 |
+
873,
|
441 |
+
893,
|
442 |
+
902,
|
443 |
+
918,
|
444 |
+
922,
|
445 |
+
931,
|
446 |
+
1350,
|
447 |
+
1853,
|
448 |
+
1982,
|
449 |
+
2460,
|
450 |
+
2627,
|
451 |
+
3246,
|
452 |
+
3253,
|
453 |
+
3268,
|
454 |
+
3536,
|
455 |
+
3846,
|
456 |
+
3961,
|
457 |
+
4183,
|
458 |
+
4667,
|
459 |
+
6585,
|
460 |
+
6647,
|
461 |
+
7273,
|
462 |
+
9061,
|
463 |
+
9383,
|
464 |
+
10428,
|
465 |
+
10929,
|
466 |
+
11938,
|
467 |
+
12033,
|
468 |
+
12331,
|
469 |
+
12562,
|
470 |
+
13793,
|
471 |
+
14157,
|
472 |
+
14635,
|
473 |
+
15265,
|
474 |
+
15618,
|
475 |
+
16553,
|
476 |
+
16604,
|
477 |
+
18362,
|
478 |
+
18956,
|
479 |
+
20075,
|
480 |
+
21675,
|
481 |
+
22520,
|
482 |
+
26130,
|
483 |
+
26161,
|
484 |
+
26435,
|
485 |
+
28279,
|
486 |
+
29464,
|
487 |
+
31650,
|
488 |
+
32302,
|
489 |
+
32470,
|
490 |
+
36865,
|
491 |
+
42863,
|
492 |
+
47425,
|
493 |
+
49870,
|
494 |
+
50254,
|
495 |
+
50258,
|
496 |
+
50358,
|
497 |
+
50359,
|
498 |
+
50360,
|
499 |
+
50361,
|
500 |
+
50362
|
501 |
+
],
|
502 |
+
"task_to_id": {
|
503 |
+
"transcribe": 50359,
|
504 |
+
"translate": 50358
|
505 |
+
}
|
506 |
+
}
|
507 |
+
|
508 |
+
trainable params: 2,111,784 || all params: 255,250,145 || trainable%: 0.8273390011198622
|
509 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-08 18:44:47,327 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/preprocessor_config.json
|
510 |
+
[INFO|tokenization_utils_base.py:2432] 2024-01-08 18:44:47,357 >> tokenizer config file saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tokenizer_config.json
|
511 |
+
[INFO|tokenization_utils_base.py:2441] 2024-01-08 18:44:47,358 >> Special tokens file saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/special_tokens_map.json
|
512 |
+
[INFO|configuration_utils.py:483] 2024-01-08 18:44:47,419 >> Configuration saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/config.json
|
513 |
+
[INFO|trainer.py:522] 2024-01-08 18:44:50,691 >> max_steps is given, it will override any value given in num_train_epochs
|
514 |
+
[INFO|trainer.py:571] 2024-01-08 18:44:50,691 >> Using auto half precision backend
|
515 |
+
wandb: Currently logged in as: qmeeus. Use `wandb login --relogin` to force relogin
|
516 |
+
wandb: wandb version 0.16.1 is available! To upgrade, please run:
|
517 |
+
wandb: $ pip install wandb --upgrade
|
518 |
+
wandb: Tracking run with wandb version 0.15.12
|
519 |
+
wandb: Run data is saved locally in /usr/data/condor/execute/dir_314523/whisper_slu/wandb/run-20240108_184452-35ireexg
|
520 |
+
wandb: Run `wandb offline` to turn off syncing.
|
521 |
+
wandb: Syncing run run-2024-01-08_18-44-50
|
522 |
+
wandb: ⭐️ View project at https://wandb.ai/qmeeus/Whisper%20PEFT%20Fine-Tuning
|
523 |
+
wandb: 🚀 View run at https://wandb.ai/qmeeus/Whisper%20PEFT%20Fine-Tuning/runs/35ireexg
|
524 |
+
[INFO|trainer.py:718] 2024-01-08 18:44:53,398 >> The following columns in the training set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
525 |
+
[INFO|trainer.py:1712] 2024-01-08 18:44:53,456 >> ***** Running training *****
|
526 |
+
[INFO|trainer.py:1713] 2024-01-08 18:44:53,456 >> Num examples = 71,615
|
527 |
+
[INFO|trainer.py:1714] 2024-01-08 18:44:53,456 >> Num Epochs = 9
|
528 |
+
[INFO|trainer.py:1715] 2024-01-08 18:44:53,456 >> Instantaneous batch size per device = 4
|
529 |
+
[INFO|trainer.py:1718] 2024-01-08 18:44:53,456 >> Total train batch size (w. parallel, distributed & accumulation) = 128
|
530 |
+
[INFO|trainer.py:1719] 2024-01-08 18:44:53,456 >> Gradient Accumulation steps = 32
|
531 |
+
[INFO|trainer.py:1720] 2024-01-08 18:44:53,456 >> Total optimization steps = 5,000
|
532 |
+
[INFO|trainer.py:1721] 2024-01-08 18:44:53,459 >> Number of trainable parameters = 2,111,784
|
533 |
+
[INFO|integration_utils.py:722] 2024-01-08 18:44:53,462 >> Automatic Weights & Biases logging enabled, to disable set os.environ["WANDB_DISABLED"] = "true"
|
534 |
+
[WARNING|logging.py:314] 2024-01-08 18:44:53,481 >> You're using a WhisperTokenizerFast tokenizer. Please note that with a fast tokenizer, using the `__call__` method is faster than using a method to encode the text followed by a call to the `pad` method to get a padded encoding.
|
535 |
+
[INFO|trainer.py:718] 2024-01-08 19:13:47,119 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
536 |
+
[INFO|trainer.py:2895] 2024-01-08 19:19:31,366 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-200
|
537 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-08 19:19:31,494 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-200/preprocessor_config.json
|
538 |
+
[INFO|trainer.py:718] 2024-01-08 19:48:40,699 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
539 |
+
[INFO|trainer.py:2895] 2024-01-08 19:54:22,629 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-400
|
540 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-08 19:54:22,697 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-400/preprocessor_config.json
|
541 |
+
[INFO|trainer.py:718] 2024-01-08 20:23:52,708 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
542 |
+
[INFO|trainer.py:2895] 2024-01-08 20:29:34,863 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-600
|
543 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-08 20:29:34,923 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-600/preprocessor_config.json
|
544 |
+
[INFO|trainer.py:718] 2024-01-08 20:57:32,183 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
545 |
+
[INFO|trainer.py:2895] 2024-01-08 21:03:16,687 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-800
|
546 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-08 21:03:16,748 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-800/preprocessor_config.json
|
547 |
+
[INFO|trainer.py:718] 2024-01-08 21:31:12,469 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
548 |
+
[INFO|trainer.py:2895] 2024-01-08 21:36:50,658 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-1000
|
549 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-08 21:36:50,723 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-1000/preprocessor_config.json
|
550 |
+
[INFO|trainer.py:718] 2024-01-08 22:04:44,620 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
551 |
+
[INFO|trainer.py:2895] 2024-01-08 22:10:25,435 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-1200
|
552 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-08 22:10:25,503 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-1200/preprocessor_config.json
|
553 |
+
[INFO|trainer.py:718] 2024-01-08 22:38:14,532 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
554 |
+
[INFO|trainer.py:2895] 2024-01-08 22:43:55,646 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-1400
|
555 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-08 22:43:55,713 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-1400/preprocessor_config.json
|
556 |
+
[INFO|trainer.py:718] 2024-01-08 23:11:49,094 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
557 |
+
[INFO|trainer.py:2895] 2024-01-08 23:17:29,789 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-1600
|
558 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-08 23:17:29,855 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-1600/preprocessor_config.json
|
559 |
+
[INFO|trainer.py:718] 2024-01-08 23:45:21,350 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
560 |
+
[INFO|trainer.py:2895] 2024-01-08 23:50:59,797 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-1800
|
561 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-08 23:50:59,864 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-1800/preprocessor_config.json
|
562 |
+
[INFO|trainer.py:718] 2024-01-09 00:18:55,674 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
563 |
+
[INFO|trainer.py:2895] 2024-01-09 00:24:38,854 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-2000
|
564 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 00:24:38,925 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-2000/preprocessor_config.json
|
565 |
+
[INFO|trainer.py:718] 2024-01-09 00:52:30,504 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
566 |
+
[INFO|trainer.py:2895] 2024-01-09 00:58:08,825 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-2200
|
567 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 00:58:08,891 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-2200/preprocessor_config.json
|
568 |
+
[INFO|trainer.py:718] 2024-01-09 01:26:03,365 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
569 |
+
[INFO|trainer.py:2895] 2024-01-09 01:31:41,568 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-2400
|
570 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 01:31:41,637 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-2400/preprocessor_config.json
|
571 |
+
[INFO|trainer.py:718] 2024-01-09 01:59:37,802 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
572 |
+
[INFO|trainer.py:2895] 2024-01-09 02:05:15,416 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-2600
|
573 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 02:05:15,487 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-2600/preprocessor_config.json
|
574 |
+
[INFO|trainer.py:718] 2024-01-09 02:33:13,316 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
575 |
+
[INFO|trainer.py:2895] 2024-01-09 02:38:52,241 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-2800
|
576 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 02:38:52,309 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-2800/preprocessor_config.json
|
577 |
+
[INFO|trainer.py:718] 2024-01-09 03:06:54,838 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
578 |
+
[INFO|trainer.py:2895] 2024-01-09 03:12:32,446 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-3000
|
579 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 03:12:32,518 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-3000/preprocessor_config.json
|
580 |
+
[INFO|trainer.py:718] 2024-01-09 03:40:35,202 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
581 |
+
[INFO|trainer.py:2895] 2024-01-09 03:46:14,094 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-3200
|
582 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 03:46:14,164 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-3200/preprocessor_config.json
|
583 |
+
[INFO|trainer.py:718] 2024-01-09 04:14:09,998 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
584 |
+
[INFO|trainer.py:2895] 2024-01-09 04:19:47,911 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-3400
|
585 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 04:19:47,978 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-3400/preprocessor_config.json
|
586 |
+
[INFO|trainer.py:718] 2024-01-09 04:47:50,188 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
587 |
+
[INFO|trainer.py:2895] 2024-01-09 04:53:29,921 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-3600
|
588 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 04:53:29,988 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-3600/preprocessor_config.json
|
589 |
+
[INFO|trainer.py:718] 2024-01-09 05:21:33,159 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
590 |
+
[INFO|trainer.py:2895] 2024-01-09 05:27:11,558 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-3800
|
591 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 05:27:11,628 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-3800/preprocessor_config.json
|
592 |
+
[INFO|trainer.py:718] 2024-01-09 05:55:12,769 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
593 |
+
[INFO|trainer.py:2895] 2024-01-09 06:00:50,862 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-4000
|
594 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 06:00:50,923 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-4000/preprocessor_config.json
|
595 |
+
[INFO|trainer.py:718] 2024-01-09 06:28:50,219 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
596 |
+
[INFO|trainer.py:2895] 2024-01-09 06:34:27,483 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-4200
|
597 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 06:34:27,548 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-4200/preprocessor_config.json
|
598 |
+
[INFO|trainer.py:718] 2024-01-09 07:02:24,846 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
599 |
+
[INFO|trainer.py:2895] 2024-01-09 07:08:04,451 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-4400
|
600 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 07:08:04,518 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-4400/preprocessor_config.json
|
601 |
+
[INFO|trainer.py:718] 2024-01-09 07:36:02,929 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
602 |
+
[INFO|trainer.py:2895] 2024-01-09 07:41:42,554 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-4600
|
603 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 07:41:42,623 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-4600/preprocessor_config.json
|
604 |
+
[INFO|trainer.py:718] 2024-01-09 08:09:40,000 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
605 |
+
[INFO|trainer.py:2895] 2024-01-09 08:15:17,334 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-4800
|
606 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 08:15:17,402 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-4800/preprocessor_config.json
|
607 |
+
[INFO|trainer.py:718] 2024-01-09 08:43:18,208 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
608 |
+
[INFO|trainer.py:2895] 2024-01-09 08:48:55,880 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-5000
|
609 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 08:48:55,951 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/tmp-checkpoint-5000/preprocessor_config.json
|
610 |
+
[INFO|trainer.py:1953] 2024-01-09 08:48:56,055 >>
|
611 |
+
|
612 |
+
Training completed. Do not forget to share your model on huggingface.co/models =)
|
613 |
+
|
614 |
+
|
615 |
+
[INFO|trainer.py:2895] 2024-01-09 08:48:56,060 >> Saving model checkpoint to /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora
|
616 |
+
[INFO|feature_extraction_utils.py:425] 2024-01-09 08:48:56,146 >> Feature extractor saved in /esat/audioslave/qmeeus/exp/whisper_slu/pipeline/whisper-small-spoken-ner-lora/preprocessor_config.json
|
617 |
+
[INFO|trainer.py:718] 2024-01-09 08:48:56,152 >> The following columns in the evaluation set don't have a corresponding argument in `PeftModel.forward` and have been ignored: input_length. If input_length are not expected by `PeftModel.forward`, you can safely ignore this message.
|
618 |
+
wandb: Waiting for W&B process to finish... (success).
|
619 |
+
wandb:
|
620 |
+
wandb: Run history:
|
621 |
+
wandb: eval/f1_score ▁▁▄▅▄▅▅▅▅▆▆▆▇▇▇▇▆██▇▇█████
|
622 |
+
wandb: eval/label_f1 ▆▆▄▃▁▅▄▂▄█▆▅▆▅▆▄▄▅▆▆▆▆▃▄▄▄
|
623 |
+
wandb: eval/loss █▅▄▃▃▂▂▂▂▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
|
624 |
+
wandb: eval/runtime █▆▆█▂▄▅▄▂▇▂▂▁▃▁▃▂▃▂▂▁▃▃▁▁▃
|
625 |
+
wandb: eval/samples_per_second ▁▃▃▁▇▅▄▅▇▂▇▇█▆█▆▇▆▇▇█▆▆██▆
|
626 |
+
wandb: eval/steps_per_second ▁▄▃▁▇▅▄▅▆▂▆▇▇▆▇▆▇▅▆▇█▅▅█▇▅
|
627 |
+
wandb: eval/wer █▅▆▅▂▂▂▃▁▃▁▁▁▃▃▂▃▃▃▃▃▃▃▃▃▃
|
628 |
+
wandb: train/epoch ▁▁▁▁▂▂▂▂▂▃▃▃▃▃▄▄▄▄▄▅▅▅▅▅▅▆▆▆▆▆▇▇▇▇▇▇████
|
629 |
+
wandb: train/global_step ▁▁▁▁▂▂▂▂▂▃▃▃▃▃▄▄▄▄▄▅▅▅▅▅▅▆▆▆▆▆▇▇▇▇▇▇████
|
630 |
+
wandb: train/learning_rate ████████▇▇▇▇▇▆▆▆▅▅▅▅▄▄▄▄▃▃▃▃▂▂▂▂▂▁▁▁▁▁▁▁
|
631 |
+
wandb: train/loss █▃▃▃▂▂▂▂▂▂▂▂▂▂▁▂▂▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
|
632 |
+
wandb: train/total_flos ▁
|
633 |
+
wandb: train/train_loss ▁
|
634 |
+
wandb: train/train_runtime ▁
|
635 |
+
wandb: train/train_samples_per_second ▁
|
636 |
+
wandb: train/train_steps_per_second ▁
|
637 |
+
wandb:
|
638 |
+
wandb: Run summary:
|
639 |
+
wandb: eval/f1_score 0.6872
|
640 |
+
wandb: eval/label_f1 0.83254
|
641 |
+
wandb: eval/loss 0.22641
|
642 |
+
wandb: eval/runtime 339.3736
|
643 |
+
wandb: eval/samples_per_second 2.947
|
644 |
+
wandb: eval/steps_per_second 0.368
|
645 |
+
wandb: eval/wer 0.098
|
646 |
+
wandb: train/epoch 8.94
|
647 |
+
wandb: train/global_step 5000
|
648 |
+
wandb: train/learning_rate 0.0
|
649 |
+
wandb: train/loss 0.1961
|
650 |
+
wandb: train/total_flos 1.9683074514013055e+20
|
651 |
+
wandb: train/train_loss 0.21677
|
652 |
+
wandb: train/train_runtime 50642.5955
|
653 |
+
wandb: train/train_samples_per_second 12.638
|
654 |
+
wandb: train/train_steps_per_second 0.099
|
655 |
+
wandb:
|
656 |
+
wandb: 🚀 View run run-2024-01-08_18-44-50 at: https://wandb.ai/qmeeus/Whisper%20PEFT%20Fine-Tuning/runs/35ireexg
|
657 |
+
wandb: ️⚡ View job at https://wandb.ai/qmeeus/Whisper%20PEFT%20Fine-Tuning/jobs/QXJ0aWZhY3RDb2xsZWN0aW9uOjEyODM1Nzc0OA==/version_details/v2
|
658 |
+
wandb: Synced 5 W&B file(s), 0 media file(s), 2 artifact file(s) and 0 other file(s)
|
659 |
+
wandb: Find logs at: ./wandb/run-20240108_184452-35ireexg/logs
|
logs/whisper-spoken-ner-small-pipe-lora.job
ADDED
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
Universe = vanilla
|
2 |
+
|
3 |
+
NiceUser = True
|
4 |
+
RequestCpus = 8
|
5 |
+
RequestMemory = 64G
|
6 |
+
RequestDisk = 100G
|
7 |
+
RequestWallTime = 255600
|
8 |
+
RequestGPUs = 1
|
9 |
+
Requirements = (GPUs_GlobalMemoryMB >= 17000) && (GPUs_GlobalMemoryMB <= 30000) && ((machine != "spchcl21.esat.kuleuven.be"))
|
10 |
+
|
11 |
+
ShouldTransferFiles = yes
|
12 |
+
TransferInputFiles = scp://audioslave/usr/data/qmeeus/repos/peft/examples/whisper_slu,scp://audioslave/usr/data/qmeeus/repos/spoken-ner/data
|
13 |
+
|
14 |
+
Initialdir =
|
15 |
+
Executable = scripts/entrypoint.sh
|
16 |
+
Arguments = "scripts/pipeline/peft/run_pipe_spoken_ner_small_peft.sh"
|
17 |
+
Environment = "LOGDIR=logs RUN_NAME=whisper-spoken-ner-small-pipe-lora LOGLEVEL=INFO OUTDIR=/esat/audioslave/qmeeus/exp/whisper_slu"
|
18 |
+
|
19 |
+
Notification = Complete
|
20 |
+
Log = /users/spraak/qmeeus/condor_logs/condor-umber.log
|
21 |
+
Output = logs/whisper-spoken-ner-small-pipe-lora.out
|
22 |
+
Error = logs/whisper-spoken-ner-small-pipe-lora.err
|
23 |
+
|
24 |
+
Queue 1
|
logs/whisper-spoken-ner-small-pipe-lora.out
ADDED
@@ -0,0 +1,393 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
/usr/data/condor/execute/dir_314523/whisper_slu
|
2 |
+
PeftModel(
|
3 |
+
(base_model): LoraModel(
|
4 |
+
(model): WhisperSLU(
|
5 |
+
(model): WhisperModel(
|
6 |
+
(encoder): WhisperEncoder(
|
7 |
+
(conv1): Conv1d(80, 768, kernel_size=(3,), stride=(1,), padding=(1,))
|
8 |
+
(conv2): Conv1d(768, 768, kernel_size=(3,), stride=(2,), padding=(1,))
|
9 |
+
(embed_positions): Embedding(1500, 768)
|
10 |
+
(layers): ModuleList(
|
11 |
+
(0-11): 12 x WhisperEncoderLayer(
|
12 |
+
(self_attn): WhisperAttention(
|
13 |
+
(k_proj): Linear(in_features=768, out_features=768, bias=False)
|
14 |
+
(v_proj): Linear(in_features=768, out_features=768, bias=True)
|
15 |
+
(q_proj): Linear(in_features=768, out_features=768, bias=True)
|
16 |
+
(out_proj): Linear(in_features=768, out_features=768, bias=True)
|
17 |
+
)
|
18 |
+
(self_attn_layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
|
19 |
+
(activation_fn): GELUActivation()
|
20 |
+
(fc1): Linear(in_features=768, out_features=3072, bias=True)
|
21 |
+
(fc2): Linear(in_features=3072, out_features=768, bias=True)
|
22 |
+
(final_layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
|
23 |
+
)
|
24 |
+
)
|
25 |
+
(layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
|
26 |
+
)
|
27 |
+
(decoder): WhisperDecoder(
|
28 |
+
(embed_tokens): Embedding(51865, 768, padding_idx=50257)
|
29 |
+
(embed_positions): WhisperPositionalEmbedding(448, 768)
|
30 |
+
(layers): ModuleList(
|
31 |
+
(0-11): 12 x WhisperDecoderLayer(
|
32 |
+
(self_attn): WhisperAttention(
|
33 |
+
(k_proj): lora.Linear(
|
34 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=False)
|
35 |
+
(lora_dropout): ModuleDict(
|
36 |
+
(default): Dropout(p=0.1, inplace=False)
|
37 |
+
)
|
38 |
+
(lora_A): ModuleDict(
|
39 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
40 |
+
)
|
41 |
+
(lora_B): ModuleDict(
|
42 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
43 |
+
)
|
44 |
+
(lora_embedding_A): ParameterDict()
|
45 |
+
(lora_embedding_B): ParameterDict()
|
46 |
+
)
|
47 |
+
(v_proj): lora.Linear(
|
48 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=True)
|
49 |
+
(lora_dropout): ModuleDict(
|
50 |
+
(default): Dropout(p=0.1, inplace=False)
|
51 |
+
)
|
52 |
+
(lora_A): ModuleDict(
|
53 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
54 |
+
)
|
55 |
+
(lora_B): ModuleDict(
|
56 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
57 |
+
)
|
58 |
+
(lora_embedding_A): ParameterDict()
|
59 |
+
(lora_embedding_B): ParameterDict()
|
60 |
+
)
|
61 |
+
(q_proj): lora.Linear(
|
62 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=True)
|
63 |
+
(lora_dropout): ModuleDict(
|
64 |
+
(default): Dropout(p=0.1, inplace=False)
|
65 |
+
)
|
66 |
+
(lora_A): ModuleDict(
|
67 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
68 |
+
)
|
69 |
+
(lora_B): ModuleDict(
|
70 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
71 |
+
)
|
72 |
+
(lora_embedding_A): ParameterDict()
|
73 |
+
(lora_embedding_B): ParameterDict()
|
74 |
+
)
|
75 |
+
(out_proj): lora.Linear(
|
76 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=True)
|
77 |
+
(lora_dropout): ModuleDict(
|
78 |
+
(default): Dropout(p=0.1, inplace=False)
|
79 |
+
)
|
80 |
+
(lora_A): ModuleDict(
|
81 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
82 |
+
)
|
83 |
+
(lora_B): ModuleDict(
|
84 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
85 |
+
)
|
86 |
+
(lora_embedding_A): ParameterDict()
|
87 |
+
(lora_embedding_B): ParameterDict()
|
88 |
+
)
|
89 |
+
)
|
90 |
+
(activation_fn): GELUActivation()
|
91 |
+
(self_attn_layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
|
92 |
+
(encoder_attn): WhisperAttention(
|
93 |
+
(k_proj): lora.Linear(
|
94 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=False)
|
95 |
+
(lora_dropout): ModuleDict(
|
96 |
+
(default): Dropout(p=0.1, inplace=False)
|
97 |
+
)
|
98 |
+
(lora_A): ModuleDict(
|
99 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
100 |
+
)
|
101 |
+
(lora_B): ModuleDict(
|
102 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
103 |
+
)
|
104 |
+
(lora_embedding_A): ParameterDict()
|
105 |
+
(lora_embedding_B): ParameterDict()
|
106 |
+
)
|
107 |
+
(v_proj): lora.Linear(
|
108 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=True)
|
109 |
+
(lora_dropout): ModuleDict(
|
110 |
+
(default): Dropout(p=0.1, inplace=False)
|
111 |
+
)
|
112 |
+
(lora_A): ModuleDict(
|
113 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
114 |
+
)
|
115 |
+
(lora_B): ModuleDict(
|
116 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
117 |
+
)
|
118 |
+
(lora_embedding_A): ParameterDict()
|
119 |
+
(lora_embedding_B): ParameterDict()
|
120 |
+
)
|
121 |
+
(q_proj): lora.Linear(
|
122 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=True)
|
123 |
+
(lora_dropout): ModuleDict(
|
124 |
+
(default): Dropout(p=0.1, inplace=False)
|
125 |
+
)
|
126 |
+
(lora_A): ModuleDict(
|
127 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
128 |
+
)
|
129 |
+
(lora_B): ModuleDict(
|
130 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
131 |
+
)
|
132 |
+
(lora_embedding_A): ParameterDict()
|
133 |
+
(lora_embedding_B): ParameterDict()
|
134 |
+
)
|
135 |
+
(out_proj): lora.Linear(
|
136 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=True)
|
137 |
+
(lora_dropout): ModuleDict(
|
138 |
+
(default): Dropout(p=0.1, inplace=False)
|
139 |
+
)
|
140 |
+
(lora_A): ModuleDict(
|
141 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
142 |
+
)
|
143 |
+
(lora_B): ModuleDict(
|
144 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
145 |
+
)
|
146 |
+
(lora_embedding_A): ParameterDict()
|
147 |
+
(lora_embedding_B): ParameterDict()
|
148 |
+
)
|
149 |
+
)
|
150 |
+
(encoder_attn_layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
|
151 |
+
(fc1): lora.Linear(
|
152 |
+
(base_layer): Linear(in_features=768, out_features=3072, bias=True)
|
153 |
+
(lora_dropout): ModuleDict(
|
154 |
+
(default): Dropout(p=0.1, inplace=False)
|
155 |
+
)
|
156 |
+
(lora_A): ModuleDict(
|
157 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
158 |
+
)
|
159 |
+
(lora_B): ModuleDict(
|
160 |
+
(default): Linear(in_features=8, out_features=3072, bias=False)
|
161 |
+
)
|
162 |
+
(lora_embedding_A): ParameterDict()
|
163 |
+
(lora_embedding_B): ParameterDict()
|
164 |
+
)
|
165 |
+
(fc2): lora.Linear(
|
166 |
+
(base_layer): Linear(in_features=3072, out_features=768, bias=True)
|
167 |
+
(lora_dropout): ModuleDict(
|
168 |
+
(default): Dropout(p=0.1, inplace=False)
|
169 |
+
)
|
170 |
+
(lora_A): ModuleDict(
|
171 |
+
(default): Linear(in_features=3072, out_features=8, bias=False)
|
172 |
+
)
|
173 |
+
(lora_B): ModuleDict(
|
174 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
175 |
+
)
|
176 |
+
(lora_embedding_A): ParameterDict()
|
177 |
+
(lora_embedding_B): ParameterDict()
|
178 |
+
)
|
179 |
+
(final_layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
|
180 |
+
)
|
181 |
+
)
|
182 |
+
(layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
|
183 |
+
)
|
184 |
+
)
|
185 |
+
(proj_out): Linear(in_features=768, out_features=51865, bias=False)
|
186 |
+
(classifier): WhisperClassificationHead(
|
187 |
+
(embed_positions): WhisperPositionalEmbedding(448, 768)
|
188 |
+
(layers): ModuleList(
|
189 |
+
(0-1): 2 x WhisperEncoderLayer(
|
190 |
+
(self_attn): WhisperAttention(
|
191 |
+
(k_proj): lora.Linear(
|
192 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=False)
|
193 |
+
(lora_dropout): ModuleDict(
|
194 |
+
(default): Dropout(p=0.1, inplace=False)
|
195 |
+
)
|
196 |
+
(lora_A): ModuleDict(
|
197 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
198 |
+
)
|
199 |
+
(lora_B): ModuleDict(
|
200 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
201 |
+
)
|
202 |
+
(lora_embedding_A): ParameterDict()
|
203 |
+
(lora_embedding_B): ParameterDict()
|
204 |
+
)
|
205 |
+
(v_proj): lora.Linear(
|
206 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=True)
|
207 |
+
(lora_dropout): ModuleDict(
|
208 |
+
(default): Dropout(p=0.1, inplace=False)
|
209 |
+
)
|
210 |
+
(lora_A): ModuleDict(
|
211 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
212 |
+
)
|
213 |
+
(lora_B): ModuleDict(
|
214 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
215 |
+
)
|
216 |
+
(lora_embedding_A): ParameterDict()
|
217 |
+
(lora_embedding_B): ParameterDict()
|
218 |
+
)
|
219 |
+
(q_proj): lora.Linear(
|
220 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=True)
|
221 |
+
(lora_dropout): ModuleDict(
|
222 |
+
(default): Dropout(p=0.1, inplace=False)
|
223 |
+
)
|
224 |
+
(lora_A): ModuleDict(
|
225 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
226 |
+
)
|
227 |
+
(lora_B): ModuleDict(
|
228 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
229 |
+
)
|
230 |
+
(lora_embedding_A): ParameterDict()
|
231 |
+
(lora_embedding_B): ParameterDict()
|
232 |
+
)
|
233 |
+
(out_proj): lora.Linear(
|
234 |
+
(base_layer): Linear(in_features=768, out_features=768, bias=True)
|
235 |
+
(lora_dropout): ModuleDict(
|
236 |
+
(default): Dropout(p=0.1, inplace=False)
|
237 |
+
)
|
238 |
+
(lora_A): ModuleDict(
|
239 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
240 |
+
)
|
241 |
+
(lora_B): ModuleDict(
|
242 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
243 |
+
)
|
244 |
+
(lora_embedding_A): ParameterDict()
|
245 |
+
(lora_embedding_B): ParameterDict()
|
246 |
+
)
|
247 |
+
)
|
248 |
+
(self_attn_layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
|
249 |
+
(activation_fn): GELUActivation()
|
250 |
+
(fc1): lora.Linear(
|
251 |
+
(base_layer): Linear(in_features=768, out_features=2048, bias=True)
|
252 |
+
(lora_dropout): ModuleDict(
|
253 |
+
(default): Dropout(p=0.1, inplace=False)
|
254 |
+
)
|
255 |
+
(lora_A): ModuleDict(
|
256 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
257 |
+
)
|
258 |
+
(lora_B): ModuleDict(
|
259 |
+
(default): Linear(in_features=8, out_features=2048, bias=False)
|
260 |
+
)
|
261 |
+
(lora_embedding_A): ParameterDict()
|
262 |
+
(lora_embedding_B): ParameterDict()
|
263 |
+
)
|
264 |
+
(fc2): lora.Linear(
|
265 |
+
(base_layer): Linear(in_features=2048, out_features=768, bias=True)
|
266 |
+
(lora_dropout): ModuleDict(
|
267 |
+
(default): Dropout(p=0.1, inplace=False)
|
268 |
+
)
|
269 |
+
(lora_A): ModuleDict(
|
270 |
+
(default): Linear(in_features=2048, out_features=8, bias=False)
|
271 |
+
)
|
272 |
+
(lora_B): ModuleDict(
|
273 |
+
(default): Linear(in_features=8, out_features=768, bias=False)
|
274 |
+
)
|
275 |
+
(lora_embedding_A): ParameterDict()
|
276 |
+
(lora_embedding_B): ParameterDict()
|
277 |
+
)
|
278 |
+
(final_layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
|
279 |
+
)
|
280 |
+
)
|
281 |
+
(layer_norm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)
|
282 |
+
(out_proj): lora.Linear(
|
283 |
+
(base_layer): Linear(in_features=768, out_features=37, bias=True)
|
284 |
+
(lora_dropout): ModuleDict(
|
285 |
+
(default): Dropout(p=0.1, inplace=False)
|
286 |
+
)
|
287 |
+
(lora_A): ModuleDict(
|
288 |
+
(default): Linear(in_features=768, out_features=8, bias=False)
|
289 |
+
)
|
290 |
+
(lora_B): ModuleDict(
|
291 |
+
(default): Linear(in_features=8, out_features=37, bias=False)
|
292 |
+
)
|
293 |
+
(lora_embedding_A): ParameterDict()
|
294 |
+
(lora_embedding_B): ParameterDict()
|
295 |
+
)
|
296 |
+
(crf): ConditionalRandomField()
|
297 |
+
)
|
298 |
+
)
|
299 |
+
)
|
300 |
+
)
|
301 |
+
{'loss': 0.4292, 'learning_rate': 5e-05, 'epoch': 0.18}
|
302 |
+
{'loss': 0.2746, 'learning_rate': 4.994863481875841e-05, 'epoch': 0.36}
|
303 |
+
{'eval_loss': 0.26023727655410767, 'eval_f1_score': 0.6564825695260478, 'eval_label_f1': 0.8343125734430082, 'eval_wer': 0.10898676368139949, 'eval_runtime': 344.2386, 'eval_samples_per_second': 2.905, 'eval_steps_per_second': 0.363, 'epoch': 0.36}
|
304 |
+
{'loss': 0.2568, 'learning_rate': 4.979475034558115e-05, 'epoch': 0.54}
|
305 |
+
{'loss': 0.2481, 'learning_rate': 4.9538978924776634e-05, 'epoch': 0.71}
|
306 |
+
{'eval_loss': 0.246540829539299, 'eval_f1_score': 0.6577916992952232, 'eval_label_f1': 0.8347689898198903, 'eval_wer': 0.10217509095131203, 'eval_runtime': 341.9234, 'eval_samples_per_second': 2.925, 'eval_steps_per_second': 0.366, 'epoch': 0.71}
|
307 |
+
{'loss': 0.2412, 'learning_rate': 4.9182371575975736e-05, 'epoch': 0.89}
|
308 |
+
{'loss': 0.2385, 'learning_rate': 4.8726393675266716e-05, 'epoch': 1.07}
|
309 |
+
{'eval_loss': 0.24104812741279602, 'eval_f1_score': 0.6684952978056427, 'eval_label_f1': 0.8322884012539185, 'eval_wer': 0.10484557628299404, 'eval_runtime': 342.1463, 'eval_samples_per_second': 2.923, 'eval_steps_per_second': 0.365, 'epoch': 1.07}
|
310 |
+
{'loss': 0.2325, 'learning_rate': 4.817291893365055e-05, 'epoch': 1.25}
|
311 |
+
{'loss': 0.2316, 'learning_rate': 4.752422169756048e-05, 'epoch': 1.43}
|
312 |
+
{'eval_loss': 0.23740312457084656, 'eval_f1_score': 0.6724477729601892, 'eval_label_f1': 0.8316909735908553, 'eval_wer': 0.10221379363727842, 'eval_runtime': 344.4982, 'eval_samples_per_second': 2.903, 'eval_steps_per_second': 0.363, 'epoch': 1.43}
|
313 |
+
{'loss': 0.2304, 'learning_rate': 4.678296760308474e-05, 'epoch': 1.61}
|
314 |
+
{'loss': 0.2291, 'learning_rate': 4.595220262229601e-05, 'epoch': 1.79}
|
315 |
+
{'eval_loss': 0.2348490208387375, 'eval_f1_score': 0.6698076168040833, 'eval_label_f1': 0.8292108362779742, 'eval_wer': 0.09683412028794798, 'eval_runtime': 338.1818, 'eval_samples_per_second': 2.957, 'eval_steps_per_second': 0.37, 'epoch': 1.79}
|
316 |
+
{'loss': 0.2275, 'learning_rate': 4.503534054669892e-05, 'epoch': 1.97}
|
317 |
+
{'loss': 0.2205, 'learning_rate': 4.4036148959228365e-05, 'epoch': 2.14}
|
318 |
+
{'eval_loss': 0.2333754003047943, 'eval_f1_score': 0.6744822196170379, 'eval_label_f1': 0.8339194998046112, 'eval_wer': 0.09636968805635111, 'eval_runtime': 340.8077, 'eval_samples_per_second': 2.934, 'eval_steps_per_second': 0.367, 'epoch': 2.14}
|
319 |
+
{'loss': 0.2224, 'learning_rate': 4.2958733752443195e-05, 'epoch': 2.32}
|
320 |
+
{'loss': 0.2211, 'learning_rate': 4.180752225653292e-05, 'epoch': 2.5}
|
321 |
+
{'eval_loss': 0.2319139689207077, 'eval_f1_score': 0.672933803368586, 'eval_label_f1': 0.8327457892675283, 'eval_wer': 0.09606006656861986, 'eval_runtime': 341.1059, 'eval_samples_per_second': 2.932, 'eval_steps_per_second': 0.366, 'epoch': 2.5}
|
322 |
+
{'loss': 0.2196, 'learning_rate': 4.058724504646834e-05, 'epoch': 2.68}
|
323 |
+
{'loss': 0.2163, 'learning_rate': 3.9302916503054246e-05, 'epoch': 2.86}
|
324 |
+
{'eval_loss': 0.2304619699716568, 'eval_f1_score': 0.6731669266770671, 'eval_label_f1': 0.829953198127925, 'eval_wer': 0.09807260623887298, 'eval_runtime': 340.688, 'eval_samples_per_second': 2.935, 'eval_steps_per_second': 0.367, 'epoch': 2.86}
|
325 |
+
{'loss': 0.219, 'learning_rate': 3.7959814207763135e-05, 'epoch': 3.04}
|
326 |
+
{'loss': 0.2108, 'learning_rate': 3.656345725602089e-05, 'epoch': 3.22}
|
327 |
+
{'eval_loss': 0.22986993193626404, 'eval_f1_score': 0.6734375, 'eval_label_f1': 0.8328125000000001, 'eval_wer': 0.09544082359315736, 'eval_runtime': 338.4413, 'eval_samples_per_second': 2.955, 'eval_steps_per_second': 0.369, 'epoch': 3.22}
|
328 |
+
{'loss': 0.2132, 'learning_rate': 3.5119583578059846e-05, 'epoch': 3.4}
|
329 |
+
{'loss': 0.2104, 'learning_rate': 3.363412636053269e-05, 'epoch': 3.57}
|
330 |
+
{'eval_loss': 0.22974026203155518, 'eval_f1_score': 0.6792156862745098, 'eval_label_f1': 0.8368627450980393, 'eval_wer': 0.09919498413189876, 'eval_runtime': 343.1743, 'eval_samples_per_second': 2.914, 'eval_steps_per_second': 0.364, 'epoch': 3.57}
|
331 |
+
{'loss': 0.2128, 'learning_rate': 3.211318966577581e-05, 'epoch': 3.75}
|
332 |
+
{'loss': 0.2124, 'learning_rate': 3.056302334890786e-05, 'epoch': 3.93}
|
333 |
+
{'eval_loss': 0.2278670072555542, 'eval_f1_score': 0.6781925343811395, 'eval_label_f1': 0.8345776031434184, 'eval_wer': 0.09447325644399722, 'eval_runtime': 338.3142, 'eval_samples_per_second': 2.956, 'eval_steps_per_second': 0.369, 'epoch': 3.93}
|
334 |
+
{'loss': 0.2077, 'learning_rate': 2.8989997375834482e-05, 'epoch': 4.11}
|
335 |
+
{'loss': 0.2027, 'learning_rate': 2.7400575647692046e-05, 'epoch': 4.29}
|
336 |
+
{'eval_loss': 0.22793905436992645, 'eval_f1_score': 0.6789638932496076, 'eval_label_f1': 0.8335949764521193, 'eval_wer': 0.09443455375803081, 'eval_runtime': 338.197, 'eval_samples_per_second': 2.957, 'eval_steps_per_second': 0.37, 'epoch': 4.29}
|
337 |
+
{'loss': 0.2077, 'learning_rate': 2.5801289439291388e-05, 'epoch': 4.47}
|
338 |
+
{'loss': 0.2055, 'learning_rate': 2.419871056070862e-05, 'epoch': 4.65}
|
339 |
+
{'eval_loss': 0.2275087535381317, 'eval_f1_score': 0.6832347140039449, 'eval_label_f1': 0.8347140039447732, 'eval_wer': 0.09493768867559409, 'eval_runtime': 337.608, 'eval_samples_per_second': 2.962, 'eval_steps_per_second': 0.37, 'epoch': 4.65}
|
340 |
+
{'loss': 0.2073, 'learning_rate': 2.2599424352307957e-05, 'epoch': 4.83}
|
341 |
+
{'loss': 0.209, 'learning_rate': 2.1010002624165527e-05, 'epoch': 5.0}
|
342 |
+
{'eval_loss': 0.22685159742832184, 'eval_f1_score': 0.6821766561514195, 'eval_label_f1': 0.833596214511041, 'eval_wer': 0.09826611966870501, 'eval_runtime': 338.9186, 'eval_samples_per_second': 2.951, 'eval_steps_per_second': 0.369, 'epoch': 5.0}
|
343 |
+
{'loss': 0.2011, 'learning_rate': 1.9436976651092144e-05, 'epoch': 5.18}
|
344 |
+
{'loss': 0.2017, 'learning_rate': 1.7886810334224192e-05, 'epoch': 5.36}
|
345 |
+
{'eval_loss': 0.22723452746868134, 'eval_f1_score': 0.6834645669291338, 'eval_label_f1': 0.8346456692913385, 'eval_wer': 0.09791779549500736, 'eval_runtime': 337.6021, 'eval_samples_per_second': 2.962, 'eval_steps_per_second': 0.37, 'epoch': 5.36}
|
346 |
+
{'loss': 0.2034, 'learning_rate': 1.6365873639467315e-05, 'epoch': 5.54}
|
347 |
+
{'loss': 0.2029, 'learning_rate': 1.4880416421940155e-05, 'epoch': 5.72}
|
348 |
+
{'eval_loss': 0.22658005356788635, 'eval_f1_score': 0.6818718049547778, 'eval_label_f1': 0.8320880849390484, 'eval_wer': 0.09656320148618314, 'eval_runtime': 338.8852, 'eval_samples_per_second': 2.951, 'eval_steps_per_second': 0.369, 'epoch': 5.72}
|
349 |
+
{'loss': 0.202, 'learning_rate': 1.3436542743979125e-05, 'epoch': 5.9}
|
350 |
+
{'loss': 0.201, 'learning_rate': 1.2040185792236874e-05, 'epoch': 6.08}
|
351 |
+
{'eval_loss': 0.22656936943531036, 'eval_f1_score': 0.6800472255017709, 'eval_label_f1': 0.8327430145611965, 'eval_wer': 0.09776298475114173, 'eval_runtime': 337.9074, 'eval_samples_per_second': 2.959, 'eval_steps_per_second': 0.37, 'epoch': 6.08}
|
352 |
+
{'loss': 0.1979, 'learning_rate': 1.0697083496945765e-05, 'epoch': 6.26}
|
353 |
+
{'loss': 0.1985, 'learning_rate': 9.412754953531663e-06, 'epoch': 6.43}
|
354 |
+
{'eval_loss': 0.22672241926193237, 'eval_f1_score': 0.6856465005931198, 'eval_label_f1': 0.8335310399367339, 'eval_wer': 0.0994659029336636, 'eval_runtime': 339.7265, 'eval_samples_per_second': 2.944, 'eval_steps_per_second': 0.368, 'epoch': 6.43}
|
355 |
+
{'loss': 0.2006, 'learning_rate': 8.192477743467078e-06, 'epoch': 6.61}
|
356 |
+
{'loss': 0.1996, 'learning_rate': 7.041266247556813e-06, 'epoch': 6.79}
|
357 |
+
{'eval_loss': 0.226467102766037, 'eval_f1_score': 0.686437327006722, 'eval_label_f1': 0.8343218663503361, 'eval_wer': 0.0988079572722347, 'eval_runtime': 338.3929, 'eval_samples_per_second': 2.955, 'eval_steps_per_second': 0.369, 'epoch': 6.79}
|
358 |
+
{'loss': 0.2002, 'learning_rate': 5.9638510407716394e-06, 'epoch': 6.97}
|
359 |
+
{'loss': 0.197, 'learning_rate': 4.9646594533010875e-06, 'epoch': 7.15}
|
360 |
+
{'eval_loss': 0.2263847291469574, 'eval_f1_score': 0.6843354430379748, 'eval_label_f1': 0.8346518987341772, 'eval_wer': 0.09857574115643626, 'eval_runtime': 338.0877, 'eval_samples_per_second': 2.958, 'eval_steps_per_second': 0.37, 'epoch': 7.15}
|
361 |
+
{'loss': 0.1983, 'learning_rate': 4.047797377703985e-06, 'epoch': 7.33}
|
362 |
+
{'loss': 0.1985, 'learning_rate': 3.217032396915265e-06, 'epoch': 7.51}
|
363 |
+
{'eval_loss': 0.22642947733402252, 'eval_f1_score': 0.6837539432176655, 'eval_label_f1': 0.8351735015772871, 'eval_wer': 0.09849833578450344, 'eval_runtime': 337.2581, 'eval_samples_per_second': 2.965, 'eval_steps_per_second': 0.371, 'epoch': 7.51}
|
364 |
+
{'loss': 0.1967, 'learning_rate': 2.475778302439524e-06, 'epoch': 7.69}
|
365 |
+
{'loss': 0.1999, 'learning_rate': 1.827081066349459e-06, 'epoch': 7.86}
|
366 |
+
{'eval_loss': 0.2263396978378296, 'eval_f1_score': 0.6861429135412555, 'eval_label_f1': 0.8345834978286616, 'eval_wer': 0.09776298475114173, 'eval_runtime': 339.5995, 'eval_samples_per_second': 2.945, 'eval_steps_per_second': 0.368, 'epoch': 7.86}
|
367 |
+
{'loss': 0.1953, 'learning_rate': 1.273606324733284e-06, 'epoch': 8.04}
|
368 |
+
{'loss': 0.1963, 'learning_rate': 8.176284240242638e-07, 'epoch': 8.22}
|
369 |
+
{'eval_loss': 0.22643861174583435, 'eval_f1_score': 0.6864139020537124, 'eval_label_f1': 0.8317535545023697, 'eval_wer': 0.09784039012307454, 'eval_runtime': 339.619, 'eval_samples_per_second': 2.944, 'eval_steps_per_second': 0.368, 'epoch': 8.22}
|
370 |
+
{'loss': 0.1984, 'learning_rate': 4.6102107522336403e-07, 'epoch': 8.4}
|
371 |
+
{'loss': 0.1977, 'learning_rate': 2.052496544188487e-07, 'epoch': 8.58}
|
372 |
+
{'eval_loss': 0.22642208635807037, 'eval_f1_score': 0.6874753062030818, 'eval_label_f1': 0.8328723824575267, 'eval_wer': 0.09791779549500736, 'eval_runtime': 337.3275, 'eval_samples_per_second': 2.964, 'eval_steps_per_second': 0.371, 'epoch': 8.58}
|
373 |
+
{'loss': 0.1979, 'learning_rate': 5.136518124159162e-08, 'epoch': 8.76}
|
374 |
+
{'loss': 0.1961, 'learning_rate': 0.0, 'epoch': 8.94}
|
375 |
+
{'eval_loss': 0.22641009092330933, 'eval_f1_score': 0.6872037914691944, 'eval_label_f1': 0.8325434439178515, 'eval_wer': 0.09799520086694016, 'eval_runtime': 337.666, 'eval_samples_per_second': 2.962, 'eval_steps_per_second': 0.37, 'epoch': 8.94}
|
376 |
+
{'train_runtime': 50642.5955, 'train_samples_per_second': 12.638, 'train_steps_per_second': 0.099, 'train_loss': 0.21677429428100586, 'epoch': 8.94}
|
377 |
+
***** train metrics *****
|
378 |
+
epoch = 8.94
|
379 |
+
train_loss = 0.2168
|
380 |
+
train_runtime = 14:04:02.59
|
381 |
+
train_samples_per_second = 12.638
|
382 |
+
train_steps_per_second = 0.099
|
383 |
+
{'eval_loss': 0.22641009092330933, 'eval_f1_score': 0.6872037914691944, 'eval_label_f1': 0.8325434439178515, 'eval_wer': 0.09799520086694016, 'eval_runtime': 339.3736, 'eval_samples_per_second': 2.947, 'eval_steps_per_second': 0.368, 'epoch': 8.94}
|
384 |
+
***** eval metrics *****
|
385 |
+
epoch = 8.94
|
386 |
+
eval_f1_score = 0.6872
|
387 |
+
eval_label_f1 = 0.8325
|
388 |
+
eval_loss = 0.2264
|
389 |
+
eval_runtime = 0:05:39.37
|
390 |
+
eval_samples = 1000
|
391 |
+
eval_samples_per_second = 2.947
|
392 |
+
eval_steps_per_second = 0.368
|
393 |
+
eval_wer = 0.098
|