doccano

深度学习 PV:

Doccano是一种用于文本标注的开源工具，旨在简化和加速标注任务的进行。它提供了一个直观的用户界面，使标注人员能够轻松地对文本数据进行标注，并创建高质量的训练数据集用于机器学习和自然语言处理任务。

链接：https://github.com/doccano/doccano

一、安装部署

环境

操作系统：Centos7.9

python：3.10

doccano：1.6.2

pip安装

注：百度源没有相应安装包

pip install doccano==1.6.2 -i https://pypi.tuna.tsinghua.edu.cn/simple

初始化

doccano init

设置超级管理员账号密码

doccano createuser --username admin --password 123456

启动服务

doccano webserver --port 8000

ERNIE-UIE 关系抽取微调数据标注

创建序列标注任务

导入增量训练数据集

注：如果导入不成功，长时间转圈，需要去控制台执行doccano task

创建实体标签和关系标签

数据标注

数据集导出

1
2
3
4
5
{"id": 11, "text": "钢筋调直宜采用机械方法,也可以采用冷拉方法", "relations": [{"id": 1, "from_id": 45, "to_id": 44, "type": "材料"}, {"id": 2, "from_id": 46, "to_id": 44, "type": "材料"}], "entities": [{"id": 44, "start_offset": 0, "end_offset": 4, "label": "工程"}, {"id": 45, "start_offset": 7, "end_offset": 11, "label": "工艺"}, {"id": 46, "start_offset": 17, "end_offset": 21, "label": "工艺"}]}
{"id": 12, "text": "受力钢筋的接头形式应按设计要求采用,若设计无要求时,钢筋宜采用焊接接头和机械连接接头,也可采用绑扎接头。", "relations": [{"id": 3, "from_id": 53, "to_id": 22, "type": "材料"}, {"id": 4, "from_id": 54, "to_id": 22, "type": "材料"}, {"id": 5, "from_id": 55, "to_id": 22, "type": "材料"}], "entities": [{"id": 22, "start_offset": 0, "end_offset": 9, "label": "工程"}, {"id": 53, "start_offset": 31, "end_offset": 35, "label": "工艺"}, {"id": 54, "start_offset": 36, "end_offset": 42, "label": "工艺"}, {"id": 55, "start_offset": 47, "end_offset": 51, "label": "工艺"}]}
{"id": 13, "text": "多层非焊接钢筋骨架的各层钢筋之间,应保持层距准确,宜采用短钢筋支垫。", "relations": [{"id": 6, "from_id": 60, "to_id": 59, "type": "工艺"}], "entities": [{"id": 59, "start_offset": 0, "end_offset": 9, "label": "工程"}, {"id": 60, "start_offset": 28, "end_offset": 33, "label": "工艺"}]}
{"id": 14, "text": "预制桩的修筑工艺包括一体化成孔、自灌注。", "relations": [{"id": 7, "from_id": 62, "to_id": 28, "type": "工艺"}, {"id": 8, "from_id": 61, "to_id": 28, "type": "工艺"}], "entities": [{"id": 28, "start_offset": 0, "end_offset": 3, "label": "工程"}, {"id": 61, "start_offset": 10, "end_offset": 15, "label": "工艺"}, {"id": 62, "start_offset": 16, "end_offset": 19, "label": "工艺"}]}
{"id": 15, "text": "目前我国水运工程的模板用材已向多样化发展,除钢材和木材外,胶木板、竹胶板、塑料等已得到广泛运用,并取得了较好的技术经济效益。", "relations": [{"id": 9, "from_id": 63, "to_id": 52, "type": "材料"}, {"id": 10, "from_id": 64, "to_id": 52, "type": "材料"}, {"id": 11, "from_id": 65, "to_id": 52, "type": "材料"}, {"id": 12, "from_id": 67, "to_id": 52, "type": "材料"}], "entities": [{"id": 52, "start_offset": 4, "end_offset": 8, "label": "工程"}, {"id": 63, "start_offset": 22, "end_offset": 24, "label": "材料"}, {"id": 64, "start_offset": 25, "end_offset": 27, "label": "材料"}, {"id": 65, "start_offset": 29, "end_offset": 32, "label": "材料"}, {"id": 66, "start_offset": 33, "end_offset": 36, "label": "材料"}, {"id": 67, "start_offset": 37, "end_offset": 39, "label": "材料"}]}

数据集格式转换

参考https://github.com/PaddlePaddle/PaddleNLP/tree/develop/legacy/model_zoo/uie

进入PaddleNLP-UIE路径，将json文件放置在路径下，创建data文件夹用于存储数据集，执行：

1
2
3
4
5
6
python doccano.py \
    --doccano_file ./data/doccano_ext.json \
    --task_type ext \
    --save_dir ./data \
    --splits 0.8 0.2 0 \
    --schema_lang ch

由于测试样本较少（5条），未自动化分验证集dev，手动将测试集内容复制到验证集当中。

开启训练

在UIE路径下创建checkpoint/model_best用于存放模型

执行（GPU）：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
export finetuned_model=./checkpoint/model_best

python -u -m paddle.distributed.launch --gpus "0,1" finetune.py \
    --device gpu \
    --logging_steps 10 \
    --save_steps 100 \
    --eval_steps 100 \
    --seed 42 \
    --model_name_or_path uie-base \
    --output_dir $finetuned_model \
    --train_path data/train.txt \
    --dev_path data/dev.txt  \
    --max_seq_length 512  \
    --per_device_eval_batch_size 16 \
    --per_device_train_batch_size  16 \
    --num_train_epochs 100 \
    --learning_rate 1e-5 \
    --do_train \
    --do_eval \
    --do_export \
    --export_model_dir $finetuned_model \
    --label_names "start_positions" "end_positions" \
    --overwrite_output_dir \
    --disable_tqdm True \
    --metric_for_best_model eval_f1 \
    --load_best_model_at_end  True \
    --save_total_limit 1 \

训练结果：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
(py39_ppner_2_7_2) [root@jdz uie]# python -u -m paddle.distributed.launch --gpus "1" finetune.py     --device gpu     --logging_steps 10     --save_steps 100     --eval_steps 100     --seed 42     --model_name_or_path uie-base     --output_dir $finetuned_model     --train_path data/train.txt     --dev_path data/dev.txt      --max_seq_length 512      --per_device_eval_batch_size 16     --per_device_train_batch_size  16     --num_train_epochs 100     --learning_rate 1e-5     --do_train     --do_eval     --do_export     --export_model_dir $finetuned_model     --label_names "start_positions" "end_positions"     --overwrite_output_dir     --disable_tqdm True     --metric_for_best_model eval_f1     --load_best_model_at_end  True     --save_total_limit 1 
LAUNCH INFO 2024-06-26 18:05:23,778 -----------  Configuration  ----------------------
LAUNCH INFO 2024-06-26 18:05:23,779 auto_parallel_config: None
LAUNCH INFO 2024-06-26 18:05:23,779 auto_tuner_json: None
LAUNCH INFO 2024-06-26 18:05:23,779 devices: 1
LAUNCH INFO 2024-06-26 18:05:23,779 elastic_level: -1
LAUNCH INFO 2024-06-26 18:05:23,779 elastic_timeout: 30
LAUNCH INFO 2024-06-26 18:05:23,779 enable_gpu_log: True
LAUNCH INFO 2024-06-26 18:05:23,779 gloo_port: 6767
LAUNCH INFO 2024-06-26 18:05:23,779 host: None
LAUNCH INFO 2024-06-26 18:05:23,779 ips: None
LAUNCH INFO 2024-06-26 18:05:23,779 job_id: default
LAUNCH INFO 2024-06-26 18:05:23,779 legacy: False
LAUNCH INFO 2024-06-26 18:05:23,779 log_dir: log
LAUNCH INFO 2024-06-26 18:05:23,780 log_level: INFO
LAUNCH INFO 2024-06-26 18:05:23,780 log_overwrite: False
LAUNCH INFO 2024-06-26 18:05:23,780 master: None
LAUNCH INFO 2024-06-26 18:05:23,780 max_restart: 3
LAUNCH INFO 2024-06-26 18:05:23,780 nnodes: 1
LAUNCH INFO 2024-06-26 18:05:23,780 nproc_per_node: None
LAUNCH INFO 2024-06-26 18:05:23,780 rank: -1
LAUNCH INFO 2024-06-26 18:05:23,780 run_mode: collective
LAUNCH INFO 2024-06-26 18:05:23,780 server_num: None
LAUNCH INFO 2024-06-26 18:05:23,780 servers: 
LAUNCH INFO 2024-06-26 18:05:23,780 sort_ip: False
LAUNCH INFO 2024-06-26 18:05:23,780 start_port: 6070
LAUNCH INFO 2024-06-26 18:05:23,780 trainer_num: None
LAUNCH INFO 2024-06-26 18:05:23,780 trainers: 
LAUNCH INFO 2024-06-26 18:05:23,780 training_script: finetune.py
LAUNCH INFO 2024-06-26 18:05:23,781 training_script_args: ['--device', 'gpu', '--logging_steps', '10', '--save_steps', '100', '--eval_steps', '100', '--seed', '42', '--model_name_or_path', 'uie-base', '--output_dir', './checkpoint/model_best', '--train_path', 'data/train.txt', '--dev_path', 'data/dev.txt', '--max_seq_length', '512', '--per_device_eval_batch_size', '16', '--per_device_train_batch_size', '16', '--num_train_epochs', '100', '--learning_rate', '1e-5', '--do_train', '--do_eval', '--do_export', '--export_model_dir', './checkpoint/model_best', '--label_names', 'start_positions', 'end_positions', '--overwrite_output_dir', '--disable_tqdm', 'True', '--metric_for_best_model', 'eval_f1', '--load_best_model_at_end', 'True', '--save_total_limit', '1']
LAUNCH INFO 2024-06-26 18:05:23,781 with_gloo: 1
LAUNCH INFO 2024-06-26 18:05:23,781 --------------------------------------------------
LAUNCH INFO 2024-06-26 18:05:23,782 Job: default, mode collective, replicas 1[1:1], elastic False
LAUNCH INFO 2024-06-26 18:05:23,797 Run Pod: pwbjet, replicas 1, status ready
LAUNCH INFO 2024-06-26 18:05:23,824 Watching Pod: pwbjet, replicas 1, status running
/root/anaconda3/envs/py39_ppner_2_7_2/lib/python3.9/site-packages/_distutils_hack/__init__.py:33: UserWarning: Setuptools is replacing distutils.
  warnings.warn("Setuptools is replacing distutils.")
[2024-06-26 18:05:27,933] [ WARNING] - evaluation_strategy reset to IntervalStrategy.STEPS for do_eval is True. you can also set evaluation_strategy='epoch'.
[2024-06-26 18:05:27,933] [    INFO] - The default value for the training argument `--report_to` will change in v5 (from all installed integrations to none). In v5, you will need to use `--report_to all` to get the same behavior as now. You should start updating your code and make this info disappear :-).
[2024-06-26 18:05:27,934] [    INFO] - ============================================================
[2024-06-26 18:05:27,934] [    INFO] -      Model Configuration Arguments      
[2024-06-26 18:05:27,934] [    INFO] - paddle commit id              :fbf852dd832bc0e63ae31cd4aa37defd829e4c03
[2024-06-26 18:05:27,934] [    INFO] - export_model_dir              :./checkpoint/model_best
[2024-06-26 18:05:27,934] [    INFO] - model_name_or_path            :uie-base
[2024-06-26 18:05:27,934] [    INFO] - multilingual                  :False
[2024-06-26 18:05:27,934] [    INFO] - 
[2024-06-26 18:05:27,934] [    INFO] - ============================================================
[2024-06-26 18:05:27,934] [    INFO] -       Data Configuration Arguments      
[2024-06-26 18:05:27,934] [    INFO] - paddle commit id              :fbf852dd832bc0e63ae31cd4aa37defd829e4c03
[2024-06-26 18:05:27,935] [    INFO] - dev_path                      :data/dev.txt
[2024-06-26 18:05:27,935] [    INFO] - dynamic_max_length            :None
[2024-06-26 18:05:27,935] [    INFO] - max_seq_length                :512
[2024-06-26 18:05:27,935] [    INFO] - train_path                    :data/train.txt
[2024-06-26 18:05:27,935] [    INFO] - 
[2024-06-26 18:05:27,935] [ WARNING] - Process rank: -1, device: gpu, world_size: 1, distributed training: False, 16-bits training: False
[2024-06-26 18:05:27,935] [    INFO] - We are using (<class 'paddlenlp.transformers.ernie.tokenizer.ErnieTokenizer'>, False) to load 'uie-base'.
[2024-06-26 18:05:27,936] [    INFO] - Already cached /root/.paddlenlp/models/uie-base/ernie_3.0_base_zh_vocab.txt
[2024-06-26 18:05:27,968] [    INFO] - tokenizer config file saved in /root/.paddlenlp/models/uie-base/tokenizer_config.json
[2024-06-26 18:05:27,969] [    INFO] - Special tokens file saved in /root/.paddlenlp/models/uie-base/special_tokens_map.json
[2024-06-26 18:05:27,970] [    INFO] - Already cached /root/.paddlenlp/models/uie-base/model_state.pdparams
[2024-06-26 18:05:27,970] [    INFO] - Loading weights file model_state.pdparams from cache at /root/.paddlenlp/models/uie-base/model_state.pdparams
[2024-06-26 18:05:28,898] [    INFO] - Loaded weights file from disk, setting weights to model.
W0626 18:05:29.062965 285242 gpu_resources.cc:119] Please NOTE: device: 1, GPU Compute Capability: 8.6, Driver API Version: 12.2, Runtime API Version: 12.0
W0626 18:05:29.064332 285242 gpu_resources.cc:164] device: 1, cuDNN Version: 8.9.
[2024-06-26 18:05:30,516] [    INFO] - All model checkpoint weights were used when initializing UIE.

[2024-06-26 18:05:30,517] [    INFO] - All the weights of UIE were initialized from the model checkpoint at uie-base.
If your task is similar to the task the model of the checkpoint was trained on, you can already use UIE for predictions without further training.
[2024-06-26 18:05:30,562] [    INFO] - The global seed is set to 42, local seed is set to 43 and random seed is set to 42.
[2024-06-26 18:05:30,655] [   DEBUG] - ============================================================
[2024-06-26 18:05:30,655] [   DEBUG] -     Training Configuration Arguments    
[2024-06-26 18:05:30,656] [   DEBUG] - paddle commit id              : fbf852dd832bc0e63ae31cd4aa37defd829e4c03
[2024-06-26 18:05:30,656] [   DEBUG] - paddlenlp commit id           : b39e701e21d11ff66ac3abfc81d384b6af8f8240
[2024-06-26 18:05:30,656] [   DEBUG] - _no_sync_in_gradient_accumulation: True
[2024-06-26 18:05:30,656] [   DEBUG] - activation_quantize_type      : None
[2024-06-26 18:05:30,656] [   DEBUG] - adam_beta1                    : 0.9
[2024-06-26 18:05:30,656] [   DEBUG] - adam_beta2                    : 0.999
[2024-06-26 18:05:30,656] [   DEBUG] - adam_epsilon                  : 1e-08
[2024-06-26 18:05:30,656] [   DEBUG] - algo_list                     : None
[2024-06-26 18:05:30,656] [   DEBUG] - amp_custom_black_list         : None
[2024-06-26 18:05:30,656] [   DEBUG] - amp_custom_white_list         : None
[2024-06-26 18:05:30,656] [   DEBUG] - amp_master_grad               : False
[2024-06-26 18:05:30,656] [   DEBUG] - batch_num_list                : None
[2024-06-26 18:05:30,657] [   DEBUG] - batch_size_list               : None
[2024-06-26 18:05:30,657] [   DEBUG] - bf16                          : False
[2024-06-26 18:05:30,657] [   DEBUG] - bf16_full_eval                : False
[2024-06-26 18:05:30,657] [   DEBUG] - bias_correction               : False
[2024-06-26 18:05:30,657] [   DEBUG] - current_device                : gpu:1
[2024-06-26 18:05:30,657] [   DEBUG] - data_parallel_rank            : 0
[2024-06-26 18:05:30,657] [   DEBUG] - dataloader_drop_last          : False
[2024-06-26 18:05:30,657] [   DEBUG] - dataloader_num_workers        : 0
[2024-06-26 18:05:30,657] [   DEBUG] - dataset_rank                  : 0
[2024-06-26 18:05:30,657] [   DEBUG] - dataset_world_size            : 1
[2024-06-26 18:05:30,657] [   DEBUG] - device                        : gpu
[2024-06-26 18:05:30,657] [   DEBUG] - disable_tqdm                  : True
[2024-06-26 18:05:30,657] [   DEBUG] - distributed_dataloader        : False
[2024-06-26 18:05:30,657] [   DEBUG] - do_compress                   : False
[2024-06-26 18:05:30,658] [   DEBUG] - do_eval                       : True
[2024-06-26 18:05:30,658] [   DEBUG] - do_export                     : True
[2024-06-26 18:05:30,658] [   DEBUG] - do_predict                    : False
[2024-06-26 18:05:30,658] [   DEBUG] - do_train                      : True
[2024-06-26 18:05:30,658] [   DEBUG] - eval_accumulation_steps       : None
[2024-06-26 18:05:30,658] [   DEBUG] - eval_batch_size               : 16
[2024-06-26 18:05:30,658] [   DEBUG] - eval_steps                    : 100
[2024-06-26 18:05:30,658] [   DEBUG] - evaluation_strategy           : IntervalStrategy.STEPS
[2024-06-26 18:05:30,658] [   DEBUG] - flatten_param_grads           : False
[2024-06-26 18:05:30,658] [   DEBUG] - force_reshard_pp              : False
[2024-06-26 18:05:30,658] [   DEBUG] - fp16                          : False
[2024-06-26 18:05:30,658] [   DEBUG] - fp16_full_eval                : False
[2024-06-26 18:05:30,658] [   DEBUG] - fp16_opt_level                : O1
[2024-06-26 18:05:30,658] [   DEBUG] - gradient_accumulation_steps   : 1
[2024-06-26 18:05:30,659] [   DEBUG] - greater_is_better             : True
[2024-06-26 18:05:30,659] [   DEBUG] - hybrid_parallel_topo_order    : None
[2024-06-26 18:05:30,659] [   DEBUG] - ignore_data_skip              : False
[2024-06-26 18:05:30,659] [   DEBUG] - ignore_load_lr_and_optim      : False
[2024-06-26 18:05:30,659] [   DEBUG] - input_dtype                   : int64
[2024-06-26 18:05:30,659] [   DEBUG] - input_infer_model_path        : None
[2024-06-26 18:05:30,659] [   DEBUG] - label_names                   : ['start_positions', 'end_positions']
[2024-06-26 18:05:30,659] [   DEBUG] - lazy_data_processing          : True
[2024-06-26 18:05:30,659] [   DEBUG] - learning_rate                 : 1e-05
[2024-06-26 18:05:30,659] [   DEBUG] - load_best_model_at_end        : True
[2024-06-26 18:05:30,659] [   DEBUG] - load_sharded_model            : False
[2024-06-26 18:05:30,659] [   DEBUG] - local_process_index           : 0
[2024-06-26 18:05:30,659] [   DEBUG] - local_rank                    : -1
[2024-06-26 18:05:30,659] [   DEBUG] - log_level                     : -1
[2024-06-26 18:05:30,660] [   DEBUG] - log_level_replica             : -1
[2024-06-26 18:05:30,660] [   DEBUG] - log_on_each_node              : True
[2024-06-26 18:05:30,660] [   DEBUG] - logging_dir                   : ./checkpoint/model_best/runs/Jun26_18-05-27_jdz
[2024-06-26 18:05:30,660] [   DEBUG] - logging_first_step            : False
[2024-06-26 18:05:30,660] [   DEBUG] - logging_steps                 : 10
[2024-06-26 18:05:30,660] [   DEBUG] - logging_strategy              : IntervalStrategy.STEPS
[2024-06-26 18:05:30,660] [   DEBUG] - logical_process_index         : 0
[2024-06-26 18:05:30,660] [   DEBUG] - lr_end                        : 1e-07
[2024-06-26 18:05:30,660] [   DEBUG] - lr_scheduler_type             : SchedulerType.LINEAR
[2024-06-26 18:05:30,660] [   DEBUG] - max_evaluate_steps            : -1
[2024-06-26 18:05:30,660] [   DEBUG] - max_grad_norm                 : 1.0
[2024-06-26 18:05:30,660] [   DEBUG] - max_steps                     : -1
[2024-06-26 18:05:30,660] [   DEBUG] - metric_for_best_model         : eval_f1
[2024-06-26 18:05:30,660] [   DEBUG] - minimum_eval_times            : None
[2024-06-26 18:05:30,660] [   DEBUG] - moving_rate                   : 0.9
[2024-06-26 18:05:30,661] [   DEBUG] - no_cuda                       : False
[2024-06-26 18:05:30,661] [   DEBUG] - num_cycles                    : 0.5
[2024-06-26 18:05:30,661] [   DEBUG] - num_train_epochs              : 100.0
[2024-06-26 18:05:30,661] [   DEBUG] - onnx_format                   : True
[2024-06-26 18:05:30,661] [   DEBUG] - optim                         : OptimizerNames.ADAMW
[2024-06-26 18:05:30,661] [   DEBUG] - optimizer_name_suffix         : None
[2024-06-26 18:05:30,661] [   DEBUG] - output_dir                    : ./checkpoint/model_best
[2024-06-26 18:05:30,661] [   DEBUG] - overwrite_output_dir          : True
[2024-06-26 18:05:30,661] [   DEBUG] - past_index                    : -1
[2024-06-26 18:05:30,661] [   DEBUG] - per_device_eval_batch_size    : 16
[2024-06-26 18:05:30,661] [   DEBUG] - per_device_train_batch_size   : 16
[2024-06-26 18:05:30,661] [   DEBUG] - pipeline_parallel_config      : 
[2024-06-26 18:05:30,661] [   DEBUG] - pipeline_parallel_degree      : -1
[2024-06-26 18:05:30,661] [   DEBUG] - pipeline_parallel_rank        : 0
[2024-06-26 18:05:30,662] [   DEBUG] - power                         : 1.0
[2024-06-26 18:05:30,662] [   DEBUG] - prediction_loss_only          : False
[2024-06-26 18:05:30,662] [   DEBUG] - process_index                 : 0
[2024-06-26 18:05:30,662] [   DEBUG] - prune_embeddings              : False
[2024-06-26 18:05:30,662] [   DEBUG] - recompute                     : False
[2024-06-26 18:05:30,662] [   DEBUG] - remove_unused_columns         : True
[2024-06-26 18:05:30,662] [   DEBUG] - report_to                     : ['visualdl']
[2024-06-26 18:05:30,662] [   DEBUG] - resume_from_checkpoint        : None
[2024-06-26 18:05:30,662] [   DEBUG] - round_type                    : round
[2024-06-26 18:05:30,662] [   DEBUG] - run_name                      : ./checkpoint/model_best
[2024-06-26 18:05:30,662] [   DEBUG] - save_on_each_node             : False
[2024-06-26 18:05:30,662] [   DEBUG] - save_sharded_model            : False
[2024-06-26 18:05:30,662] [   DEBUG] - save_steps                    : 100
[2024-06-26 18:05:30,662] [   DEBUG] - save_strategy                 : IntervalStrategy.STEPS
[2024-06-26 18:05:30,663] [   DEBUG] - save_total_limit              : 1
[2024-06-26 18:05:30,663] [   DEBUG] - scale_loss                    : 32768
[2024-06-26 18:05:30,663] [   DEBUG] - seed                          : 42
[2024-06-26 18:05:30,663] [   DEBUG] - sep_parallel_degree           : -1
[2024-06-26 18:05:30,663] [   DEBUG] - sharding                      : []
[2024-06-26 18:05:30,663] [   DEBUG] - sharding_degree               : -1
[2024-06-26 18:05:30,663] [   DEBUG] - sharding_parallel_config      : 
[2024-06-26 18:05:30,663] [   DEBUG] - sharding_parallel_degree      : -1
[2024-06-26 18:05:30,663] [   DEBUG] - sharding_parallel_rank        : 0
[2024-06-26 18:05:30,663] [   DEBUG] - should_load_dataset           : True
[2024-06-26 18:05:30,663] [   DEBUG] - should_load_sharding_stage1_model: False
[2024-06-26 18:05:30,663] [   DEBUG] - should_log                    : True
[2024-06-26 18:05:30,663] [   DEBUG] - should_save                   : True
[2024-06-26 18:05:30,663] [   DEBUG] - should_save_model_state       : True
[2024-06-26 18:05:30,664] [   DEBUG] - should_save_sharding_stage1_model: False
[2024-06-26 18:05:30,664] [   DEBUG] - skip_memory_metrics           : True
[2024-06-26 18:05:30,664] [   DEBUG] - skip_profile_timer            : True
[2024-06-26 18:05:30,664] [   DEBUG] - strategy                      : dynabert+ptq
[2024-06-26 18:05:30,664] [   DEBUG] - tensor_parallel_config        : 
[2024-06-26 18:05:30,664] [   DEBUG] - tensor_parallel_degree        : -1
[2024-06-26 18:05:30,664] [   DEBUG] - tensor_parallel_rank          : 0
[2024-06-26 18:05:30,664] [   DEBUG] - to_static                     : False
[2024-06-26 18:05:30,664] [   DEBUG] - train_batch_size              : 16
[2024-06-26 18:05:30,664] [   DEBUG] - unified_checkpoint            : False
[2024-06-26 18:05:30,664] [   DEBUG] - unified_checkpoint_config     : 
[2024-06-26 18:05:30,664] [   DEBUG] - use_auto_parallel             : False
[2024-06-26 18:05:30,664] [   DEBUG] - use_hybrid_parallel           : False
[2024-06-26 18:05:30,664] [   DEBUG] - use_pact                      : True
[2024-06-26 18:05:30,665] [   DEBUG] - warmup_ratio                  : 0.1
[2024-06-26 18:05:30,665] [   DEBUG] - warmup_steps                  : 0
[2024-06-26 18:05:30,665] [   DEBUG] - weight_decay                  : 0.0
[2024-06-26 18:05:30,665] [   DEBUG] - weight_name_suffix            : None
[2024-06-26 18:05:30,665] [   DEBUG] - weight_quantize_type          : channel_wise_abs_max
[2024-06-26 18:05:30,665] [   DEBUG] - width_mult_list               : None
[2024-06-26 18:05:30,665] [   DEBUG] - world_size                    : 1
[2024-06-26 18:05:30,665] [   DEBUG] - 
[2024-06-26 18:05:30,666] [    INFO] - Starting training from resume_from_checkpoint : None
/root/anaconda3/envs/py39_ppner_2_7_2/lib/python3.9/site-packages/paddle/distributed/parallel.py:410: UserWarning: The program will return to single-card operation. Please check 1, whether you use spawn or fleetrun to start the program. 2, Whether it is a multi-card program. 3, Is the current environment multi-card.
  warnings.warn(
[2024-06-26 18:05:30,667] [    INFO] - [timelog] checkpoint loading time: 0.00s (2024-06-26 18:05:30) 
[2024-06-26 18:05:30,667] [    INFO] - ***** Running training *****
[2024-06-26 18:05:30,667] [    INFO] -   Num examples = 68
[2024-06-26 18:05:30,667] [    INFO] -   Num Epochs = 100
[2024-06-26 18:05:30,667] [    INFO] -   Instantaneous batch size per device = 16
[2024-06-26 18:05:30,668] [    INFO] -   Total train batch size (w. parallel, distributed & accumulation) = 16
[2024-06-26 18:05:30,668] [    INFO] -   Gradient Accumulation steps = 1
[2024-06-26 18:05:30,668] [    INFO] -   Total optimization steps = 500
[2024-06-26 18:05:30,668] [    INFO] -   Total num train samples = 6,800
[2024-06-26 18:05:30,670] [   DEBUG] -   Number of trainable parameters = 117,946,370 (per device)
/root/anaconda3/envs/py39_ppner_2_7_2/lib/python3.9/site-packages/paddlenlp/transformers/tokenizer_utils_base.py:2538: FutureWarning: The `max_seq_len` argument is deprecated and will be removed in a future version, please use `max_length` instead.
  warnings.warn(
/root/anaconda3/envs/py39_ppner_2_7_2/lib/python3.9/site-packages/paddlenlp/transformers/tokenizer_utils_base.py:1938: FutureWarning: The `pad_to_max_length` argument is deprecated and will be removed in a future version, use `padding=True` or `padding='longest'` to pad to the longest sequence in the batch, or use `padding='max_length'` to pad to a max length. In this case, you can give a specific length with `max_length` (e.g. `max_length=45`) or leave max_length to None to pad to the maximal input size of the model (e.g. 512 for Bert).
  warnings.warn(
[2024-06-26 18:05:34,727] [    INFO] - loss: 0.00208795, learning_rate: 1e-05, global_step: 10, interval_runtime: 4.0565, interval_samples_per_second: 39.44293710564109, interval_steps_per_second: 2.4651835691025683, progress_or_epoch: 2.0
[2024-06-26 18:05:37,278] [    INFO] - loss: 0.00129266, learning_rate: 1e-05, global_step: 20, interval_runtime: 2.55, interval_samples_per_second: 62.74611413497253, interval_steps_per_second: 3.921632133435783, progress_or_epoch: 4.0
[2024-06-26 18:05:39,833] [    INFO] - loss: 0.00092623, learning_rate: 1e-05, global_step: 30, interval_runtime: 2.5559, interval_samples_per_second: 62.599374107598685, interval_steps_per_second: 3.912460881724918, progress_or_epoch: 6.0
[2024-06-26 18:05:42,413] [    INFO] - loss: 0.00058895, learning_rate: 1e-05, global_step: 40, interval_runtime: 2.5801, interval_samples_per_second: 62.012657597381434, interval_steps_per_second: 3.8757910998363396, progress_or_epoch: 8.0
[2024-06-26 18:05:44,946] [    INFO] - loss: 0.00047016, learning_rate: 1e-05, global_step: 50, interval_runtime: 2.5327, interval_samples_per_second: 63.17332385701746, interval_steps_per_second: 3.948332741063591, progress_or_epoch: 10.0
[2024-06-26 18:05:47,514] [    INFO] - loss: 0.00031726, learning_rate: 1e-05, global_step: 60, interval_runtime: 2.5675, interval_samples_per_second: 62.316911606640005, interval_steps_per_second: 3.8948069754150003, progress_or_epoch: 12.0
[2024-06-26 18:05:50,081] [    INFO] - loss: 0.00024869, learning_rate: 1e-05, global_step: 70, interval_runtime: 2.5673, interval_samples_per_second: 62.32305770213569, interval_steps_per_second: 3.8951911063834808, progress_or_epoch: 14.0
[2024-06-26 18:05:52,658] [    INFO] - loss: 0.00041933, learning_rate: 1e-05, global_step: 80, interval_runtime: 2.5765, interval_samples_per_second: 62.10074334302723, interval_steps_per_second: 3.881296458939202, progress_or_epoch: 16.0
[2024-06-26 18:05:55,223] [    INFO] - loss: 0.00017784, learning_rate: 1e-05, global_step: 90, interval_runtime: 2.5647, interval_samples_per_second: 62.38556549647052, interval_steps_per_second: 3.8990978435294075, progress_or_epoch: 18.0
[2024-06-26 18:05:57,816] [    INFO] - loss: 0.00019145, learning_rate: 1e-05, global_step: 100, interval_runtime: 2.5931, interval_samples_per_second: 61.70326891265709, interval_steps_per_second: 3.856454307041068, progress_or_epoch: 20.0
[2024-06-26 18:05:57,817] [    INFO] - ***** Running Evaluation *****
[2024-06-26 18:05:57,817] [    INFO] -   Num examples = 4
[2024-06-26 18:05:57,817] [    INFO] -   Total prediction steps = 1
[2024-06-26 18:05:57,817] [    INFO] -   Pre device batch size = 16
[2024-06-26 18:05:57,818] [    INFO] -   Total Batch size = 16
[2024-06-26 18:05:57,889] [    INFO] - eval_loss: 0.005224619060754776, eval_precision: 0.4, eval_recall: 0.4, eval_f1: 0.4000000000000001, eval_runtime: 0.0705, eval_samples_per_second: 56.74937846074747, eval_steps_per_second: 14.187344615186868, progress_or_epoch: 20.0
[2024-06-26 18:05:57,889] [    INFO] - Saving model checkpoint to ./checkpoint/model_best/checkpoint-100
[2024-06-26 18:05:57,890] [    INFO] - tokenizer config file saved in ./checkpoint/model_best/checkpoint-100/tokenizer_config.json
[2024-06-26 18:05:57,890] [    INFO] - Special tokens file saved in ./checkpoint/model_best/checkpoint-100/special_tokens_map.json
[2024-06-26 18:05:57,902] [    INFO] - Configuration saved in ./checkpoint/model_best/checkpoint-100/config.json
[2024-06-26 18:06:00,646] [    INFO] - Model weights saved in ./checkpoint/model_best/checkpoint-100/model_state.pdparams
[2024-06-26 18:06:00,648] [    INFO] - Saving optimizer files.
[2024-06-26 18:06:07,375] [    INFO] - [timelog] checkpoint saving time: 9.48s (2024-06-26 18:06:07) 
[2024-06-26 18:06:09,924] [    INFO] - loss: 0.00022237, learning_rate: 1e-05, global_step: 110, interval_runtime: 12.1083, interval_samples_per_second: 13.21409688201894, interval_steps_per_second: 0.8258810551261837, progress_or_epoch: 22.0
[2024-06-26 18:06:12,478] [    INFO] - loss: 0.00017666, learning_rate: 1e-05, global_step: 120, interval_runtime: 2.5542, interval_samples_per_second: 62.64218200480218, interval_steps_per_second: 3.9151363753001363, progress_or_epoch: 24.0
[2024-06-26 18:06:15,065] [    INFO] - loss: 0.00024367, learning_rate: 1e-05, global_step: 130, interval_runtime: 2.587, interval_samples_per_second: 61.84880269252828, interval_steps_per_second: 3.8655501682830176, progress_or_epoch: 26.0
[2024-06-26 18:06:17,695] [    INFO] - loss: 0.0001915, learning_rate: 1e-05, global_step: 140, interval_runtime: 2.6297, interval_samples_per_second: 60.84334418667863, interval_steps_per_second: 3.8027090116674143, progress_or_epoch: 28.0
[2024-06-26 18:06:20,298] [    INFO] - loss: 0.00017253, learning_rate: 1e-05, global_step: 150, interval_runtime: 2.5998, interval_samples_per_second: 61.542086843031, interval_steps_per_second: 3.8463804276894376, progress_or_epoch: 30.0
[2024-06-26 18:06:22,895] [    INFO] - loss: 0.00014841, learning_rate: 1e-05, global_step: 160, interval_runtime: 2.6008, interval_samples_per_second: 61.51999965531349, interval_steps_per_second: 3.844999978457093, progress_or_epoch: 32.0
[2024-06-26 18:06:25,498] [    INFO] - loss: 0.00012293, learning_rate: 1e-05, global_step: 170, interval_runtime: 2.6024, interval_samples_per_second: 61.48274409158785, interval_steps_per_second: 3.8426715057242404, progress_or_epoch: 34.0
[2024-06-26 18:06:28,099] [    INFO] - loss: 0.00010312, learning_rate: 1e-05, global_step: 180, interval_runtime: 2.601, interval_samples_per_second: 61.514997685471634, interval_steps_per_second: 3.844687355341977, progress_or_epoch: 36.0
[2024-06-26 18:06:30,698] [    INFO] - loss: 9.851e-05, learning_rate: 1e-05, global_step: 190, interval_runtime: 2.5992, interval_samples_per_second: 61.5571084161852, interval_steps_per_second: 3.847319276011575, progress_or_epoch: 38.0
[2024-06-26 18:06:33,295] [    INFO] - loss: 0.00013552, learning_rate: 1e-05, global_step: 200, interval_runtime: 2.5971, interval_samples_per_second: 61.60644154857953, interval_steps_per_second: 3.8504025967862208, progress_or_epoch: 40.0
[2024-06-26 18:06:33,295] [    INFO] - ***** Running Evaluation *****
[2024-06-26 18:06:33,295] [    INFO] -   Num examples = 4
[2024-06-26 18:06:33,295] [    INFO] -   Total prediction steps = 1
[2024-06-26 18:06:33,295] [    INFO] -   Pre device batch size = 16
[2024-06-26 18:06:33,295] [    INFO] -   Total Batch size = 16
[2024-06-26 18:06:33,358] [    INFO] - eval_loss: 0.006688571535050869, eval_precision: 0.4, eval_recall: 0.4, eval_f1: 0.4000000000000001, eval_runtime: 0.0621, eval_samples_per_second: 64.41700614712398, eval_steps_per_second: 16.104251536780996, progress_or_epoch: 40.0
[2024-06-26 18:06:33,359] [    INFO] - Saving model checkpoint to ./checkpoint/model_best/checkpoint-200
[2024-06-26 18:06:33,360] [    INFO] - tokenizer config file saved in ./checkpoint/model_best/checkpoint-200/tokenizer_config.json
[2024-06-26 18:06:33,360] [    INFO] - Special tokens file saved in ./checkpoint/model_best/checkpoint-200/special_tokens_map.json
[2024-06-26 18:06:33,370] [    INFO] - Configuration saved in ./checkpoint/model_best/checkpoint-200/config.json
[2024-06-26 18:06:34,574] [    INFO] - Model weights saved in ./checkpoint/model_best/checkpoint-200/model_state.pdparams
[2024-06-26 18:06:34,574] [    INFO] - Saving optimizer files.
[2024-06-26 18:06:39,080] [    INFO] - [timelog] checkpoint saving time: 5.72s (2024-06-26 18:06:39) 
[2024-06-26 18:06:41,672] [    INFO] - loss: 0.00011128, learning_rate: 1e-05, global_step: 210, interval_runtime: 8.3772, interval_samples_per_second: 19.099448736351285, interval_steps_per_second: 1.1937155460219553, progress_or_epoch: 42.0
[2024-06-26 18:06:44,288] [    INFO] - loss: 0.00011245, learning_rate: 1e-05, global_step: 220, interval_runtime: 2.6149, interval_samples_per_second: 61.18868160731556, interval_steps_per_second: 3.8242926004572224, progress_or_epoch: 44.0
[2024-06-26 18:06:46,911] [    INFO] - loss: 0.00012268, learning_rate: 1e-05, global_step: 230, interval_runtime: 2.6234, interval_samples_per_second: 60.98920142963071, interval_steps_per_second: 3.8118250893519194, progress_or_epoch: 46.0
[2024-06-26 18:06:49,537] [    INFO] - loss: 0.00013804, learning_rate: 1e-05, global_step: 240, interval_runtime: 2.6259, interval_samples_per_second: 60.93070388366043, interval_steps_per_second: 3.8081689927287767, progress_or_epoch: 48.0
[2024-06-26 18:06:52,117] [    INFO] - loss: 0.00020081, learning_rate: 1e-05, global_step: 250, interval_runtime: 2.5809, interval_samples_per_second: 61.99492724918558, interval_steps_per_second: 3.8746829530740987, progress_or_epoch: 50.0
[2024-06-26 18:06:54,753] [    INFO] - loss: 0.00014258, learning_rate: 1e-05, global_step: 260, interval_runtime: 2.6358, interval_samples_per_second: 60.702937625070206, interval_steps_per_second: 3.793933601566888, progress_or_epoch: 52.0
[2024-06-26 18:06:57,369] [    INFO] - loss: 0.00012738, learning_rate: 1e-05, global_step: 270, interval_runtime: 2.6164, interval_samples_per_second: 61.153369707036596, interval_steps_per_second: 3.8220856066897873, progress_or_epoch: 54.0
[2024-06-26 18:07:00,006] [    INFO] - loss: 0.00010497, learning_rate: 1e-05, global_step: 280, interval_runtime: 2.6373, interval_samples_per_second: 60.669083466377856, interval_steps_per_second: 3.791817716648616, progress_or_epoch: 56.0
[2024-06-26 18:07:02,657] [    INFO] - loss: 0.00010167, learning_rate: 1e-05, global_step: 290, interval_runtime: 2.6499, interval_samples_per_second: 60.38069724825452, interval_steps_per_second: 3.7737935780159075, progress_or_epoch: 58.0
[2024-06-26 18:07:05,297] [    INFO] - loss: 0.000171, learning_rate: 1e-05, global_step: 300, interval_runtime: 2.6399, interval_samples_per_second: 60.60732256501828, interval_steps_per_second: 3.7879576603136424, progress_or_epoch: 60.0
[2024-06-26 18:07:05,298] [    INFO] - ***** Running Evaluation *****
[2024-06-26 18:07:05,298] [    INFO] -   Num examples = 4
[2024-06-26 18:07:05,298] [    INFO] -   Total prediction steps = 1
[2024-06-26 18:07:05,298] [    INFO] -   Pre device batch size = 16
[2024-06-26 18:07:05,298] [    INFO] -   Total Batch size = 16
[2024-06-26 18:07:05,357] [    INFO] - eval_loss: 0.007026550825685263, eval_precision: 0.4, eval_recall: 0.4, eval_f1: 0.4000000000000001, eval_runtime: 0.0582, eval_samples_per_second: 68.67042956838507, eval_steps_per_second: 17.16760739209627, progress_or_epoch: 60.0
[2024-06-26 18:07:05,358] [    INFO] - Saving model checkpoint to ./checkpoint/model_best/checkpoint-300
[2024-06-26 18:07:05,358] [    INFO] - tokenizer config file saved in ./checkpoint/model_best/checkpoint-300/tokenizer_config.json
[2024-06-26 18:07:05,358] [    INFO] - Special tokens file saved in ./checkpoint/model_best/checkpoint-300/special_tokens_map.json
[2024-06-26 18:07:05,364] [    INFO] - Configuration saved in ./checkpoint/model_best/checkpoint-300/config.json
[2024-06-26 18:07:07,995] [    INFO] - Model weights saved in ./checkpoint/model_best/checkpoint-300/model_state.pdparams
[2024-06-26 18:07:07,996] [    INFO] - Saving optimizer files.
[2024-06-26 18:07:11,005] [    INFO] - [timelog] checkpoint saving time: 5.64s (2024-06-26 18:07:11) 
[2024-06-26 18:07:13,640] [    INFO] - loss: 9.265e-05, learning_rate: 1e-05, global_step: 310, interval_runtime: 8.3438, interval_samples_per_second: 19.17600366118997, interval_steps_per_second: 1.1985002288243731, progress_or_epoch: 62.0
[2024-06-26 18:07:16,285] [    INFO] - loss: 0.00012613, learning_rate: 1e-05, global_step: 320, interval_runtime: 2.6441, interval_samples_per_second: 60.512013700058034, interval_steps_per_second: 3.782000856253627, progress_or_epoch: 64.0
[2024-06-26 18:07:18,887] [    INFO] - loss: 7.656e-05, learning_rate: 1e-05, global_step: 330, interval_runtime: 2.603, interval_samples_per_second: 61.4676855948393, interval_steps_per_second: 3.8417303496774564, progress_or_epoch: 66.0
[2024-06-26 18:07:21,539] [    INFO] - loss: 9.599e-05, learning_rate: 1e-05, global_step: 340, interval_runtime: 2.6517, interval_samples_per_second: 60.33811856609532, interval_steps_per_second: 3.7711324103809574, progress_or_epoch: 68.0
[2024-06-26 18:07:24,185] [    INFO] - loss: 9.96e-05, learning_rate: 1e-05, global_step: 350, interval_runtime: 2.6462, interval_samples_per_second: 60.463986738962795, interval_steps_per_second: 3.7789991711851747, progress_or_epoch: 70.0
[2024-06-26 18:07:26,830] [    INFO] - loss: 9.502e-05, learning_rate: 1e-05, global_step: 360, interval_runtime: 2.6438, interval_samples_per_second: 60.51813635974227, interval_steps_per_second: 3.782383522483892, progress_or_epoch: 72.0
[2024-06-26 18:07:29,486] [    INFO] - loss: 8.396e-05, learning_rate: 1e-05, global_step: 370, interval_runtime: 2.6568, interval_samples_per_second: 60.22368303445632, interval_steps_per_second: 3.76398018965352, progress_or_epoch: 74.0
[2024-06-26 18:07:32,128] [    INFO] - loss: 0.000107, learning_rate: 1e-05, global_step: 380, interval_runtime: 2.6414, interval_samples_per_second: 60.57406701856363, interval_steps_per_second: 3.785879188660227, progress_or_epoch: 76.0
[2024-06-26 18:07:34,774] [    INFO] - loss: 0.00015859, learning_rate: 1e-05, global_step: 390, interval_runtime: 2.646, interval_samples_per_second: 60.46891186644317, interval_steps_per_second: 3.7793069916526982, progress_or_epoch: 78.0
[2024-06-26 18:07:37,422] [    INFO] - loss: 8.44e-05, learning_rate: 1e-05, global_step: 400, interval_runtime: 2.6486, interval_samples_per_second: 60.40979254722639, interval_steps_per_second: 3.7756120342016493, progress_or_epoch: 80.0
[2024-06-26 18:07:37,423] [    INFO] - ***** Running Evaluation *****
[2024-06-26 18:07:37,423] [    INFO] -   Num examples = 4
[2024-06-26 18:07:37,423] [    INFO] -   Total prediction steps = 1
[2024-06-26 18:07:37,423] [    INFO] -   Pre device batch size = 16
[2024-06-26 18:07:37,423] [    INFO] -   Total Batch size = 16
[2024-06-26 18:07:37,482] [    INFO] - eval_loss: 0.00755023630335927, eval_precision: 0.4, eval_recall: 0.4, eval_f1: 0.4000000000000001, eval_runtime: 0.0578, eval_samples_per_second: 69.19010227647641, eval_steps_per_second: 17.297525569119102, progress_or_epoch: 80.0
[2024-06-26 18:07:37,482] [    INFO] - Saving model checkpoint to ./checkpoint/model_best/checkpoint-400
[2024-06-26 18:07:37,483] [    INFO] - tokenizer config file saved in ./checkpoint/model_best/checkpoint-400/tokenizer_config.json
[2024-06-26 18:07:37,483] [    INFO] - Special tokens file saved in ./checkpoint/model_best/checkpoint-400/special_tokens_map.json
[2024-06-26 18:07:37,489] [    INFO] - Configuration saved in ./checkpoint/model_best/checkpoint-400/config.json
[2024-06-26 18:07:38,671] [    INFO] - Model weights saved in ./checkpoint/model_best/checkpoint-400/model_state.pdparams
[2024-06-26 18:07:38,671] [    INFO] - Saving optimizer files.
[2024-06-26 18:07:41,730] [    INFO] - [timelog] checkpoint saving time: 4.24s (2024-06-26 18:07:41) 
[2024-06-26 18:07:44,394] [    INFO] - loss: 8.615e-05, learning_rate: 1e-05, global_step: 410, interval_runtime: 6.9719, interval_samples_per_second: 22.949178353528108, interval_steps_per_second: 1.4343236470955067, progress_or_epoch: 82.0
[2024-06-26 18:07:47,030] [    INFO] - loss: 8.146e-05, learning_rate: 1e-05, global_step: 420, interval_runtime: 2.6363, interval_samples_per_second: 60.691782269542095, interval_steps_per_second: 3.793236391846381, progress_or_epoch: 84.0
[2024-06-26 18:07:49,660] [    INFO] - loss: 0.0001024, learning_rate: 1e-05, global_step: 430, interval_runtime: 2.6297, interval_samples_per_second: 60.84334970295866, interval_steps_per_second: 3.8027093564349164, progress_or_epoch: 86.0
[2024-06-26 18:07:52,296] [    INFO] - loss: 0.00011797, learning_rate: 1e-05, global_step: 440, interval_runtime: 2.6357, interval_samples_per_second: 60.70518895680559, interval_steps_per_second: 3.7940743098003495, progress_or_epoch: 88.0
[2024-06-26 18:07:54,939] [    INFO] - loss: 0.00014716, learning_rate: 1e-05, global_step: 450, interval_runtime: 2.6435, interval_samples_per_second: 60.52505176192572, interval_steps_per_second: 3.7828157351203573, progress_or_epoch: 90.0
[2024-06-26 18:07:57,583] [    INFO] - loss: 7.592e-05, learning_rate: 1e-05, global_step: 460, interval_runtime: 2.6437, interval_samples_per_second: 60.520177521644605, interval_steps_per_second: 3.782511095102788, progress_or_epoch: 92.0
[2024-06-26 18:08:00,303] [    INFO] - loss: 8.616e-05, learning_rate: 1e-05, global_step: 470, interval_runtime: 2.7201, interval_samples_per_second: 58.82132721602211, interval_steps_per_second: 3.676332951001382, progress_or_epoch: 94.0
[2024-06-26 18:08:02,968] [    INFO] - loss: 7.984e-05, learning_rate: 1e-05, global_step: 480, interval_runtime: 2.6645, interval_samples_per_second: 60.048504909189305, interval_steps_per_second: 3.7530315568243315, progress_or_epoch: 96.0
[2024-06-26 18:08:05,618] [    INFO] - loss: 7.743e-05, learning_rate: 1e-05, global_step: 490, interval_runtime: 2.6507, interval_samples_per_second: 60.3610372488141, interval_steps_per_second: 3.7725648280508812, progress_or_epoch: 98.0
[2024-06-26 18:08:08,285] [    INFO] - loss: 7.793e-05, learning_rate: 1e-05, global_step: 500, interval_runtime: 2.6661, interval_samples_per_second: 60.01194714924565, interval_steps_per_second: 3.750746696827853, progress_or_epoch: 100.0
[2024-06-26 18:08:08,285] [    INFO] - ***** Running Evaluation *****
[2024-06-26 18:08:08,285] [    INFO] -   Num examples = 4
[2024-06-26 18:08:08,286] [    INFO] -   Total prediction steps = 1
[2024-06-26 18:08:08,286] [    INFO] -   Pre device batch size = 16
[2024-06-26 18:08:08,286] [    INFO] -   Total Batch size = 16
[2024-06-26 18:08:08,344] [    INFO] - eval_loss: 0.007735834456980228, eval_precision: 0.4, eval_recall: 0.4, eval_f1: 0.4000000000000001, eval_runtime: 0.0574, eval_samples_per_second: 69.71943866123114, eval_steps_per_second: 17.429859665307784, progress_or_epoch: 100.0
[2024-06-26 18:08:08,344] [    INFO] - Saving model checkpoint to ./checkpoint/model_best/checkpoint-500
[2024-06-26 18:08:08,345] [    INFO] - tokenizer config file saved in ./checkpoint/model_best/checkpoint-500/tokenizer_config.json
[2024-06-26 18:08:08,345] [    INFO] - Special tokens file saved in ./checkpoint/model_best/checkpoint-500/special_tokens_map.json
[2024-06-26 18:08:08,351] [    INFO] - Configuration saved in ./checkpoint/model_best/checkpoint-500/config.json
[2024-06-26 18:08:09,518] [    INFO] - Model weights saved in ./checkpoint/model_best/checkpoint-500/model_state.pdparams
[2024-06-26 18:08:09,518] [    INFO] - Saving optimizer files.
[2024-06-26 18:08:11,773] [    INFO] - [timelog] checkpoint saving time: 3.42s (2024-06-26 18:08:11) 
[2024-06-26 18:08:11,774] [    INFO] - 
Training completed. 

[2024-06-26 18:08:11,774] [    INFO] - Loading best model from ./checkpoint/model_best/checkpoint-100 (score: 0.4000000000000001).
[2024-06-26 18:08:12,116] [    INFO] - set state-dict :([], [])
[2024-06-26 18:08:12,118] [    INFO] - train_runtime: 161.4472, train_samples_per_second: 42.1190341985795, train_steps_per_second: 3.0969878087190805, train_loss: 0.00023241847997996956, progress_or_epoch: 100.0
[2024-06-26 18:08:12,134] [    INFO] - Saving model checkpoint to ./checkpoint/model_best
[2024-06-26 18:08:12,135] [    INFO] - tokenizer config file saved in ./checkpoint/model_best/tokenizer_config.json
[2024-06-26 18:08:12,135] [    INFO] - Special tokens file saved in ./checkpoint/model_best/special_tokens_map.json
[2024-06-26 18:08:12,142] [    INFO] - Configuration saved in ./checkpoint/model_best/config.json
[2024-06-26 18:08:13,814] [    INFO] - Model weights saved in ./checkpoint/model_best/model_state.pdparams
[2024-06-26 18:08:13,815] [    INFO] - ***** train metrics *****
[2024-06-26 18:08:13,815] [    INFO] -   progress_or_epoch        =      100.0
[2024-06-26 18:08:13,815] [    INFO] -   train_loss               =     0.0002
[2024-06-26 18:08:13,815] [    INFO] -   train_runtime            = 0:02:41.44
[2024-06-26 18:08:13,815] [    INFO] -   train_samples_per_second =     42.119
[2024-06-26 18:08:13,815] [    INFO] -   train_steps_per_second   =      3.097
[2024-06-26 18:08:13,820] [    INFO] - ***** Running Evaluation *****
[2024-06-26 18:08:13,820] [    INFO] -   Num examples = 4
[2024-06-26 18:08:13,820] [    INFO] -   Total prediction steps = 1
[2024-06-26 18:08:13,820] [    INFO] -   Pre device batch size = 16
[2024-06-26 18:08:13,820] [    INFO] -   Total Batch size = 16
[2024-06-26 18:08:13,888] [    INFO] - eval_loss: 0.005224619060754776, eval_precision: 0.4, eval_recall: 0.4, eval_f1: 0.4000000000000001, eval_runtime: 0.0687, eval_samples_per_second: 58.19804494272889, eval_steps_per_second: 14.549511235682223, progress_or_epoch: 100.0
[2024-06-26 18:08:13,889] [    INFO] - ***** eval metrics *****
[2024-06-26 18:08:13,889] [    INFO] -   eval_f1                 =        0.4
[2024-06-26 18:08:13,889] [    INFO] -   eval_loss               =     0.0052
[2024-06-26 18:08:13,889] [    INFO] -   eval_precision          =        0.4
[2024-06-26 18:08:13,889] [    INFO] -   eval_recall             =        0.4
[2024-06-26 18:08:13,889] [    INFO] -   eval_runtime            = 0:00:00.06
[2024-06-26 18:08:13,890] [    INFO] -   eval_samples_per_second =     58.198
[2024-06-26 18:08:13,890] [    INFO] -   eval_steps_per_second   =    14.5495
[2024-06-26 18:08:13,890] [    INFO] -   progress_or_epoch       =      100.0
/root/anaconda3/envs/py39_ppner_2_7_2/lib/python3.9/site-packages/paddle/jit/dy2static/program_translator.py:712: UserWarning: full_graph=False don't support input_spec arguments. It will not produce any effect.
You can set full_graph=True, then you can assign input spec.

  warnings.warn(
[2024-06-26 18:08:13,895] [    INFO] - Exporting inference model to ./checkpoint/model_best/model
I0626 18:08:15.824677 285242 program_interpreter.cc:212] New Executor is Running.
[2024-06-26 18:08:18,295] [    INFO] - Inference model exported.
[2024-06-26 18:08:18,297] [    INFO] - tokenizer config file saved in ./checkpoint/model_best/tokenizer_config.json
[2024-06-26 18:08:18,297] [    INFO] - Special tokens file saved in ./checkpoint/model_best/special_tokens_map.json
LAUNCH INFO 2024-06-26 18:08:20,044 Pod completed
LAUNCH INFO 2024-06-26 18:08:20,045 Exit code 0

模型文件

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
(py39_ppner_2_7_2) [root@jdz uie]# ll checkpoint/model_best/
total 919580
-rw-r--r-- 1 root root       207 Jun 26 18:08 all_results.json
drwxr-xr-x 2 root root       251 Jun 26 18:06 checkpoint-100
drwxr-xr-x 2 root root       251 Jun 26 18:06 checkpoint-200
drwxr-xr-x 2 root root       251 Jun 26 18:07 checkpoint-300
drwxr-xr-x 2 root root       251 Jun 26 18:07 checkpoint-400
drwxr-xr-x 2 root root       251 Jun 26 18:08 checkpoint-500
-rw-r--r-- 1 root root       559 Jun 26 18:08 config.json
-rw-r--r-- 1 root root 469428391 Jun 26 18:08 model.pdiparams
-rw-r--r-- 1 root root     17581 Jun 26 18:08 model.pdiparams.info
-rw-r--r-- 1 root root    153135 Jun 26 18:08 model.pdmodel
-rw-r--r-- 1 root root 471806543 Jun 26 18:12 model_state.pdparams
drwxr-xr-x 7 root root       136 Jun 26 18:05 runs
-rw-r--r-- 1 root root       112 Jun 26 18:08 special_tokens_map.json
drwxr-xr-x 2 root root        90 Jun 26 18:12 static
-rw-r--r-- 1 root root       197 Jun 26 18:08 tokenizer_config.json
-rw-r--r-- 1 root root     16736 Jun 26 18:08 trainer_state.json
-rw-r--r-- 1 root root      2598 Jun 26 18:08 training_args.bin
-rw-r--r-- 1 root root       207 Jun 26 18:08 train_results.json
-rw-r--r-- 1 root root    186807 Jun 26 18:08 vocab.txt
(py39_ppner_2_7_2) [root@jdz uie]#

调用api测试

1
2
3
4
5
6
7
8
9
from pprint import pprint
from paddlenlp import Taskflow

schema = [{'工程': ['工艺']}]

ie = Taskflow('information_extraction', schema=schema, task_path='./checkpoint/model_best')

ie.set_schema(schema) # Reset schema
pprint(ie("""受力钢筋的接头形式应按设计要求采用,若设计无要求时,钢筋宜采用焊接接头和机械连接接头,也可采用绑扎接头。"""))

测试结果

1
2
3
4
5
6
7
8
[{'工程': [{'end': 9,
          'probability': 0.9548929987860788,
          'relations': {'工艺': [{'end': 42,
                                'probability': 0.2548182944884658,
                                'start': 36,
                                'text': '机械连接接头'}]},
          'start': 0,
          'text': '受力钢筋的接头形式'}]}]

Powered by Hexo and Hexo-theme-hiker

Copyright © 2017 - 2024 青域 All Rights Reserved.

UV : | PV :