|
- Unparsed args: ['--lstm_layer', '2']
- fix_gaz_emb : True
- fix_entity_emb : True
- contrast : False
- number_layer : 1
- bilstm_flag : True
- gat_nhidden : 300
- gat_nhead : 4
- gat_layer : 2
- pre_model : BERT
- ner_encode_type : lstm
- strategy : n
- model_type : lstm
- entity : True
- use_bert : True
- attention : True
- tri_fuse : False
- alpha : 0.1
- dropout : 0.6
- droplstm : 0.0
- dropbert : 0.1
- dropgat : 0.0
- gaz_dropout : 0.2
- kg_dropout : 0.2
- use_lstm : False
- dataset_name : meizhou2
- train_file : data/train_ch.txt
- test_file : data/train_ch.txt
- dev_file : data/train_ch.txt
- gaz_file : None
- kg_file : None
- char_embedding_path : None
- data_stored_directory : data/generated_data/
- param_stored_directory : data/model_param/
- entity_type_embdding_path : None
- language : CH
- norm_char_emb : True
- norm_gaz_emb : True
- norm_kg_emb : False
- number_normalized : True
- max_sentence_length : 200
- batch_size : 80
- max_epoch : 8
- lr : 2e-05
- lr_other : 0.0001
- lr_decay : 0.005
- use_clip : True
- clip : 5.0
- gradient_accumulation_steps : 1
- optimizer : Adam
- l2_penalty : 5e-08
- refresh : False
- use_gpu : True
- visible_gpu : 0
- random_seed : 1122
- Gaz file is None, load nothing
- kg file is None, load nothing
- +++++++++++++++++
- 31991
- ********************
- 28791
- +++++++++++++++++
- 31991
- build entity pretrain emb...
- Embedding: None
- pretrain num:0, prefect match:0, case_match:0, oov:2, oov%:1.0
- build entity id...
- DATA SUMMARY START:
- Dataset name: meizhou2
- Tag scheme: BIO
- Max Sentence Length: 200
- Char alphabet size: 2066
- Gaz alphabet size: 2
- Label alphabet size: 18
- Char embedding size: 200
- Gaz embedding size: 100
- Number normalized: True
- Norm char emb: True
- Norm gaz emb: True
- Train instance number: 28791
- Dev instance number: 3200
- Test instance number: 31991
- Train cut number: 0
- Dev cut number: 0
- Test cut number: 0
- DATA SUMMARY END.
- Data setting saved to file: data/generated_data/meizhou2_dataset.dset
- build BLSTM_GCN_CRF model...
- using model lstm......
- Epoch: 0/8
- Learning rate is setted as: 2e-05
- Instance: 2000; Time: 141.64s; loss: 199860.9004
- Instance: 4000; Time: 142.81s; loss: 39976.8633
- Instance: 6000; Time: 141.88s; loss: 15983.0391
- Instance: 8000; Time: 138.36s; loss: 11318.1230
- Instance: 10000; Time: 140.15s; loss: 8829.7715
- Instance: 12000; Time: 141.17s; loss: 7679.9570
- Instance: 14000; Time: 245.13s; loss: 8640.3672
- Instance: 16000; Time: 243.02s; loss: 7533.3242
- Instance: 18000; Time: 247.71s; loss: 6240.6543
- Instance: 20000; Time: 249.88s; loss: 6633.7930
- Instance: 22000; Time: 244.29s; loss: 6350.1348
- Instance: 24000; Time: 244.72s; loss: 5716.4219
- Instance: 26000; Time: 252.25s; loss: 7500.5625
- Instance: 28000; Time: 196.59s; loss: 6531.4141
- Instance: 28791; Time: 55.50s; loss: 2736.0039
- Epoch: 0 training finished. Time: 2825.10s, speed: 10.19st/s, total loss: 341531.330078125
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 3200
- gold_num = 5318 pred_num = 5381 right_num = 4813
- Dev: time: 173.02s, speed: 18.51st/s; acc: 0.9779, p: 0.8944, r: 0.9050, f: 0.8997
- Exceed previous best f score: -1
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 31991
- gold_num = 53249 pred_num = 53911 right_num = 48593
- Test: speed: 21.95st/s; acc: 0.9803, p: 0.9014, r: 0.9126, f: 0.9069
- Epoch: 1/8
- Learning rate is setted as: 1.9900000000000003e-05
- Instance: 2000; Time: 146.88s; loss: 5543.7031
- Instance: 4000; Time: 144.02s; loss: 5295.7422
- Instance: 6000; Time: 147.63s; loss: 5186.9727
- Instance: 8000; Time: 145.27s; loss: 6350.7891
- Instance: 10000; Time: 142.79s; loss: 5850.3008
- Instance: 12000; Time: 147.47s; loss: 5875.8516
- Instance: 14000; Time: 146.48s; loss: 4977.0312
- Instance: 16000; Time: 144.45s; loss: 5206.4961
- Instance: 18000; Time: 136.22s; loss: 5317.0996
- Instance: 20000; Time: 143.61s; loss: 5815.6719
- Instance: 22000; Time: 143.21s; loss: 5483.8574
- Instance: 24000; Time: 139.14s; loss: 4798.3789
- Instance: 26000; Time: 146.97s; loss: 5484.0820
- Instance: 28000; Time: 145.63s; loss: 4868.5273
- Instance: 28791; Time: 58.69s; loss: 2198.4375
- Epoch: 1 training finished. Time: 2078.46s, speed: 13.85st/s, total loss: 78252.94140625
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 3200
- gold_num = 5318 pred_num = 5421 right_num = 4881
- Dev: time: 173.88s, speed: 18.41st/s; acc: 0.9804, p: 0.9004, r: 0.9178, f: 0.9090
- Exceed previous best f score: 0.8997102532947004
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 31991
- gold_num = 53249 pred_num = 53947 right_num = 49527
- Test: speed: 22.19st/s; acc: 0.9843, p: 0.9181, r: 0.9301, f: 0.9240
- Epoch: 2/8
- Learning rate is setted as: 1.98005e-05
- Instance: 2000; Time: 144.51s; loss: 4425.7363
- Instance: 4000; Time: 145.64s; loss: 4242.8320
- Instance: 6000; Time: 145.17s; loss: 3645.6055
- Instance: 8000; Time: 145.39s; loss: 4536.0547
- Instance: 10000; Time: 142.12s; loss: 3948.2734
- Instance: 12000; Time: 142.66s; loss: 4486.8164
- Instance: 14000; Time: 140.74s; loss: 5224.7773
- Instance: 16000; Time: 146.03s; loss: 4202.8008
- Instance: 18000; Time: 144.23s; loss: 3781.5859
- Instance: 20000; Time: 145.77s; loss: 4679.0625
- Instance: 22000; Time: 142.87s; loss: 4320.5078
- Instance: 24000; Time: 147.20s; loss: 4811.2109
- Instance: 26000; Time: 143.51s; loss: 4387.2773
- Instance: 28000; Time: 145.43s; loss: 4680.3516
- Instance: 28791; Time: 56.41s; loss: 1347.7227
- Epoch: 2 training finished. Time: 2077.70s, speed: 13.86st/s, total loss: 62720.615234375
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 3200
- gold_num = 5318 pred_num = 5412 right_num = 4911
- Dev: time: 172.45s, speed: 18.57st/s; acc: 0.9811, p: 0.9074, r: 0.9235, f: 0.9154
- Exceed previous best f score: 0.9090231865164354
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 31991
- gold_num = 53249 pred_num = 54383 right_num = 50170
- Test: speed: 22.17st/s; acc: 0.9863, p: 0.9225, r: 0.9422, f: 0.9323
- Epoch: 3/8
- Learning rate is setted as: 1.97014975e-05
- Instance: 2000; Time: 144.28s; loss: 3710.2109
- Instance: 4000; Time: 145.81s; loss: 3395.3281
- Instance: 6000; Time: 147.02s; loss: 3908.8047
- Instance: 8000; Time: 144.75s; loss: 3330.5508
- Instance: 10000; Time: 147.22s; loss: 4760.4414
- Instance: 12000; Time: 148.68s; loss: 3616.5312
- Instance: 14000; Time: 146.58s; loss: 3569.9531
- Instance: 16000; Time: 145.70s; loss: 3991.0742
- Instance: 18000; Time: 148.02s; loss: 3815.6797
- Instance: 20000; Time: 145.30s; loss: 3615.4062
- Instance: 22000; Time: 134.96s; loss: 4148.2070
- Instance: 24000; Time: 145.78s; loss: 3436.9258
- Instance: 26000; Time: 143.66s; loss: 3751.6914
- Instance: 28000; Time: 143.92s; loss: 4164.1172
- Instance: 28791; Time: 58.28s; loss: 1417.1992
- Epoch: 3 training finished. Time: 2089.96s, speed: 13.78st/s, total loss: 54632.12109375
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 3200
- gold_num = 5318 pred_num = 5465 right_num = 4930
- Dev: time: 171.92s, speed: 18.62st/s; acc: 0.9798, p: 0.9021, r: 0.9270, f: 0.9144
- Epoch: 4/8
- Learning rate is setted as: 1.9602990012500002e-05
- Instance: 2000; Time: 140.44s; loss: 3196.7070
- Instance: 4000; Time: 144.42s; loss: 2834.2500
- Instance: 6000; Time: 146.85s; loss: 3319.5039
- Instance: 8000; Time: 144.44s; loss: 3207.5469
- Instance: 10000; Time: 145.72s; loss: 3401.5547
- Instance: 12000; Time: 145.14s; loss: 3285.8594
- Instance: 14000; Time: 147.23s; loss: 3582.6289
- Instance: 16000; Time: 146.30s; loss: 3465.7539
- Instance: 18000; Time: 137.99s; loss: 3674.1016
- Instance: 20000; Time: 141.17s; loss: 2832.6250
- Instance: 22000; Time: 143.89s; loss: 3754.0508
- Instance: 24000; Time: 145.83s; loss: 3083.0703
- Instance: 26000; Time: 141.97s; loss: 3570.8750
- Instance: 28000; Time: 143.26s; loss: 3924.1211
- Instance: 28791; Time: 58.42s; loss: 1189.4062
- Epoch: 4 training finished. Time: 2073.07s, speed: 13.89st/s, total loss: 48322.0546875
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 3200
- gold_num = 5318 pred_num = 5516 right_num = 4952
- Dev: time: 173.74s, speed: 18.43st/s; acc: 0.9802, p: 0.8978, r: 0.9312, f: 0.9142
- Epoch: 5/8
- Learning rate is setted as: 1.95049750624375e-05
- Instance: 2000; Time: 142.47s; loss: 3394.5430
- Instance: 4000; Time: 145.88s; loss: 2807.8672
- Instance: 6000; Time: 139.25s; loss: 2718.2617
- Instance: 8000; Time: 144.29s; loss: 2946.4688
- Instance: 10000; Time: 144.46s; loss: 3193.0859
- Instance: 12000; Time: 145.30s; loss: 2635.9062
- Instance: 14000; Time: 139.55s; loss: 3757.5156
- Instance: 16000; Time: 145.64s; loss: 3250.6445
- Instance: 18000; Time: 142.98s; loss: 2971.8711
- Instance: 20000; Time: 138.15s; loss: 2936.1523
- Instance: 22000; Time: 136.72s; loss: 2784.7773
- Instance: 24000; Time: 142.14s; loss: 2855.4805
- Instance: 26000; Time: 142.55s; loss: 3038.8359
- Instance: 28000; Time: 143.44s; loss: 2963.4648
- Instance: 28791; Time: 54.92s; loss: 1365.7305
- Epoch: 5 training finished. Time: 2047.74s, speed: 14.06st/s, total loss: 43620.60546875
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 3200
- gold_num = 5318 pred_num = 5410 right_num = 4899
- Dev: time: 170.85s, speed: 18.74st/s; acc: 0.9803, p: 0.9055, r: 0.9212, f: 0.9133
- Epoch: 6/8
- Learning rate is setted as: 1.9407450187125313e-05
- Instance: 2000; Time: 146.30s; loss: 2602.5820
- Instance: 4000; Time: 144.28s; loss: 2706.0391
- Instance: 6000; Time: 145.28s; loss: 2810.6953
- Instance: 8000; Time: 144.14s; loss: 2949.9141
- Instance: 10000; Time: 143.02s; loss: 2380.1484
- Instance: 12000; Time: 144.13s; loss: 2550.6406
- Instance: 14000; Time: 137.63s; loss: 3093.6133
- Instance: 16000; Time: 140.22s; loss: 2510.4844
- Instance: 18000; Time: 144.50s; loss: 2561.6797
- Instance: 20000; Time: 141.67s; loss: 2929.8555
- Instance: 22000; Time: 143.59s; loss: 2562.2188
- Instance: 24000; Time: 144.56s; loss: 2955.7422
- Instance: 26000; Time: 138.52s; loss: 3390.7461
- Instance: 28000; Time: 141.73s; loss: 2496.7656
- Instance: 28791; Time: 57.64s; loss: 1127.0664
- Epoch: 6 training finished. Time: 2057.22s, speed: 14.00st/s, total loss: 39628.19140625
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 3200
- gold_num = 5318 pred_num = 5462 right_num = 4923
- Dev: time: 171.62s, speed: 18.66st/s; acc: 0.9799, p: 0.9013, r: 0.9257, f: 0.9134
- Epoch: 7/8
- Learning rate is setted as: 1.9310412936189687e-05
- Instance: 2000; Time: 143.74s; loss: 2317.4453
- Instance: 4000; Time: 143.68s; loss: 2887.3438
- Instance: 6000; Time: 145.25s; loss: 2296.8789
- Instance: 8000; Time: 141.36s; loss: 2952.8164
- Instance: 10000; Time: 136.37s; loss: 2320.1289
- Instance: 12000; Time: 141.02s; loss: 2443.2344
- Instance: 14000; Time: 144.93s; loss: 2814.8633
- Instance: 16000; Time: 143.24s; loss: 2567.0352
- Instance: 18000; Time: 138.27s; loss: 2265.0820
- Instance: 20000; Time: 135.68s; loss: 2339.2930
- Instance: 22000; Time: 143.74s; loss: 2939.8594
- Instance: 24000; Time: 141.27s; loss: 2691.2617
- Instance: 26000; Time: 143.31s; loss: 2184.6445
- Instance: 28000; Time: 143.52s; loss: 3003.6758
- Instance: 28791; Time: 57.43s; loss: 929.7617
- Epoch: 7 training finished. Time: 2042.79s, speed: 14.09st/s, total loss: 36953.32421875
- $$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$
- 3200
- gold_num = 5318 pred_num = 5454 right_num = 4928
- Dev: time: 172.08s, speed: 18.61st/s; acc: 0.9798, p: 0.9036, r: 0.9267, f: 0.9150
|