Edit model card

deberta-v3-large-sentiment

This model is a fine-tuned version of microsoft/deberta-v3-large on an tweet_eval dataset.

Model description

Test set results:

Model Emotion Hate Irony Offensive Sentiment
deberta-v3-large 86.3 61.3 87.1 86.4 73.9
BERTweet 79.3 - 82.1 79.5 73.4
RoB-RT 79.5 52.3 61.7 80.5 69.3

source:papers_with_code

Intended uses & limitations

Classifying attributes of interest on tweeter like data.

Training and evaluation data

tweet_eval dataset.

Training procedure

Fine tuned and evaluated with run_glue.py

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 5e-06
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 50
  • num_epochs: 10.0

Training results

Training Loss Epoch Step Validation Loss Accuracy
1.0614 0.07 100 1.0196 0.4345
0.8601 0.14 200 0.7561 0.6460
0.734 0.21 300 0.6796 0.6955
0.6753 0.28 400 0.6521 0.7000
0.6408 0.35 500 0.6119 0.7440
0.5991 0.42 600 0.6034 0.7370
0.6069 0.49 700 0.5976 0.7375
0.6122 0.56 800 0.5871 0.7425
0.5908 0.63 900 0.5935 0.7445
0.5884 0.7 1000 0.5792 0.7520
0.5839 0.77 1100 0.5780 0.7555
0.5772 0.84 1200 0.5727 0.7570
0.5895 0.91 1300 0.5601 0.7550
0.5757 0.98 1400 0.5613 0.7525
0.5121 1.05 1500 0.5867 0.7600
0.5254 1.12 1600 0.5595 0.7630
0.5074 1.19 1700 0.5594 0.7585
0.4947 1.26 1800 0.5697 0.7575
0.5019 1.33 1900 0.5665 0.7580
0.5005 1.4 2000 0.5484 0.7655
0.5125 1.47 2100 0.5626 0.7605
0.5241 1.54 2200 0.5561 0.7560
0.5198 1.61 2300 0.5602 0.7600
0.5124 1.68 2400 0.5654 0.7490
0.5096 1.75 2500 0.5803 0.7515
0.4885 1.82 2600 0.5889 0.75
0.5111 1.89 2700 0.5508 0.7665
0.4868 1.96 2800 0.5621 0.7635
0.4599 2.04 2900 0.5995 0.7615
0.4147 2.11 3000 0.6202 0.7530
0.4233 2.18 3100 0.5875 0.7625
0.4324 2.25 3200 0.5794 0.7610
0.4141 2.32 3300 0.5902 0.7460
0.4306 2.39 3400 0.6053 0.7545
0.4266 2.46 3500 0.5979 0.7570
0.4227 2.53 3600 0.5920 0.7650
0.4226 2.6 3700 0.6166 0.7455
0.3978 2.67 3800 0.6126 0.7560
0.3954 2.74 3900 0.6152 0.7550
0.4209 2.81 4000 0.5980 0.75
0.3982 2.88 4100 0.6096 0.7490
0.4016 2.95 4200 0.6541 0.7425
0.3966 3.02 4300 0.6377 0.7545
0.3074 3.09 4400 0.6860 0.75
0.3551 3.16 4500 0.6160 0.7550
0.3323 3.23 4600 0.6714 0.7520
0.3171 3.3 4700 0.6538 0.7535
0.3403 3.37 4800 0.6774 0.7465
0.3396 3.44 4900 0.6726 0.7465
0.3259 3.51 5000 0.6465 0.7480
0.3392 3.58 5100 0.6860 0.7460
0.3251 3.65 5200 0.6697 0.7495
0.3253 3.72 5300 0.6770 0.7430
0.3455 3.79 5400 0.7177 0.7360
0.3323 3.86 5500 0.6943 0.7400
0.3335 3.93 5600 0.6507 0.7555
0.3368 4.0 5700 0.6580 0.7485
0.2479 4.07 5800 0.7667 0.7430
0.2613 4.14 5900 0.7513 0.7505
0.2557 4.21 6000 0.7927 0.7485
0.243 4.28 6100 0.7792 0.7450
0.2473 4.35 6200 0.8107 0.7355
0.2447 4.42 6300 0.7851 0.7370
0.2515 4.49 6400 0.7529 0.7465
0.274 4.56 6500 0.7390 0.7465
0.2674 4.63 6600 0.7658 0.7460
0.2416 4.7 6700 0.7915 0.7485
0.2432 4.77 6800 0.7989 0.7435
0.2595 4.84 6900 0.7850 0.7380
0.2736 4.91 7000 0.7577 0.7395
0.2783 4.98 7100 0.7650 0.7405
0.2304 5.05 7200 0.8542 0.7385
0.1937 5.12 7300 0.8390 0.7345
0.1878 5.19 7400 0.9150 0.7330
0.1921 5.26 7500 0.8792 0.7405
0.1916 5.33 7600 0.8892 0.7410
0.2011 5.4 7700 0.9012 0.7325
0.211 5.47 7800 0.8608 0.7420
0.2194 5.54 7900 0.8852 0.7320
0.205 5.61 8000 0.8803 0.7385
0.1981 5.68 8100 0.8681 0.7330
0.1908 5.75 8200 0.9020 0.7435
0.1942 5.82 8300 0.8780 0.7410
0.1958 5.89 8400 0.8937 0.7345
0.1883 5.96 8500 0.9121 0.7360
0.1819 6.04 8600 0.9409 0.7430
0.145 6.11 8700 1.1390 0.7265
0.1696 6.18 8800 0.9189 0.7430
0.1488 6.25 8900 0.9718 0.7400
0.1637 6.32 9000 0.9702 0.7450
0.1547 6.39 9100 1.0033 0.7410
0.1605 6.46 9200 0.9973 0.7355
0.1552 6.53 9300 1.0491 0.7290
0.1731 6.6 9400 1.0271 0.7335
0.1738 6.67 9500 0.9575 0.7430
0.1669 6.74 9600 0.9614 0.7350
0.1347 6.81 9700 1.0263 0.7365
0.1593 6.88 9800 1.0173 0.7360
0.1549 6.95 9900 1.0398 0.7350
0.1675 7.02 10000 0.9975 0.7380
0.1182 7.09 10100 1.1059 0.7350
0.1351 7.16 10200 1.0933 0.7400
0.1496 7.23 10300 1.0731 0.7355
0.1197 7.3 10400 1.1089 0.7360
0.1111 7.37 10500 1.1381 0.7405
0.1494 7.44 10600 1.0252 0.7425
0.1235 7.51 10700 1.0906 0.7360
0.133 7.58 10800 1.1796 0.7375
0.1248 7.65 10900 1.1332 0.7420
0.1268 7.72 11000 1.1304 0.7415
0.1368 7.79 11100 1.1345 0.7380
0.1228 7.86 11200 1.2018 0.7320
0.1281 7.93 11300 1.1884 0.7350
0.1449 8.0 11400 1.1571 0.7345
0.1025 8.07 11500 1.1538 0.7345
0.1199 8.14 11600 1.2113 0.7390
0.1016 8.21 11700 1.2882 0.7370
0.114 8.28 11800 1.2872 0.7390
0.1019 8.35 11900 1.2876 0.7380
0.1142 8.42 12000 1.2791 0.7385
0.1135 8.49 12100 1.2883 0.7380
0.1139 8.56 12200 1.2829 0.7360
0.1107 8.63 12300 1.2698 0.7365
0.1183 8.7 12400 1.2660 0.7345
0.1064 8.77 12500 1.2889 0.7365
0.0895 8.84 12600 1.3480 0.7330
0.1244 8.91 12700 1.2872 0.7325
0.1209 8.98 12800 1.2681 0.7375
0.1144 9.05 12900 1.2711 0.7370
0.1034 9.12 13000 1.2801 0.7360
0.113 9.19 13100 1.2801 0.7350
0.0994 9.26 13200 1.2920 0.7360
0.0966 9.33 13300 1.2761 0.7335
0.0939 9.4 13400 1.2909 0.7365
0.0975 9.47 13500 1.2953 0.7360
0.0842 9.54 13600 1.3179 0.7335
0.0871 9.61 13700 1.3149 0.7385
0.1162 9.68 13800 1.3124 0.7350
0.085 9.75 13900 1.3207 0.7355
0.0966 9.82 14000 1.3248 0.7335
0.1064 9.89 14100 1.3261 0.7335
0.1046 9.96 14200 1.3255 0.7360

Framework versions

  • Transformers 4.20.0.dev0
  • Pytorch 1.9.0
  • Datasets 2.2.2
  • Tokenizers 0.11.6
Downloads last month
19
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.