rubert-tiny2-russe-toxicity

This model is a fine-tuned version of cointegrated/rubert-tiny2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2087
• Precision Macro: 0.9395
• Recall Macro: 0.9394
• F1 Macro: 0.9394
• F1 Neutral: 0.9398
• F1 Toxic: 0.9390

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision Macro	Recall Macro	F1 Macro	F1 Neutral	F1 Toxic
0.6127	0.23	100	0.4477	0.8252	0.825	0.8250	0.8274	0.8226
0.3728	0.46	200	0.2852	0.8877	0.8875	0.8875	0.8886	0.8864
0.2888	0.69	300	0.2405	0.9039	0.9038	0.9037	0.9047	0.9028
0.2724	0.92	400	0.2288	0.9110	0.9094	0.9093	0.9122	0.9064
0.2387	1.15	500	0.2161	0.9132	0.9125	0.9125	0.9143	0.9106
0.2189	1.38	600	0.2098	0.9206	0.9206	0.9206	0.9208	0.9205
0.1783	1.61	700	0.2045	0.9233	0.9231	0.9231	0.9238	0.9224
0.1911	1.84	800	0.2098	0.9184	0.9175	0.9175	0.9155	0.9194
0.1749	2.07	900	0.1947	0.9271	0.9269	0.9269	0.9277	0.9260
0.1728	2.3	1000	0.1893	0.9263	0.9263	0.9262	0.9265	0.9260
0.1628	2.53	1100	0.1926	0.9276	0.9275	0.9275	0.9270	0.9280
0.1545	2.76	1200	0.1889	0.9299	0.9294	0.9294	0.9306	0.9281
0.1659	2.99	1300	0.1893	0.9263	0.9263	0.9263	0.9263	0.9263
0.1433	3.22	1400	0.1952	0.9264	0.9263	0.9262	0.9256	0.9269
0.1243	3.45	1500	0.1838	0.9300	0.93	0.9300	0.9303	0.9296
0.1357	3.68	1600	0.1846	0.9370	0.9356	0.9356	0.9374	0.9338
0.1419	3.91	1700	0.1793	0.9339	0.9331	0.9331	0.9345	0.9317
0.1205	4.14	1800	0.1896	0.9306	0.9306	0.9306	0.9306	0.9307
0.1159	4.37	1900	0.1905	0.9354	0.935	0.9350	0.9360	0.9340
0.1185	4.6	2000	0.1893	0.9400	0.9387	0.9387	0.9403	0.9371
0.102	4.83	2100	0.1944	0.9415	0.9406	0.9406	0.9419	0.9393
0.1222	5.06	2200	0.1880	0.9390	0.9387	0.9387	0.9394	0.9381
0.1006	5.29	2300	0.1964	0.9331	0.9331	0.9331	0.9330	0.9333
0.108	5.52	2400	0.1901	0.9382	0.9381	0.9381	0.9384	0.9379
0.1059	5.75	2500	0.1907	0.9401	0.94	0.9400	0.9404	0.9396
0.1033	5.98	2600	0.1905	0.9409	0.94	0.9400	0.9413	0.9386
0.0969	6.21	2700	0.1969	0.9345	0.9344	0.9344	0.9338	0.9349
0.0826	6.44	2800	0.1952	0.9407	0.9406	0.9406	0.9410	0.9402
0.1055	6.67	2900	0.1967	0.9375	0.9375	0.9375	0.9373	0.9377
0.0982	6.9	3000	0.2063	0.9394	0.9381	0.9381	0.9397	0.9364
0.0862	7.13	3100	0.2056	0.9414	0.9412	0.9412	0.9418	0.9407
0.0811	7.36	3200	0.2054	0.9414	0.9412	0.9412	0.9418	0.9407
0.0908	7.59	3300	0.2078	0.9394	0.9394	0.9394	0.9396	0.9392
0.092	7.82	3400	0.2003	0.9400	0.94	0.9400	0.9401	0.9399
0.0834	8.05	3500	0.2039	0.9394	0.9394	0.9394	0.9392	0.9396
0.0747	8.28	3600	0.2018	0.9413	0.9413	0.9412	0.9415	0.9410
0.0898	8.51	3700	0.2009	0.94	0.94	0.94	0.94	0.94
0.077	8.74	3800	0.2022	0.9410	0.9406	0.9406	0.9415	0.9398
0.087	8.97	3900	0.2055	0.9382	0.9381	0.9381	0.9385	0.9377
0.0647	9.2	4000	0.2062	0.9401	0.94	0.9400	0.9405	0.9395
0.0736	9.43	4100	0.2081	0.9389	0.9387	0.9387	0.9393	0.9382
0.08	9.66	4200	0.2087	0.9395	0.9394	0.9394	0.9399	0.9388
0.083	9.89	4300	0.2087	0.9395	0.9394	0.9394	0.9398	0.9390

Framework versions

• Transformers 4.28.1
• Pytorch 2.0.0+cu118
• Datasets 2.11.0
• Tokenizers 0.13.3