Terjman-Ultra-v2.2

This model is a fine-tuned version of facebook/nllb-200-1.3B on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 2.0034
• Bleu: 24.9877
• Chrf: 44.6972
• Ter: 77.1984
• Gen Len: 12.3388

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: 0.0005
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 32
• total_train_batch_size: 128
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Chrf	Ter	Gen Len
70.7656	0.0361	100	2.9806	12.9152	30.9913	98.5067	12.7812
52.8738	0.0723	200	2.7797	15.5994	34.6105	93.0509	12.5988
45.6954	0.1084	300	2.6691	17.7178	36.6882	88.7668	12.5506
39.9565	0.1446	400	2.5662	17.5642	37.6628	90.0519	12.7071
36.7684	0.1807	500	2.4275	19.4366	39.1214	86.0378	12.5859
33.4308	0.2169	600	2.3610	20.5894	40.0986	85.4547	12.5965
31.3199	0.2530	700	2.2796	20.063	39.6827	85.9359	12.7247
30.5014	0.2892	800	2.2517	20.6695	40.3914	85.2728	12.7459
29.2522	0.3253	900	2.2143	21.2474	41.3276	82.497	12.5588
27.6444	0.3615	1000	2.1680	21.6217	41.2203	84.6141	12.8859
26.1025	0.3976	1100	2.1871	21.9405	41.8135	83.7077	12.8235
25.8736	0.4338	1200	2.1521	23.0621	42.5788	82.0175	12.7929
24.5748	0.4699	1300	2.1207	22.9369	42.4532	80.0937	12.3553
24.0533	0.5061	1400	2.1252	22.9731	42.9912	80.8733	12.4365
23.8092	0.5422	1500	2.1173	22.1319	42.3421	81.1727	12.3988
23.2533	0.5783	1600	2.1203	22.9435	42.7683	80.9051	12.4294
22.5494	0.6145	1700	2.1041	23.5743	43.1741	78.8658	12.44
22.4968	0.6506	1800	2.1143	24.0091	43.7091	79.2722	12.4471
21.8837	0.6868	1900	2.1086	23.5233	43.1374	83.0314	12.5106
21.4213	0.7229	2000	2.0802	23.4877	43.2273	80.5064	12.4506
21.0951	0.7591	2100	2.0604	23.8738	43.2852	83.3723	12.5212
20.861	0.7952	2200	2.0520	23.8624	43.4149	79.3725	12.46
20.164	0.8314	2300	2.0542	24.3772	43.6011	77.7732	12.4059
20.4746	0.8675	2400	2.0430	23.9582	43.1838	84.7154	12.5318
19.9062	0.9037	2500	2.0450	23.6104	43.4165	79.6449	12.4435
19.6233	0.9398	2600	2.0134	24.3974	44.2913	78.2905	12.3906
19.2435	0.9760	2700	2.0239	24.2622	44.0087	77.9154	12.3341
17.9055	1.0119	2800	2.0400	23.8797	43.6364	78.3719	12.3059
17.3272	1.0481	2900	2.0592	24.2694	43.7129	77.6987	12.3494
17.3509	1.0842	3000	2.0404	24.202	43.8023	78.4414	12.3882
17.4569	1.1204	3100	2.0216	24.254	43.937	77.67	12.4412
17.1146	1.1565	3200	2.0148	24.2023	44.4105	78.7747	12.4565
16.6774	1.1927	3300	2.0354	23.7494	43.796	79.1691	12.3871
17.0017	1.2288	3400	2.0189	24.2081	44.1283	77.7256	12.4047
16.6622	1.2650	3500	2.0140	24.217	44.2066	77.9338	12.38
16.5509	1.3011	3600	2.0312	24.3622	44.2009	78.341	12.4424
16.4062	1.3372	3700	2.0121	24.3472	44.0205	77.7659	12.4012
16.5797	1.3734	3800	2.0434	23.9545	43.8269	78.3792	12.4188
16.3888	1.4095	3900	2.0162	23.7386	43.79	78.9535	12.4576
16.0585	1.4457	4000	2.0166	23.9605	43.8616	79.1566	12.42
16.4091	1.4818	4100	2.0234	24.2887	44.2176	78.4537	12.4176
16.2554	1.5180	4200	2.0220	24.278	44.3497	77.9841	12.4553
16.1241	1.5541	4300	1.9927	24.9649	44.8357	77.048	12.3647
16.0068	1.5903	4400	2.0072	25.0727	44.8539	77.268	12.3729
15.7749	1.6264	4500	2.0135	24.4552	44.5047	78.7697	12.4294
15.9322	1.6626	4600	2.0011	24.7077	44.3159	78.0441	12.4153
15.8415	1.6987	4700	2.0028	24.2381	44.0753	78.4017	12.4247
15.544	1.7349	4800	2.0001	24.6574	44.4939	77.7002	12.4094
15.8296	1.7710	4900	1.9900	25.0563	45.2428	77.0852	12.3988
15.4856	1.8072	5000	1.9845	24.8863	44.87	77.323	12.3518
15.3165	1.8433	5100	1.9749	24.5669	44.8923	77.8361	12.3906
15.8505	1.8795	5200	1.9879	24.6842	44.8395	78.055	12.4165
15.2058	1.9156	5300	1.9728	24.8448	44.7121	77.4636	12.3529
15.1863	1.9517	5400	1.9831	25.0569	44.8317	77.37	12.3659
15.3051	1.9879	5500	1.9772	25.0335	45.0737	77.0836	12.3494
14.3992	2.0239	5600	2.0031	25.2894	45.2749	76.6897	12.3541
14.2126	2.0600	5700	1.9986	25.1707	45.1943	77.2161	12.3741
14.5549	2.0962	5800	2.0035	25.2928	45.129	77.1665	12.3576
14.4486	2.1323	5900	1.9989	25.0251	44.9919	77.573	12.3459
14.3505	2.1684	6000	1.9978	25.1946	44.9452	77.4272	12.3365
14.5783	2.2046	6100	2.0078	25.0654	44.8266	77.6317	12.3765
14.2261	2.2407	6200	2.0079	25.0622	44.788	77.5574	12.3729
14.3163	2.2769	6300	2.0104	25.0457	44.7239	78.0145	12.4
14.4612	2.3130	6400	2.0035	25.184	45.1284	77.4162	12.3741
14.1588	2.3492	6500	2.0060	24.9345	44.7316	77.8871	12.3824
14.3309	2.3853	6600	2.0102	25.2199	45.0809	77.634	12.3859
14.0311	2.4215	6700	2.0017	25.1946	44.8508	77.3804	12.3565
14.0974	2.4576	6800	2.0046	25.005	44.902	78.0266	12.3882
14.2057	2.4938	6900	2.0035	24.9742	45.0221	78.0698	12.3894
14.4113	2.5299	7000	2.0027	25.2538	45.0121	76.9214	12.3329
14.4146	2.5661	7100	2.0040	25.0706	44.7691	77.1347	12.3565
13.8906	2.6022	7200	2.0076	25.0349	44.6927	77.2741	12.3788
14.5977	2.6384	7300	2.0051	25.0629	44.7696	77.1818	12.3776
14.2155	2.6745	7400	2.0059	25.0825	44.8729	77.0082	12.36
14.0901	2.7106	7500	2.0040	25.125	44.8754	76.8288	12.3588
14.1026	2.7468	7600	2.0045	24.9539	44.6773	77.1001	12.3365
14.5714	2.7829	7700	2.0049	25.0891	44.7547	76.9704	12.3494
14.0298	2.8191	7800	2.0036	25.1397	44.8099	76.9669	12.3565
14.1156	2.8552	7900	2.0033	24.9858	44.6656	77.0585	12.3365
14.032	2.8914	8000	2.0036	25.0805	44.6327	77.1592	12.3471
14.0703	2.9275	8100	2.0033	24.9824	44.5966	77.2585	12.3471
14.157	2.9637	8200	2.0034	24.9877	44.6972	77.1984	12.3388

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.21.0