Terjman-Ultra-v2.2-bs-4-lr-0.005-ep-3-wp-0.1-gacc-32-gnm-1.0-mx-512-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: 2.1788
• Bleu: 22.9949
• Chrf: 42.7982
• Ter: 81.779
• Gen Len: 12.4471

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: 0.005
• 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
43.8491	0.0361	100	2.5695	17.2289	37.3828	89.9572	12.6812
37.8172	0.0723	200	2.4585	18.2812	37.9854	88.7129	12.6447
41.0438	0.1084	300	2.5096	18.8162	38.8708	87.3215	12.5471
42.4993	0.1446	400	2.5739	17.0759	37.6617	92.179	12.7612
46.4173	0.1807	500	2.5927	17.7493	37.0782	90.7712	12.7153
50.3046	0.2169	600	2.7970	15.8428	34.6685	92.0702	12.5482
51.0591	0.2530	700	2.7518	13.7852	32.3087	122.4174	12.7729
53.5123	0.2892	800	2.8505	15.0263	33.5287	103.083	13.0224
54.1577	0.3253	900	2.8056	15.0231	33.3181	111.4563	13.2682
51.5253	0.3615	1000	2.7933	15.0815	32.7664	113.3758	13.4812
48.9478	0.3976	1100	2.7655	16.2933	34.243	94.6244	12.8918
47.938	0.4338	1200	2.7612	16.456	34.2244	100.909	13.7153
45.326	0.4699	1300	2.6848	16.6214	34.6197	98.8592	12.8941
43.4495	0.5061	1400	2.6833	16.4706	34.2734	93.3621	12.7494
42.9919	0.5422	1500	2.6799	17.0703	34.625	97.7785	12.7
41.6122	0.5783	1600	2.6150	17.075	34.9678	95.8607	12.8212
39.7837	0.6145	1700	2.5757	18.1852	35.6426	95.2936	13.0435
39.4488	0.6506	1800	2.6387	17.3216	36.3273	101.5896	13.5753
37.5611	0.6868	1900	2.6270	17.7145	35.9937	88.7923	12.7388
35.9422	0.7229	2000	2.6597	17.6019	36.3482	92.8667	12.8165
35.7019	0.7591	2100	2.5452	17.7785	36.4548	106.5095	13.5741
34.3932	0.7952	2200	2.5148	15.9903	34.5921	98.0078	12.8953
33.1799	0.8314	2300	2.4486	18.1216	35.6711	87.5631	12.2859
32.7152	0.8675	2400	2.4996	18.7002	37.2594	90.4309	12.8471
31.5079	0.9037	2500	2.4359	18.1631	37.2894	88.5531	12.5518
30.6564	0.9398	2600	2.4656	18.9625	38.9211	88.7146	12.7259
29.5573	0.9760	2700	2.4242	18.6898	37.6113	88.1869	12.4588
27.8906	1.0119	2800	2.4335	19.9443	38.6642	95.4597	13.2976
26.8031	1.0481	2900	2.4132	20.279	39.0184	87.0889	12.6753
26.3385	1.0842	3000	2.4161	19.4296	38.2119	87.4356	12.5035
26.1276	1.1204	3100	2.4171	19.6523	37.982	86.3196	12.7235
25.269	1.1565	3200	2.4185	19.2776	38.4189	88.6809	12.7
24.6057	1.1927	3300	2.4177	18.9888	37.8358	86.3955	12.4953
24.2829	1.2288	3400	2.3645	19.4833	38.1572	88.7418	12.6929
23.7796	1.2650	3500	2.3611	19.9181	38.7512	91.6993	12.9671
23.3656	1.3011	3600	2.4021	20.3132	39.1634	85.3322	12.4976
22.9616	1.3372	3700	2.3549	19.5902	38.7572	86.4413	12.5306
22.6219	1.3734	3800	2.3330	20.4724	40.3681	85.8608	12.5459
22.1655	1.4095	3900	2.2768	21.042	40.3518	84.8207	12.4212
21.465	1.4457	4000	2.2969	21.1681	39.6366	85.3886	12.4294
21.3829	1.4818	4100	2.3233	20.0998	39.8767	85.702	12.4871
21.2687	1.5180	4200	2.2875	21.3475	40.1695	85.7872	12.6141
20.4221	1.5541	4300	2.2167	21.6864	40.2743	83.9404	12.4353
20.4634	1.5903	4400	2.2713	20.9198	40.4753	86.215	12.5776
19.923	1.6264	4500	2.2585	20.287	40.0784	84.9928	12.4341
19.5794	1.6626	4600	2.2550	20.648	40.5857	86.0858	12.7694
19.4394	1.6987	4700	2.2104	21.9344	41.1899	83.1	12.4929
18.6122	1.7349	4800	2.2093	22.767	41.4871	82.5461	12.6235
18.6992	1.7710	4900	2.2190	20.9887	41.0	85.8985	12.8235
18.2776	1.8072	5000	2.2146	20.9857	41.2827	83.8972	12.4824
17.6287	1.8433	5100	2.1996	21.9022	41.5448	85.0591	12.7012
17.9071	1.8795	5200	2.2085	22.0738	41.5302	82.5875	12.5071
16.8171	1.9156	5300	2.2099	21.7048	41.25	82.833	12.5141
16.9345	1.9517	5400	2.1895	22.0361	41.4987	82.2035	12.5094
16.7839	1.9879	5500	2.1839	21.9444	41.5776	85.0539	12.7247
13.937	2.0239	5600	2.2415	21.8045	41.7823	82.9828	12.48
13.8213	2.0600	5700	2.2454	21.8772	41.4813	83.3942	12.5624
14.1685	2.0962	5800	2.2389	21.463	41.9267	84.3062	12.5435
13.8483	2.1323	5900	2.2024	22.6018	41.8719	81.5109	12.3553
13.5261	2.1684	6000	2.2463	22.011	41.9225	82.6643	12.4235
13.6918	2.2046	6100	2.2273	22.7619	42.3633	81.7888	12.4012
13.2274	2.2407	6200	2.2184	22.3117	41.7462	82.1613	12.4035
13.2595	2.2769	6300	2.2420	22.554	41.7048	82.5436	12.4624
13.229	2.3130	6400	2.2358	22.8454	42.2844	82.2621	12.5412
12.8033	2.3492	6500	2.2154	23.1289	42.5379	81.5014	12.4094
12.8587	2.3853	6600	2.2180	21.9732	41.935	83.0024	12.5071
12.4347	2.4215	6700	2.2110	22.446	42.1488	81.8244	12.4976
12.4909	2.4576	6800	2.1824	22.3089	42.3126	81.7373	12.4812
12.4221	2.4938	6900	2.1763	22.7566	42.4346	82.1514	12.5341
12.3757	2.5299	7000	2.2033	22.7765	42.3882	84.5552	12.7024
12.1889	2.5661	7100	2.1985	23.0604	42.4498	81.9895	12.4859
11.8595	2.6022	7200	2.1895	22.7264	42.4201	81.7662	12.5082
12.3815	2.6384	7300	2.1769	22.531	42.608	82.6642	12.5141
12.107	2.6745	7400	2.1906	22.9567	42.8802	81.617	12.44
11.7338	2.7106	7500	2.1932	22.8328	42.879	81.8005	12.4553
11.6615	2.7468	7600	2.1772	22.9404	42.6161	81.7249	12.4271
12.2163	2.7829	7700	2.1904	22.9762	42.8218	81.7552	12.4565
11.766	2.8191	7800	2.1768	22.8033	42.8082	82.034	12.4082
11.5269	2.8552	7900	2.1870	22.8658	42.5434	81.7954	12.4435
11.6161	2.8914	8000	2.1844	23.2422	42.986	81.6856	12.4788
11.6672	2.9275	8100	2.1809	22.9331	42.725	81.7888	12.4353
11.6776	2.9637	8200	2.1788	22.9949	42.7982	81.779	12.4471

Framework versions

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