Terjman-Large-v2.2

This model is a fine-tuned version of atlasia/Terjman-Large-v1.2 on BounharAbdelaziz/Terjman-v2-English-Darija-Dataset-350K. It achieves the following results on atlasia/TerjamaBench set:

• Loss: 2.8068
• Bleu: 20.1692
• Chrf: 41.5829
• Ter: 82.2755
• Gen Len: 29.0212

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 8
• 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: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Bleu	Chrf	Ter	Gen Len
25.1947	0.0361	100	4.3934	15.9485	35.999	89.4884	10.3094
23.9289	0.0723	200	4.3239	16.0039	36.1914	89.2378	10.3871
22.4636	0.1084	300	4.1613	16.4537	36.6901	88.6224	10.3341
20.7804	0.1446	400	3.8688	16.6385	36.9491	88.2061	10.2976
19.3631	0.1807	500	3.6065	16.8079	37.4258	87.9841	10.24
18.5934	0.2169	600	3.4287	17.1599	37.9096	87.8884	10.2471
17.5375	0.2530	700	3.3204	17.3512	38.3927	87.8847	10.5671
17.2196	0.2892	800	3.2327	17.8504	38.8577	88.2548	11.4647
16.648	0.3253	900	3.1727	18.0511	39.2475	87.3063	11.8929
15.9364	0.3615	1000	3.1162	18.5393	39.5155	86.2896	11.5035
15.4445	0.3976	1100	3.0761	18.8806	40.0084	86.0285	10.6118
15.1001	0.4338	1200	3.0376	19.0202	40.0596	85.4085	10.9141
14.4878	0.4699	1300	3.0135	19.1544	40.1196	85.0821	11.0424
14.0768	0.5061	1400	2.9858	18.8328	40.13	85.0403	11.6153
13.9057	0.5422	1500	2.9611	19.0325	40.3066	84.6503	12.9035
13.582	0.5783	1600	2.9354	19.2564	40.6244	84.4523	14.2729
13.2156	0.6145	1700	2.9283	19.2737	40.6857	84.1659	13.6271
13.216	0.6506	1800	2.9195	19.1842	40.5645	84.1892	15.3176
12.8209	0.6868	1900	2.9140	18.9829	40.2408	84.6937	15.2318
12.6086	0.7229	2000	2.9068	19.1247	40.4541	84.4727	15.5129
12.4518	0.7591	2100	2.8839	19.4021	40.8301	84.419	14.9365
12.4289	0.7952	2200	2.8866	19.7503	41.0128	83.2866	16.9776
12.2287	0.8314	2300	2.8734	19.8382	40.9191	83.0623	17.1
12.23	0.8675	2400	2.8700	19.6373	40.807	83.2168	19.0612
12.0073	0.9037	2500	2.8616	19.6527	40.7406	83.1873	18.9859
11.9013	0.9398	2600	2.8486	20.3566	41.2676	82.5944	20.0953
11.724	0.9760	2700	2.8548	19.9016	40.9086	82.4162	20.6
11.5015	1.0119	2800	2.8643	19.7775	41.0928	82.6536	19.4812
11.4529	1.0481	2900	2.8572	19.9022	41.1604	82.5963	20.5929
11.3655	1.0842	3000	2.8501	20.2911	41.3615	82.0111	20.8847
11.44	1.1204	3100	2.8478	20.2161	41.2086	82.363	24.5824
11.3663	1.1565	3200	2.8399	20.2121	41.4432	81.9885	23.4976
11.2401	1.1927	3300	2.8444	20.3898	41.3422	81.8858	25.3129
11.2055	1.2288	3400	2.8404	20.463	41.5269	81.9694	25.6447
11.0982	1.2650	3500	2.8417	20.5416	41.8609	81.852	24.7647
11.0639	1.3011	3600	2.8364	20.3427	41.7149	81.642	25.38
11.1082	1.3372	3700	2.8382	20.1427	41.525	82.3573	27.2624
11.0479	1.3734	3800	2.8349	20.092	41.4815	82.2276	26.3412
11.1308	1.4095	3900	2.8270	19.8146	41.2064	82.8144	26.9353
10.9374	1.4457	4000	2.8278	20.334	41.6593	81.7711	27.2165
11.0379	1.4818	4100	2.8208	20.3952	41.6785	81.7776	27.8235
11.0583	1.5180	4200	2.8225	20.3378	41.6073	82.1717	27.2506
10.7442	1.5541	4300	2.8193	20.326	41.5546	82.0312	27.2482
10.9269	1.5903	4400	2.8220	20.3777	41.6625	81.8084	27.2341
10.7753	1.6264	4500	2.8237	20.3371	41.63	81.788	26.0553
10.7841	1.6626	4600	2.8224	20.2222	41.4059	82.0374	26.6306
10.8673	1.6987	4700	2.8190	20.0474	41.3502	81.9854	27.2165
10.7534	1.7349	4800	2.8230	20.1404	41.3887	82.1984	27.2447
10.7428	1.7710	4900	2.8164	20.1693	41.4398	82.2848	26.64
10.7501	1.8072	5000	2.8152	20.2346	41.3471	81.9833	26.6235
10.5935	1.8433	5100	2.8179	20.0303	41.3017	82.2892	26.6341
10.8916	1.8795	5200	2.8142	20.1371	41.2921	82.3067	26.6388
10.4615	1.9156	5300	2.8147	20.1184	41.4823	82.4293	26.0271
10.6519	1.9517	5400	2.8145	19.8947	41.2411	82.4574	27.1906
10.7264	1.9879	5500	2.8182	19.9463	41.3386	82.2442	27.2424
10.3405	2.0239	5600	2.8159	19.9448	41.2743	82.3859	27.8
10.3744	2.0600	5700	2.8108	20.1041	41.5535	82.2007	26.04
10.4719	2.0962	5800	2.8152	19.9187	41.3583	82.2967	27.2235
10.5306	2.1323	5900	2.8127	19.944	41.3826	82.3566	27.8153
10.4882	2.1684	6000	2.8126	20.0051	41.4424	82.3244	27.2376
10.6208	2.2046	6100	2.8109	20.2372	41.6177	81.9337	26.6412
10.458	2.2407	6200	2.8103	20.259	41.6903	82.0083	26.6188
10.5497	2.2769	6300	2.8176	20.1553	41.6126	82.3499	26.6318
10.4733	2.3130	6400	2.8158	20.2925	41.6992	82.0426	26.6329
10.3814	2.3492	6500	2.8161	20.3436	41.6806	82.0206	26.6318
10.5096	2.3853	6600	2.8104	20.0678	41.5413	82.3613	26.6353
10.345	2.4215	6700	2.8162	20.0445	41.5661	82.3574	26.6424
10.3022	2.4576	6800	2.8150	20.1996	41.6086	82.1407	27.2247
10.511	2.4938	6900	2.8124	20.2234	41.6513	82.0956	26.0294
10.5411	2.5299	7000	2.8110	20.0885	41.6026	82.3022	27.1976
10.3113	2.5661	7100	2.8125	20.2328	41.6496	82.2802	27.8376
10.2469	2.6022	7200	2.8140	20.1713	41.6658	82.3415	26.6341
10.6242	2.6384	7300	2.8082	20.397	41.7355	81.8477	27.8071
10.3379	2.6745	7400	2.8141	20.2633	41.6424	82.2042	27.2282
10.2108	2.7106	7500	2.8112	20.1278	41.5304	82.2911	27.2106
10.3156	2.7468	7600	2.8113	20.2513	41.4369	82.1582	27.6129
10.5468	2.7829	7700	2.8092	20.2669	41.6353	82.0547	26.6212
10.3466	2.8191	7800	2.8095	20.1406	41.5998	82.2192	27.2294
10.1576	2.8552	7900	2.8102	20.4323	41.717	82.0407	27.2282
10.2929	2.8914	8000	2.8081	20.2843	41.6445	82.1534	27.2294
10.1551	2.9275	8100	2.8124	20.042	41.5029	82.3521	27.6212
10.3701	2.9637	8200	2.8074	20.149	41.6323	82.1568	27.7918
10.3841	2.9998	8300	2.8071	20.2359	41.7569	82.0191	27.2
10.2077	3.0358	8400	2.8088	20.1197	41.6256	82.2721	27.2141
10.2587	3.0719	8500	2.8089	20.1437	41.6294	82.2269	27.2082
10.4181	3.1081	8600	2.8087	20.3797	41.6715	81.84	27.2141
10.2875	3.1442	8700	2.8113	20.3151	41.7896	82.1183	27.2376
10.3484	3.1804	8800	2.8118	20.2499	41.5947	82.1579	29.0153
10.2257	3.2165	8900	2.8107	20.119	41.5849	82.1647	27.8012
10.2536	3.2527	9000	2.8098	20.208	41.5198	82.1859	28.4047
10.226	3.2888	9100	2.8083	20.0872	41.5215	82.2154	27.7941
10.2555	3.3250	9200	2.8066	20.0697	41.5266	82.2524	28.3953
10.1948	3.3611	9300	2.8077	19.9162	41.4577	82.5675	28.4
10.346	3.3973	9400	2.8087	20.0162	41.3612	82.5519	28.9906
10.2357	3.4334	9500	2.8088	20.1612	41.6107	82.364	27.8271
10.4022	3.4695	9600	2.8049	20.0821	41.5959	82.3291	26.6165
10.2421	3.5057	9700	2.8070	20.0572	41.4946	82.3907	26.6259
10.1731	3.5418	9800	2.8069	20.0779	41.4285	82.3776	27.7812
10.275	3.5780	9900	2.8061	20.1649	41.5337	82.2389	27.2106
10.1281	3.6141	10000	2.8081	20.2355	41.5411	82.2055	27.2212
10.178	3.6503	10100	2.8069	20.1021	41.4715	82.5423	27.2153
10.1036	3.6864	10200	2.8087	20.0886	41.532	82.4482	26.6365
10.2586	3.7226	10300	2.8066	20.2039	41.4906	82.184	27.7965
10.1633	3.7587	10400	2.8072	20.2001	41.6133	82.4142	27.2271
10.3653	3.7949	10500	2.8051	20.2178	41.7168	82.1759	26.6306
10.3772	3.8310	10600	2.8061	20.1352	41.4552	82.2656	28.9882
10.265	3.8672	10700	2.8063	20.3686	41.6936	82.0663	27.2247
10.1565	3.9033	10800	2.8066	20.3704	41.5896	82.1135	28.4235
10.1053	3.9395	10900	2.8064	20.232	41.638	82.1641	27.8224
10.2919	3.9756	11000	2.8058	20.2259	41.4303	82.3619	28.4035
10.2109	4.0116	11100	2.8063	20.1277	41.5148	82.4291	29.0071
10.2317	4.0477	11200	2.8065	20.1976	41.5571	82.3222	27.8082
10.2665	4.0839	11300	2.8045	20.1174	41.4949	82.3993	28.4118
10.2305	4.1200	11400	2.8054	20.1751	41.6274	82.3186	27.8212
10.1426	4.1562	11500	2.8066	20.2274	41.4655	82.3561	29.0047
10.2428	4.1923	11600	2.8060	20.1688	41.5318	82.415	27.8059
10.3369	4.2284	11700	2.8055	20.2572	41.6457	82.1688	27.8153
10.3457	4.2646	11800	2.8057	20.3668	41.6965	82.0632	27.2224
10.3538	4.3007	11900	2.8064	20.1635	41.5695	82.238	28.4165
10.1697	4.3369	12000	2.8059	20.3072	41.5665	82.1484	27.8094
10.2341	4.3730	12100	2.8064	20.1402	41.4892	82.2549	28.4106
10.1779	4.4092	12200	2.8064	20.2026	41.6443	82.2105	27.8412
10.287	4.4453	12300	2.8069	20.1142	41.4856	82.2969	28.4094
10.2197	4.4815	12400	2.8064	20.2797	41.5558	82.198	27.8294
10.3417	4.5176	12500	2.8065	20.2762	41.5022	82.129	28.4059
10.2043	4.5538	12600	2.8064	20.3172	41.5538	82.0916	27.8141
10.316	4.5899	12700	2.8067	20.1842	41.564	82.247	28.3953
10.1878	4.6261	12800	2.8061	20.1348	41.5324	82.3159	28.9941
10.3951	4.6622	12900	2.8063	20.1038	41.4362	82.3589	29.0212
10.2292	4.6984	13000	2.8061	19.9689	41.3646	82.5469	29.5953
10.1952	4.7345	13100	2.8068	20.2237	41.4602	82.2747	28.4012
10.2828	4.7706	13200	2.8066	20.2114	41.5359	82.2337	27.8165
10.3883	4.8068	13300	2.8063	20.1173	41.4379	82.2924	28.4024
10.2585	4.8429	13400	2.8065	20.1329	41.5104	82.3775	28.4224
10.1554	4.8791	13500	2.8064	20.2585	41.5165	82.1683	27.8224
10.2092	4.9152	13600	2.8063	20.0996	41.4114	82.3806	29.0118
10.2896	4.9514	13700	2.8067	20.0253	41.3706	82.4652	27.8082
10.2847	4.9875	13800	2.8068	20.1692	41.5829	82.2755	29.0212

Framework versions

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