wav2vec2-large-xlsr-coraa-texts-aug-exp-1

This model is a fine-tuned version of Edresson/wav2vec2-large-xlsr-coraa-portuguese on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3162
• Wer: 0.2298
• Cer: 0.1328
• Per: 0.2260

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: 3e-05
• train_batch_size: 16
• eval_batch_size: 8
• 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
• num_epochs: 200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Per
43.995	1.0	84	3.3727	1.0	0.9671	1.0
9.1824	2.0	168	2.9921	1.0	0.9671	1.0
3.1225	3.0	252	2.9321	1.0	0.9671	1.0
2.979	4.0	336	2.9394	1.0	0.9671	1.0
2.9357	5.0	420	2.9118	1.0	0.9671	1.0
2.9116	6.0	504	2.8401	1.0	0.9671	1.0
2.9116	7.0	588	2.0977	1.0	0.7348	1.0
2.6841	8.0	672	0.8210	0.5521	0.2093	0.5335
1.6358	9.0	756	0.5586	0.3329	0.1633	0.3080
1.1226	10.0	840	0.4817	0.2951	0.1537	0.2834
0.9403	11.0	924	0.4337	0.2755	0.1482	0.2643
0.8262	12.0	1008	0.4019	0.2717	0.1460	0.2613
0.8262	13.0	1092	0.3872	0.2682	0.1458	0.2590
0.7654	14.0	1176	0.3664	0.2532	0.1414	0.2445
0.6788	15.0	1260	0.3476	0.2501	0.1390	0.2405
0.647	16.0	1344	0.3508	0.2453	0.1388	0.2364
0.6058	17.0	1428	0.3425	0.2448	0.1392	0.2349
0.5853	18.0	1512	0.3393	0.2377	0.1351	0.2293
0.5853	19.0	1596	0.3276	0.2293	0.1347	0.2222
0.5356	20.0	1680	0.3275	0.2316	0.1340	0.2275
0.5323	21.0	1764	0.3245	0.2301	0.1342	0.2265
0.501	22.0	1848	0.3258	0.2232	0.1334	0.2197
0.4956	23.0	1932	0.3262	0.2242	0.1332	0.2202
0.4427	24.0	2016	0.3162	0.2298	0.1328	0.2260
0.4443	25.0	2100	0.3225	0.2285	0.1331	0.2252
0.4443	26.0	2184	0.3246	0.2270	0.1333	0.2227
0.4234	27.0	2268	0.3171	0.2250	0.1327	0.2214
0.4245	28.0	2352	0.3181	0.2265	0.1344	0.2230
0.3993	29.0	2436	0.3283	0.2250	0.1331	0.2217
0.3769	30.0	2520	0.3292	0.2354	0.1359	0.2324
0.3925	31.0	2604	0.3276	0.2344	0.1360	0.2313
0.3925	32.0	2688	0.3356	0.2301	0.1345	0.2275
0.3724	33.0	2772	0.3363	0.2273	0.1357	0.2245
0.3614	34.0	2856	0.3307	0.2311	0.1358	0.2280
0.3705	35.0	2940	0.3276	0.2316	0.1356	0.2285
0.3376	36.0	3024	0.3309	0.2379	0.1373	0.2336
0.3434	37.0	3108	0.3340	0.2291	0.1367	0.2255
0.3434	38.0	3192	0.3428	0.2326	0.1369	0.2293
0.3245	39.0	3276	0.3410	0.2260	0.1340	0.2227
0.3087	40.0	3360	0.3516	0.2285	0.1348	0.2257
0.3025	41.0	3444	0.3396	0.2268	0.1347	0.2237
0.2953	42.0	3528	0.3443	0.2255	0.1346	0.2222
0.2778	43.0	3612	0.3406	0.2255	0.1351	0.2230
0.2778	44.0	3696	0.3457	0.2346	0.1367	0.2308

Framework versions

• Transformers 4.28.0
• Pytorch 2.4.1+cu121
• Datasets 3.2.0
• Tokenizers 0.13.3