wav2vec2-large-xls-r-300m-tr

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - TR dataset. It achieves the following results on the evaluation set:

• Loss: 0.2224
• Wer: 0.2869

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: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 100.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.8222	0.64	500	3.5026	1.0
3.2136	1.28	1000	3.0593	1.0000
2.8882	1.91	1500	2.4670	0.9939
2.3743	2.55	2000	1.1844	0.8657
1.9456	3.19	2500	0.8228	0.7397
1.7781	3.83	3000	0.6826	0.6753
1.6848	4.46	3500	0.5885	0.6140
1.6228	5.1	4000	0.5274	0.5789
1.5768	5.74	4500	0.4900	0.5519
1.5431	6.38	5000	0.4508	0.5238
1.5019	7.02	5500	0.4248	0.5021
1.4684	7.65	6000	0.4009	0.4827
1.4635	8.29	6500	0.3830	0.4700
1.4291	8.93	7000	0.3707	0.4595
1.4271	9.57	7500	0.3570	0.4514
1.3938	10.2	8000	0.3479	0.4378
1.3914	10.84	8500	0.3396	0.4368
1.3767	11.48	9000	0.3253	0.4262
1.3641	12.12	9500	0.3251	0.4178
1.355	12.76	10000	0.3138	0.4136
1.336	13.39	10500	0.3121	0.4069
1.3292	14.03	11000	0.3041	0.4014
1.3249	14.67	11500	0.3014	0.3931
1.3156	15.31	12000	0.3014	0.3929
1.313	15.94	12500	0.2969	0.3968
1.3068	16.58	13000	0.2965	0.3966
1.2785	17.22	13500	0.2943	0.3850
1.2867	17.86	14000	0.2912	0.3782
1.2714	18.49	14500	0.2819	0.3747
1.2844	19.13	15000	0.2840	0.3740
1.2684	19.77	15500	0.2913	0.3828
1.26	20.41	16000	0.2739	0.3674
1.2543	21.05	16500	0.2740	0.3691
1.2532	21.68	17000	0.2709	0.3756
1.2409	22.32	17500	0.2669	0.3593
1.2404	22.96	18000	0.2673	0.3576
1.2347	23.6	18500	0.2678	0.3643
1.2351	24.23	19000	0.2715	0.3650
1.2409	24.87	19500	0.2637	0.3571
1.2152	25.51	20000	0.2785	0.3609
1.2046	26.15	20500	0.2610	0.3508
1.2082	26.79	21000	0.2619	0.3461
1.2109	27.42	21500	0.2597	0.3502
1.2014	28.06	22000	0.2608	0.3468
1.1948	28.7	22500	0.2573	0.3457
1.205	29.34	23000	0.2619	0.3464
1.2019	29.97	23500	0.2559	0.3474
1.1917	30.61	24000	0.2601	0.3462
1.1939	31.25	24500	0.2575	0.3387
1.1882	31.89	25000	0.2535	0.3368
1.191	32.53	25500	0.2489	0.3365
1.1767	33.16	26000	0.2501	0.3347
1.167	33.8	26500	0.2504	0.3347
1.1678	34.44	27000	0.2480	0.3378
1.1803	35.08	27500	0.2487	0.3345
1.167	35.71	28000	0.2442	0.3319
1.1661	36.35	28500	0.2495	0.3334
1.164	36.99	29000	0.2472	0.3292
1.1578	37.63	29500	0.2442	0.3242
1.1584	38.27	30000	0.2431	0.3314
1.1526	38.9	30500	0.2441	0.3347
1.1542	39.54	31000	0.2437	0.3330
1.1508	40.18	31500	0.2433	0.3294
1.1406	40.82	32000	0.2434	0.3271
1.1514	41.45	32500	0.2426	0.3255
1.1418	42.09	33000	0.2432	0.3233
1.1365	42.73	33500	0.2436	0.3240
1.1348	43.37	34000	0.2483	0.3257
1.1301	44.01	34500	0.2420	0.3271
1.1268	44.64	35000	0.2472	0.3225
1.1224	45.28	35500	0.2382	0.3205
1.1224	45.92	36000	0.2388	0.3184
1.1198	46.56	36500	0.2382	0.3202
1.1274	47.19	37000	0.2404	0.3172
1.1147	47.83	37500	0.2394	0.3164
1.121	48.47	38000	0.2406	0.3202
1.1109	49.11	38500	0.2384	0.3154
1.1164	49.74	39000	0.2375	0.3169
1.1105	50.38	39500	0.2387	0.3173
1.1054	51.02	40000	0.2362	0.3120
1.0893	51.66	40500	0.2399	0.3130
1.0913	52.3	41000	0.2357	0.3088
1.1017	52.93	41500	0.2345	0.3084
1.0937	53.57	42000	0.2330	0.3140
1.0945	54.21	42500	0.2399	0.3107
1.0933	54.85	43000	0.2383	0.3134
1.0912	55.48	43500	0.2372	0.3077
1.0898	56.12	44000	0.2339	0.3083
1.0903	56.76	44500	0.2367	0.3065
1.0947	57.4	45000	0.2352	0.3104
1.0751	58.04	45500	0.2334	0.3084
1.09	58.67	46000	0.2328	0.3100
1.0876	59.31	46500	0.2276	0.3050
1.076	59.95	47000	0.2309	0.3047
1.086	60.59	47500	0.2293	0.3047
1.082	61.22	48000	0.2328	0.3027
1.0714	61.86	48500	0.2290	0.3020
1.0746	62.5	49000	0.2313	0.3059
1.076	63.14	49500	0.2342	0.3050
1.0648	63.78	50000	0.2286	0.3025
1.0586	64.41	50500	0.2338	0.3044
1.0753	65.05	51000	0.2308	0.3045
1.0664	65.69	51500	0.2273	0.3009
1.0739	66.33	52000	0.2298	0.3027
1.0695	66.96	52500	0.2247	0.2996
1.06	67.6	53000	0.2276	0.3015
1.0742	68.24	53500	0.2280	0.2974
1.0618	68.88	54000	0.2291	0.2989
1.062	69.52	54500	0.2302	0.2971
1.0572	70.15	55000	0.2280	0.2990
1.055	70.79	55500	0.2278	0.2983
1.0553	71.43	56000	0.2282	0.2991
1.0509	72.07	56500	0.2261	0.2959
1.0469	72.7	57000	0.2216	0.2919
1.0476	73.34	57500	0.2267	0.2989
1.0494	73.98	58000	0.2260	0.2960
1.0517	74.62	58500	0.2297	0.2989
1.0458	75.26	59000	0.2246	0.2923
1.0382	75.89	59500	0.2255	0.2922
1.0462	76.53	60000	0.2258	0.2954
1.0375	77.17	60500	0.2251	0.2929
1.0332	77.81	61000	0.2277	0.2940
1.0423	78.44	61500	0.2243	0.2896
1.0379	79.08	62000	0.2274	0.2928
1.0398	79.72	62500	0.2237	0.2928
1.0395	80.36	63000	0.2265	0.2956
1.0397	80.99	63500	0.2240	0.2920
1.0262	81.63	64000	0.2244	0.2934
1.0335	82.27	64500	0.2265	0.2936
1.0385	82.91	65000	0.2238	0.2928
1.0289	83.55	65500	0.2219	0.2912
1.0372	84.18	66000	0.2236	0.2898
1.0279	84.82	66500	0.2219	0.2902
1.0325	85.46	67000	0.2240	0.2908
1.0202	86.1	67500	0.2206	0.2886
1.0166	86.73	68000	0.2219	0.2886
1.0259	87.37	68500	0.2235	0.2897
1.0337	88.01	69000	0.2210	0.2873
1.0264	88.65	69500	0.2216	0.2882
1.0231	89.29	70000	0.2223	0.2899
1.0281	89.92	70500	0.2214	0.2872
1.0135	90.56	71000	0.2218	0.2868
1.0291	91.2	71500	0.2209	0.2863
1.0321	91.84	72000	0.2199	0.2876
1.028	92.47	72500	0.2214	0.2858
1.0213	93.11	73000	0.2219	0.2875
1.0261	93.75	73500	0.2232	0.2869
1.0197	94.39	74000	0.2227	0.2866
1.0298	95.03	74500	0.2228	0.2868
1.0192	95.66	75000	0.2230	0.2865
1.0156	96.3	75500	0.2220	0.2869
1.0075	96.94	76000	0.2223	0.2866
1.0201	97.58	76500	0.2219	0.2866
1.0159	98.21	77000	0.2219	0.2876
1.0087	98.85	77500	0.2219	0.2873
1.0159	99.49	78000	0.2223	0.2867

Framework versions

• Transformers 4.17.0.dev0
• Pytorch 1.10.2+cu102
• Datasets 1.18.2.dev0
• Tokenizers 0.11.0