20220412 - 203254 Open-source Speech Recognition Model - Free Support for Automatic Speech Recognition Tasks

20220412 203254

Developed by lilitket

This model is a fine-tuned speech recognition model based on facebook/wav2vec2-xls-r-300m on the common_voice dataset, supporting automatic speech recognition tasks.

Speech Recognition

Transformers

Open Source License:Apache-2.0 #Speech Recognition #Low Word Error Rate #Multilingual Support

Downloads 18

Release Time : 4/12/2022

Model Overview

This is a speech recognition model based on the wav2vec2-xls-r-300m architecture, fine-tuned on the common_voice dataset, capable of converting speech to text.

Model Features

Efficient Fine-tuning

Fine-tuned based on the pre-trained wav2vec2-xls-r-300m model, leveraging the advantages of large-scale pre-training.

Low Word Error Rate

Achieved a word error rate (WER) of 1.0019 on the evaluation set, demonstrating excellent performance.

Mixed Precision Training

Utilizes native AMP mixed precision training technology to improve training efficiency.

Model Capabilities

Speech to Text

Automatic Speech Recognition

Use Cases

Speech Transcription

Automatic Meeting Transcription

Automatically converts meeting recordings into text transcripts

Word error rate as low as 1.0019

Voice Assistant

Used in the speech recognition module of voice assistant systems

🚀 20220412-203254

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

Loss: 5.0428
Wer: 1.0019

✨ Features

This model is fine - tuned from facebook/wav2vec2-xls-r-300m on the common_voice dataset, which may have better performance in speech - related tasks.

🔧 Technical Details

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 6e-06
train_batch_size: 1
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 2000
num_epochs: 1200
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
16.9455	1.5	200	16.4676	1.2534
15.444	3.01	400	14.1207	1.0
9.5452	4.51	600	8.4030	1.0
6.2565	6.02	800	6.5233	1.0
5.2827	7.52	1000	5.6058	1.0
4.7652	9.02	1200	4.9765	1.0
4.3803	10.53	1400	4.4565	1.0
4.0005	12.03	1600	4.0224	1.0
3.7041	13.53	1800	3.6903	1.0
3.4991	15.04	2000	3.4642	1.0
3.34	16.54	2200	3.3425	1.0
3.2352	18.05	2400	3.2617	1.0
3.1867	19.55	2600	3.2358	1.0
3.161	21.05	2800	3.2289	1.0
3.145	22.56	3000	3.2023	1.0
3.1203	24.06	3200	3.1964	1.0
3.1109	25.56	3400	3.1844	1.0
3.0958	27.07	3600	3.1839	1.0
3.0732	28.57	3800	3.2058	1.0
3.0535	30.08	4000	3.1843	1.0
3.0243	31.58	4200	3.1992	1.0
2.9829	33.08	4400	3.2019	1.0
2.9219	34.59	4600	3.2346	1.0
2.8313	36.09	4800	3.2781	1.0
2.7186	37.59	5000	3.3056	1.0
2.5745	39.1	5200	3.3554	1.0
2.4028	40.6	5400	3.4331	1.0
2.2645	42.11	5600	3.4418	1.0
2.1303	43.61	5800	3.5584	1.0
2.0257	45.11	6000	3.5943	1.0
1.9223	46.62	6200	3.6767	1.0
1.8344	48.12	6400	3.7363	1.0
1.7574	49.62	6600	3.8921	1.0
1.67	51.13	6800	3.9054	1.0
1.6118	52.63	7000	4.0352	1.0
1.5372	54.14	7200	3.9742	1.0
1.4846	55.64	7400	4.1078	1.0
1.4093	57.14	7600	4.1705	1.0
1.3379	58.65	7800	4.2737	1.0
1.28	60.15	8000	4.3662	1.0
1.2268	61.65	8200	4.4278	1.0
1.1641	63.16	8400	4.4831	1.0
1.1058	64.66	8600	4.5354	1.0
1.0596	66.17	8800	4.5983	1.0
0.9953	67.67	9000	4.7143	1.0
0.9406	69.17	9200	4.8536	1.0
0.9022	70.68	9400	4.7732	1.0
0.8551	72.18	9600	4.8929	1.0
0.8103	73.68	9800	4.9513	1.0
0.7665	75.19	10000	4.9530	1.0
0.7215	76.69	10200	5.1471	1.0
0.6906	78.2	10400	5.2295	1.0
0.6354	79.7	10600	5.1287	1.0
0.6196	81.2	10800	5.2081	1.0
0.6026	82.71	11000	5.4323	1.0
0.5726	84.21	11200	5.3907	1.0
0.5348	85.71	11400	5.5669	1.0
0.5344	87.22	11600	5.5685	1.0
0.4849	88.72	11800	5.5814	1.0
0.4689	90.23	12000	5.6186	1.0
0.4646	91.73	12200	5.4834	1.0
0.4266	93.23	12400	5.6463	1.0
0.4424	94.74	12600	5.6562	1.0
0.3865	96.24	12800	5.7463	1.0
0.3914	97.74	13000	5.7014	1.0
0.3661	99.25	13200	5.7543	1.0
0.3582	100.75	13400	5.9172	1.0
0.3571	102.26	13600	5.5968	1.0
0.3343	103.76	13800	5.3691	1.0
0.3123	105.26	14000	5.8917	1.0
0.3089	106.77	14200	5.8054	1.0
0.3078	108.27	14400	5.9066	1.0
0.3076	109.77	14600	5.7379	1.0
0.2924	111.28	14800	5.7931	1.0
0.2925	112.78	15000	5.9529	1.0
0.2839	114.29	15200	5.9881	1.0
0.2599	115.79	15400	6.0081	1.0
0.2685	117.29	15600	6.1049	1.0
0.2557	118.8	15800	6.1154	1.0
0.2688	120.3	16000	5.9336	1.0
0.2422	121.8	16200	6.0492	1.0
0.2408	123.31	16400	6.3155	1.0
0.2423	124.81	16600	6.3437	1.0
0.2421	126.32	16800	6.0979	1.0
0.2212	127.82	17000	5.5551	1.0
0.2239	129.32	17200	5.9007	1.0
0.2101	130.83	17400	6.0142	1.0
0.2097	132.33	17600	5.8984	1.0
0.2064	133.83	17800	5.9705	1.0
0.1898	135.34	18000	5.9915	1.0
0.2053	136.84	18200	6.1079	1.0
0.1822	138.35	18400	6.1324	1.0
0.1867	139.85	18600	6.1122	1.0
0.1831	141.35	18800	6.1476	1.0
0.1935	142.86	19000	5.7248	1.0
0.1983	144.36	19200	6.1466	1.0
0.176	145.86	19400	5.9555	1.0
0.1778	147.37	19600	6.1434	1.0
0.1758	148.87	19800	6.2104	1.0
0.1799	150.38	20000	6.0933	1.0
0.1674	151.88	20200	6.0476	1.0
0.1777	153.38	20400	5.8937	1.0
0.1616	154.89	20600	6.4417	1.0
0.1498	156.39	20800	6.3136	1.0
0.1607	157.89	21000	5.9295	1.0
0.1445	159.4	21200	6.2741	1.0
0.1636	160.9	21400	6.1931	1.0
0.1488	162.41	21600	6.0089	1.0
0.1549	163.91	21800	5.6184	1.0
0.1532	165.41	22000	6.1250	1.0
0.1581	166.92	22200	6.2635	1.0
0.146	168.42	22400	6.0498	1.0
0.148	169.92	22600	6.3486	1.0
0.1489	171.43	22800	6.1659	1.0
0.1464	172.93	23000	6.2259	1.0
0.139	174.44	23200	6.2796	1.0
0.1357	175.94	23400	6.2119	1.0
0.1435	177.44	23600	6.5722	1.0
0.1172	178.95	23800	6.4221	1.0
0.1539	180.45	24000	6.3963	1.0
0.1389	181.95	24200	6.2367	1.0
0.1274	183.46	24400	6.3693	1.0
0.1295	184.96	24600	6.0819	1.0
0.1337	186.47	24800	6.1525	1.0
0.1303	187.97	25000	6.2520	1.0
0.141	189.47	25200	6.5302	1.0
0.1322	190.98	25400	6.3731	1.0
0.1313	192.48	25600	6.3570	1.0
0.1178	193.98	25800	6.1667	1.0
0.1277	195.49	26000	6.1352	1.0
0.1169	196.99	26200	6.3132	1.0
0.1199	198.5	26400	6.6116	1.0
0.1138	200.0	26600	6.4862	1.0
0.1129	201.5	26800	6.3442	1.0
0.1142	203.01	27000	6.5077	1.0
0.1169	204.51	27200	6.5710	1.0
0.111	206.02	27400	6.0623	1.0
0.1198	207.52	27600	6.4331	1.0
0.1108	209.02	27800	5.9192	1.0
0.1121	210.53	28000	6.0724	1.0
0.1171	212.03	28200	6.3363	1.0
0.1188	213.53	28400	6.3704	1.0
0.104	215.04	28600	6.5802	1.0
0.1125	216.54	28800	5.4428	1.0
0.1115	218.05	29000	6.4286	1.0
0.1109	219.55	29200	6.6998	1.0
0.1061	221.05	29400	6.3761	1.0
0.1161	222.56	29600	5.8712	1.0
0.1091	224.06	29800	6.1844	1.0
0.0947	225.56	30000	6.5670	1.0
0.1004	227.07	30200	6.2302	1.0
0.1099	228.57	30400	6.4218	1.0
0.1154	230.08	30600	6.4911	1.0
0.0999	231.58	30800	6.4390	1.0
0.1068	233.08	31000	6.2367	1.0
0.1015	234.59	31200	6.0790	1.0
0.0958	236.09	31400	5.9926	1.0
0.1183	237.59	31600	6.3400	1.0
0.0833	239.1	31800	6.4481	1.0
0.0874	240.6	32000	6.4535	1.0
0.0958	242.11	32200	6.0597	1.0
0.1075	243.61	32400	6.3403	1.0
0.0909	245.11	32600	6.1297	1.0
0.1093	246.62	32800	6.2232	1.0
0.0995	248.12	33000	6.7110	1.0
0.1061	249.62	33200	5.8551	1.0
0.0872	251.13	33400	6.0338	1.0
0.109	252.63	33600	6.2880	1.0
0.0976	254.14	33800	5.9304	1.0
0.0977	255.64	34000	6.4527	1.0
0.0895	257.14	34200	6.3178	1.0
0.0951	258.65	34400	6.3646	1.0
0.0942	260.15	34600	6.4405	1.0
0.0876	261.65	34800	5.8373	1.0
0.0877	263.16	35000	6.5296	1.0
0.0896	264.66	35200	6.6644	1.0
0.0938	266.17	35400	6.4721	1.0
0.0864	267.67	35600	7.0132	1.0
0.0897	269.17	35800	6.3767	1.0
0.094	270.68	36000	6.1663	1.0
0.0782	272.18	36200	5.7325	1.0
0.0878	273.68	36400	6.0681	1.0
0.0877	275.19	36600	6.2621	1.0
0.0827	276.69	36800	6.0367	1.0

📄 License

This model is licensed under the Apache 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご