ai-light-dance_drums_ft_pretrain_wav2vec2-base-new-v3

This model is a fine-tuned version of gary109/ai-light-dance_drums_ft_pretrain_wav2vec2-base-new-v3 on the GARY109/AI_LIGHT_DANCE - ONSET-IDMT-SMT-DRUMS-V2+MDBDRUMS dataset. It achieves the following results on the evaluation set:

• Loss: 0.5550
• Wer: 0.3147

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 100.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1747	1.0	45	0.5638	0.3337
0.2339	2.0	90	0.5785	0.3254
0.2849	3.0	135	0.5586	0.3397
0.2396	4.0	180	0.5868	0.3266
0.2272	5.0	225	0.6052	0.3230
0.2497	6.0	270	0.5913	0.3278
0.2218	7.0	315	0.5926	0.3349
0.2584	8.0	360	0.5617	0.3218
0.2741	9.0	405	0.5901	0.3230
0.2481	10.0	450	0.5860	0.3278
0.2504	11.0	495	0.5991	0.3123
0.2125	12.0	540	0.5992	0.3218
0.2482	13.0	585	0.5756	0.3194
0.2135	14.0	630	0.5836	0.3302
0.2345	15.0	675	0.6347	0.3254
0.1912	16.0	720	0.6160	0.3206
0.2117	17.0	765	0.6268	0.3099
0.2217	18.0	810	0.6873	0.3182
0.2165	19.0	855	0.6721	0.3159
0.207	20.0	900	0.6312	0.3206
0.2263	21.0	945	0.6223	0.3290
0.2015	22.0	990	0.6319	0.3182
0.1997	23.0	1035	0.6527	0.3135
0.2318	24.0	1080	0.5987	0.3278
0.2196	25.0	1125	0.6269	0.3242
0.2298	26.0	1170	0.5774	0.3254
0.2117	27.0	1215	0.5938	0.3027
0.2553	28.0	1260	0.5831	0.3123
0.226	29.0	1305	0.6151	0.3099
0.1635	30.0	1350	0.5622	0.3230
0.5734	31.0	1395	0.6198	0.2920
0.2196	32.0	1440	0.5779	0.3039
0.2019	33.0	1485	0.5866	0.3111
0.2222	34.0	1530	0.5557	0.3063
0.2167	35.0	1575	0.5740	0.3206
0.2011	36.0	1620	0.5598	0.3004
0.2032	37.0	1665	0.5550	0.3147
0.225	38.0	1710	0.5794	0.3099
0.2068	39.0	1755	0.6223	0.3063
0.2105	40.0	1800	0.5797	0.3039
0.1968	41.0	1845	0.5681	0.2968
0.224	42.0	1890	0.5742	0.3170
0.2351	43.0	1935	0.5567	0.3111
0.2121	44.0	1980	0.5893	0.3039
0.1913	45.0	2025	0.6030	0.3027
0.1636	46.0	2070	0.5812	0.3004
0.2062	47.0	2115	0.6081	0.3004
0.2031	48.0	2160	0.5610	0.3159
0.1892	49.0	2205	0.5863	0.3147
0.1712	50.0	2250	0.5943	0.3159
0.1886	51.0	2295	0.5953	0.3051
0.1748	52.0	2340	0.5761	0.3087
0.1705	53.0	2385	0.6045	0.2872
0.1794	54.0	2430	0.5731	0.3075
0.1815	55.0	2475	0.5949	0.2849
0.1571	56.0	2520	0.5663	0.2884
0.1902	57.0	2565	0.5903	0.2956
0.2057	58.0	2610	0.5820	0.2872
0.1904	59.0	2655	0.5923	0.2896
0.1677	60.0	2700	0.5769	0.3075
0.1859	61.0	2745	0.5566	0.3147
0.2382	62.0	2790	0.5849	0.3051
0.1753	63.0	2835	0.5773	0.3075
0.1651	64.0	2880	0.5877	0.3039
0.1781	65.0	2925	0.5905	0.3027
0.1582	66.0	2970	0.5800	0.3015
0.1538	67.0	3015	0.6025	0.3075
0.1606	68.0	3060	0.5758	0.3039
0.1522	69.0	3105	0.5860	0.2932
0.1521	70.0	3150	0.5896	0.2956
0.1592	71.0	3195	0.5738	0.3027
0.2245	72.0	3240	0.5782	0.3039
0.2185	73.0	3285	0.5722	0.3027
0.1597	74.0	3330	0.5891	0.3004
0.1713	75.0	3375	0.5650	0.3027
0.1464	76.0	3420	0.5860	0.3063
0.1551	77.0	3465	0.5755	0.3027
0.1509	78.0	3510	0.5895	0.2944
0.176	79.0	3555	0.5750	0.2992
0.1695	80.0	3600	0.5759	0.3004
0.1797	81.0	3645	0.5904	0.2992
0.1371	82.0	3690	0.5923	0.3015
0.1798	83.0	3735	0.5864	0.2992
0.1386	84.0	3780	0.5733	0.3004
0.2173	85.0	3825	0.5751	0.3004
0.151	86.0	3870	0.5711	0.2968
0.1579	87.0	3915	0.5750	0.2992
0.1328	88.0	3960	0.5764	0.2944
0.1657	89.0	4005	0.5769	0.3004
0.1353	90.0	4050	0.5715	0.2956
0.1982	91.0	4095	0.5754	0.2968
0.1687	92.0	4140	0.5725	0.2980
0.1842	93.0	4185	0.5750	0.2980
0.1893	94.0	4230	0.5789	0.2944
0.1744	95.0	4275	0.5750	0.3004
0.1745	96.0	4320	0.5794	0.2980
0.1665	97.0	4365	0.5755	0.3004
0.1569	98.0	4410	0.5763	0.2968
0.1449	99.0	4455	0.5779	0.2968
0.1469	100.0	4500	0.5774	0.2968

Framework versions

• Transformers 4.25.0.dev0
• Pytorch 1.8.1+cu111
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2