hindi_wav2vec2

This model is a fine-tuned version of TheAIchemist13/hindi_wav2vec2 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.0003
Wer: 0.5333

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: 0.0003
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
lr_scheduler_warmup_steps: 100
num_epochs: 500

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	7.14	25	1.2980	1.4167
No log	14.29	50	0.7841	1.2833
No log	21.43	75	0.5604	1.05
No log	28.57	100	0.5288	1.05
No log	35.71	125	0.2642	0.8833
No log	42.86	150	0.3233	1.0167
No log	50.0	175	0.5526	1.0667
No log	57.14	200	0.1759	0.8167
No log	64.29	225	0.1275	0.6833
No log	71.43	250	0.1004	0.7
No log	78.57	275	0.1294	0.75
No log	85.71	300	0.1928	0.8333
No log	92.86	325	0.1206	0.7167
No log	100.0	350	0.1060	0.7
No log	107.14	375	0.0676	0.65
No log	114.29	400	0.1803	0.8667
No log	121.43	425	0.0502	0.6333
No log	128.57	450	0.0978	0.6833
No log	135.71	475	0.0817	0.6167
0.553	142.86	500	0.0695	0.6667
0.553	150.0	525	0.2449	0.8333
0.553	157.14	550	0.0407	0.6
0.553	164.29	575	0.0713	0.65
0.553	171.43	600	0.0317	0.6333
0.553	178.57	625	0.0383	0.6833
0.553	185.71	650	0.0217	0.6
0.553	192.86	675	0.0087	0.5667
0.553	200.0	700	0.0270	0.6167
0.553	207.14	725	0.1069	0.7
0.553	214.29	750	0.0118	0.5833
0.553	221.43	775	0.0089	0.6
0.553	228.57	800	0.0072	0.5667
0.553	235.71	825	0.0510	0.5833
0.553	242.86	850	0.0187	0.5833
0.553	250.0	875	0.0199	0.5833
0.553	257.14	900	0.0105	0.5833
0.553	264.29	925	0.0082	0.5833
0.553	271.43	950	0.0177	0.5833
0.553	278.57	975	0.0032	0.55
0.103	285.71	1000	0.0036	0.55
0.103	292.86	1025	0.0028	0.5333
0.103	300.0	1050	0.0040	0.5667
0.103	307.14	1075	0.0416	0.5667
0.103	314.29	1100	0.0055	0.5667
0.103	321.43	1125	0.0026	0.55
0.103	328.57	1150	0.0029	0.55
0.103	335.71	1175	0.0010	0.5333
0.103	342.86	1200	0.0036	0.55
0.103	350.0	1225	0.0013	0.55
0.103	357.14	1250	0.0010	0.5333
0.103	364.29	1275	0.0013	0.5333
0.103	371.43	1300	0.0007	0.5333
0.103	378.57	1325	0.0006	0.5333
0.103	385.71	1350	0.0005	0.5333
0.103	392.86	1375	0.0004	0.5333
0.103	400.0	1400	0.0004	0.5333
0.103	407.14	1425	0.0004	0.5333
0.103	414.29	1450	0.0003	0.5333
0.103	421.43	1475	0.0003	0.5333
0.0142	428.57	1500	0.0003	0.5333

Framework versions

Transformers 4.34.0
Pytorch 2.0.1+cu118
Datasets 1.18.3
Tokenizers 0.14.0

TheAIchemist13/hindi_wav2vec2

hindi_wav2vec2

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

为此模型做种