base_phase4

This model is a fine-tuned version of nguyenminhly/base_phase3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2538
• Wer: 0.1236

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: 16
• 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: 1000
• num_epochs: 70

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.7336	4.49	2000	0.3028	0.1418
0.0887	8.99	4000	0.2627	0.1411
0.0679	13.48	6000	0.2520	0.1335
0.0644	17.98	8000	0.2663	0.1333
0.0555	22.47	10000	0.2783	0.1333
0.054	26.97	12000	0.2881	0.1360
0.0502	31.46	14000	0.3052	0.1340
0.0433	35.96	16000	0.2875	0.1317
0.0444	40.45	18000	0.2732	0.1319
0.0415	44.94	20000	0.2543	0.1303
0.0344	49.44	22000	0.2584	0.1275
0.0337	53.93	24000	0.2640	0.1273
0.0299	58.43	26000	0.2744	0.1271
0.0288	62.92	28000	0.2562	0.1252
0.0363	67.42	30000	0.2538	0.1236

Framework versions

• Transformers 4.32.1
• Pytorch 2.0.1+cu117
• Datasets 2.14.4
• Tokenizers 0.13.3