MSP-VSR
This model is a fine-tuned version of on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.2109
- Wer: 0.6198
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: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 1000.0
- training_steps: 50000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 2.6988 | 0.02 | 1000 | 2.5501 | 0.9915 |
| 2.0297 | 0.04 | 2000 | 1.7684 | 0.8450 |
| 1.9479 | 0.06 | 3000 | 1.6473 | 0.8139 |
| 1.9878 | 0.08 | 4000 | 1.5835 | 0.7850 |
| 2.0281 | 0.1 | 5000 | 1.5461 | 0.7615 |
| 1.9042 | 0.12 | 6000 | 1.4989 | 0.7539 |
| 1.9134 | 0.14 | 7000 | 1.4811 | 0.7418 |
| 1.9830 | 0.16 | 8000 | 1.4620 | 0.7214 |
| 1.9236 | 0.18 | 9000 | 1.4404 | 0.7128 |
| 1.8155 | 0.2 | 10000 | 1.4440 | 0.7099 |
| 1.8339 | 0.22 | 11000 | 1.4199 | 0.7160 |
| 1.7363 | 0.24 | 12000 | 1.4074 | 0.7001 |
| 1.9054 | 0.26 | 13000 | 1.3908 | 0.6984 |
| 1.8643 | 0.28 | 14000 | 1.3891 | 0.6907 |
| 1.6696 | 0.3 | 15000 | 1.3644 | 0.6904 |
| 1.6995 | 0.32 | 16000 | 1.3502 | 0.6964 |
| 1.7230 | 0.34 | 17000 | 1.3431 | 0.6806 |
| 1.7547 | 0.36 | 18000 | 1.3435 | 0.6762 |
| 1.6977 | 0.38 | 19000 | 1.3134 | 0.6728 |
| 1.6884 | 0.4 | 20000 | 1.3223 | 0.6699 |
| 1.7758 | 0.42 | 21000 | 1.2926 | 0.6814 |
| 1.8115 | 0.44 | 22000 | 1.2995 | 0.6636 |
| 1.7980 | 0.46 | 23000 | 1.2844 | 0.6669 |
| 1.7737 | 0.48 | 24000 | 1.3020 | 0.6591 |
| 1.6655 | 0.5 | 25000 | 1.2727 | 0.6544 |
| 1.7493 | 0.52 | 26000 | 1.2830 | 0.6547 |
| 1.7016 | 0.54 | 27000 | 1.2820 | 0.6497 |
| 1.6763 | 0.56 | 28000 | 1.2767 | 0.6473 |
| 1.7027 | 0.58 | 29000 | 1.2730 | 0.6454 |
| 1.7984 | 0.6 | 30000 | 1.2610 | 0.6471 |
| 1.7301 | 0.62 | 31000 | 1.2475 | 0.6438 |
| 1.7133 | 0.64 | 32000 | 1.2569 | 0.6342 |
| 1.6079 | 0.66 | 33000 | 1.2468 | 0.6314 |
| 1.8220 | 0.68 | 34000 | 1.2299 | 0.6355 |
| 1.6110 | 0.7 | 35000 | 1.2369 | 0.6310 |
| 1.6863 | 0.72 | 36000 | 1.2286 | 0.6357 |
| 1.5938 | 0.74 | 37000 | 1.2285 | 0.6288 |
| 1.7096 | 0.76 | 38000 | 1.2211 | 0.6266 |
| 1.5388 | 0.78 | 39000 | 1.2222 | 0.6246 |
| 1.6024 | 0.8 | 40000 | 1.2288 | 0.6268 |
| 1.5912 | 0.82 | 41000 | 1.2150 | 0.6285 |
| 1.6931 | 0.84 | 42000 | 1.2121 | 0.6220 |
| 1.6600 | 0.86 | 43000 | 1.2134 | 0.6246 |
| 1.6601 | 0.88 | 44000 | 1.2062 | 0.6222 |
| 1.6800 | 0.9 | 45000 | 1.2109 | 0.6198 |
| 1.5803 | 0.92 | 46000 | 1.2080 | 0.6222 |
| 1.6128 | 0.94 | 47000 | 1.2034 | 0.6216 |
| 1.5225 | 0.96 | 48000 | 1.2063 | 0.6210 |
| 1.6296 | 0.98 | 49000 | 1.2091 | 0.6219 |
| 1.6504 | 1.0 | 50000 | 1.2102 | 0.6214 |
Framework versions
- Transformers 5.10.2
- Pytorch 2.10.0+rocm7.2.4.git3d3aa833
- Datasets 4.0.0
- Tokenizers 0.22.2