metadata
language:
- uz
license: apache-2.0
tags:
- automatic-speech-recognition
- mozilla-foundation/common_voice_10_0
- generated_from_trainer
datasets:
- common_voice_10_0
base_model: facebook/wav2vec2-xls-r-300m
model-index:
- name: xls-r-uzbek-cv10
results: []
xls-r-uzbek-cv10
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMON_VOICE_10_0 - UZ dataset. It achieves the following results on the evaluation set:
- Loss: 0.2491
- Wer: 0.2588
- Cer: 0.0513
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: 3e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- 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: 500
- num_epochs: 100.0
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Cer | Validation Loss | Wer |
---|---|---|---|---|---|
3.1215 | 0.68 | 500 | 1.0 | 3.1188 | 1.0 |
2.8562 | 1.36 | 1000 | 0.9689 | 2.5724 | 1.0002 |
1.2709 | 2.04 | 1500 | 0.1471 | 0.6278 | 0.6478 |
1.0817 | 2.72 | 2000 | 0.1304 | 0.4989 | 0.5931 |
0.9801 | 3.4 | 2500 | 0.1225 | 0.4582 | 0.5667 |
0.951 | 4.08 | 3000 | 0.1149 | 0.4239 | 0.5381 |
0.8834 | 4.76 | 3500 | 0.1092 | 0.4016 | 0.5158 |
0.857 | 5.44 | 4000 | 0.1047 | 0.3785 | 0.4992 |
0.8307 | 6.12 | 4500 | 0.1004 | 0.3720 | 0.4811 |
0.805 | 6.8 | 5000 | 0.0937 | 0.3450 | 0.4537 |
0.7828 | 7.48 | 5500 | 0.0912 | 0.3421 | 0.4460 |
0.7789 | 8.16 | 6000 | 0.0890 | 0.3295 | 0.4337 |
0.755 | 8.84 | 6500 | 0.0862 | 0.3257 | 0.4222 |
0.7464 | 9.52 | 7000 | 0.0847 | 0.3269 | 0.4155 |
0.7293 | 10.2 | 7500 | 0.0823 | 0.3121 | 0.4025 |
0.7283 | 10.88 | 8000 | 0.0789 | 0.2991 | 0.3941 |
0.7145 | 11.56 | 8500 | 0.0786 | 0.2961 | 0.3868 |
0.6963 | 12.24 | 9000 | 0.0767 | 0.2972 | 0.3784 |
0.6981 | 12.92 | 9500 | 0.0757 | 0.2880 | 0.3750 |
0.6888 | 13.6 | 10000 | 0.0745 | 0.2865 | 0.3703 |
0.6733 | 14.29 | 10500 | 0.0744 | 0.2887 | 0.3663 |
0.6701 | 14.97 | 11000 | 0.0735 | 0.2857 | 0.3624 |
0.6634 | 15.65 | 11500 | 0.0723 | 0.2822 | 0.3581 |
0.6484 | 16.33 | 12000 | 0.0706 | 0.2778 | 0.3503 |
0.6626 | 17.01 | 12500 | 0.0697 | 0.2697 | 0.3477 |
0.6341 | 17.69 | 13000 | 0.0708 | 0.2804 | 0.3511 |
0.6402 | 18.37 | 13500 | 0.0681 | 0.2665 | 0.3418 |
0.6343 | 19.05 | 14000 | 0.0687 | 0.2748 | 0.3425 |
0.6383 | 19.73 | 14500 | 0.0677 | 0.2696 | 0.3383 |
0.6178 | 20.41 | 15000 | 0.0690 | 0.2743 | 0.3417 |
0.6097 | 21.09 | 15500 | 0.0671 | 0.2663 | 0.3352 |
0.6245 | 21.77 | 16000 | 0.0665 | 0.2685 | 0.3318 |
0.6137 | 22.45 | 16500 | 0.0655 | 0.2700 | 0.3262 |
0.6018 | 23.13 | 17000 | 0.0652 | 0.2657 | 0.3225 |
0.6063 | 23.81 | 17500 | 0.0663 | 0.2664 | 0.3276 |
0.5917 | 24.49 | 18000 | 0.0658 | 0.2725 | 0.3264 |
0.5984 | 25.17 | 18500 | 0.0643 | 0.2593 | 0.3197 |
0.5949 | 25.85 | 19000 | 0.0635 | 0.2581 | 0.3161 |
0.5863 | 26.53 | 19500 | 0.0639 | 0.2543 | 0.3196 |
0.5858 | 27.21 | 20000 | 0.0628 | 0.2620 | 0.3136 |
0.5902 | 27.89 | 20500 | 0.0627 | 0.2549 | 0.3157 |
0.5794 | 28.57 | 21000 | 0.0624 | 0.2543 | 0.3136 |
0.5744 | 29.25 | 21500 | 0.0620 | 0.2542 | 0.3091 |
0.5899 | 29.93 | 22000 | 0.0624 | 0.2540 | 0.3122 |
0.5597 | 30.61 | 22500 | 0.0609 | 0.2500 | 0.3057 |
0.5595 | 31.29 | 23000 | 0.0616 | 0.2539 | 0.3087 |
0.5664 | 31.97 | 23500 | 0.0610 | 0.2504 | 0.3070 |
0.5608 | 32.65 | 24000 | 0.0611 | 0.2535 | 0.3066 |
0.5557 | 33.33 | 24500 | 0.0608 | 0.2538 | 0.3047 |
0.5741 | 34.01 | 25000 | 0.0596 | 0.2480 | 0.3009 |
0.5614 | 34.69 | 25500 | 0.0601 | 0.2516 | 0.3033 |
0.5436 | 35.37 | 26000 | 0.0601 | 0.2540 | 0.3004 |
0.555 | 36.05 | 26500 | 0.0595 | 0.2486 | 0.2993 |
0.5474 | 36.73 | 27000 | 0.0598 | 0.2536 | 0.3003 |
0.5352 | 37.41 | 27500 | 0.0597 | 0.2589 | 0.2986 |
0.5489 | 38.1 | 28000 | 0.0586 | 0.2485 | 0.2925 |
0.5438 | 38.77 | 28500 | 0.0581 | 0.2500 | 0.2908 |
0.541 | 39.46 | 29000 | 0.0577 | 0.2451 | 0.2879 |
0.5462 | 40.14 | 29500 | 0.0581 | 0.2510 | 0.2935 |
0.529 | 40.82 | 30000 | 0.0575 | 0.2435 | 0.2879 |
0.5169 | 41.5 | 30500 | 0.0572 | 0.2474 | 0.2860 |
0.5281 | 42.18 | 31000 | 0.0575 | 0.2478 | 0.2884 |
0.527 | 42.86 | 31500 | 0.0568 | 0.2492 | 0.2845 |
0.5172 | 43.54 | 32000 | 0.0575 | 0.2451 | 0.2885 |
0.5154 | 44.22 | 32500 | 0.0574 | 0.2490 | 0.2873 |
0.5129 | 44.9 | 33000 | 0.0569 | 0.2446 | 0.2853 |
0.5075 | 45.58 | 33500 | 0.0565 | 0.2485 | 0.2828 |
0.5077 | 46.26 | 34000 | 0.0559 | 0.2452 | 0.2807 |
0.5004 | 46.94 | 34500 | 0.0572 | 0.2501 | 0.2882 |
0.5319 | 47.62 | 35000 | 0.0575 | 0.2516 | 0.2856 |
0.4956 | 48.3 | 35500 | 0.0567 | 0.2495 | 0.2821 |
0.5053 | 48.98 | 36000 | 0.0565 | 0.2482 | 0.2825 |
0.5014 | 49.66 | 36500 | 0.0559 | 0.2441 | 0.2808 |
0.4945 | 50.34 | 37000 | 0.0562 | 0.2460 | 0.2807 |
0.51 | 51.02 | 37500 | 0.0547 | 0.2434 | 0.2741 |
0.5095 | 51.7 | 38000 | 0.0558 | 0.2434 | 0.2790 |
0.5026 | 52.38 | 38500 | 0.0560 | 0.2478 | 0.2787 |
0.5081 | 53.06 | 39000 | 0.0566 | 0.2485 | 0.2821 |
0.5021 | 53.74 | 39500 | 0.0551 | 0.2410 | 0.2752 |
0.4945 | 54.42 | 40000 | 0.0552 | 0.2436 | 0.2766 |
0.4882 | 55.1 | 40500 | 0.0555 | 0.2438 | 0.2769 |
0.497 | 55.78 | 41000 | 0.0550 | 0.2423 | 0.2758 |
0.4925 | 56.46 | 41500 | 0.0560 | 0.2474 | 0.2790 |
0.4894 | 57.14 | 42000 | 0.0559 | 0.2497 | 0.2797 |
0.4767 | 57.82 | 42500 | 0.0556 | 0.2528 | 0.2800 |
0.4796 | 58.5 | 43000 | 0.0549 | 0.2463 | 0.2755 |
0.4767 | 59.18 | 43500 | 0.0548 | 0.2452 | 0.2753 |
0.4786 | 59.86 | 44000 | 0.0551 | 0.2480 | 0.2769 |
0.4804 | 60.54 | 44500 | 0.0556 | 0.2514 | 0.2789 |
0.4794 | 61.22 | 45000 | 0.0539 | 0.2391 | 0.2715 |
0.4789 | 61.9 | 45500 | 0.0546 | 0.2461 | 0.2725 |
0.4683 | 62.58 | 46000 | 0.0541 | 0.2444 | 0.2707 |
0.4721 | 63.27 | 46500 | 0.0539 | 0.2468 | 0.2693 |
0.4792 | 63.94 | 47000 | 0.0546 | 0.2479 | 0.2738 |
0.4712 | 64.63 | 47500 | 0.0547 | 0.2466 | 0.2742 |
0.4607 | 65.31 | 48000 | 0.0539 | 0.2503 | 0.2707 |
0.4712 | 65.99 | 48500 | 0.0543 | 0.2458 | 0.2718 |
0.4647 | 66.67 | 49000 | 0.0538 | 0.2474 | 0.2693 |
0.4736 | 67.35 | 49500 | 0.0541 | 0.2514 | 0.2696 |
0.4718 | 68.03 | 50000 | 0.0540 | 0.2506 | 0.2692 |
0.4695 | 68.71 | 50500 | 0.0538 | 0.2499 | 0.2675 |
0.4549 | 69.39 | 51000 | 0.0534 | 0.2491 | 0.2669 |
0.4605 | 70.07 | 51500 | 0.0532 | 0.2497 | 0.2660 |
0.4538 | 70.75 | 52000 | 0.0536 | 0.2472 | 0.2684 |
0.4571 | 71.43 | 52500 | 0.0523 | 0.2441 | 0.2629 |
0.4608 | 72.11 | 53000 | 0.0529 | 0.2469 | 0.2652 |
0.4541 | 72.79 | 53500 | 0.0533 | 0.2498 | 0.2673 |
0.4424 | 73.47 | 54000 | 0.0530 | 0.2504 | 0.2658 |
0.4482 | 74.15 | 54500 | 0.0534 | 0.2517 | 0.2684 |
0.4554 | 74.83 | 55000 | 0.0529 | 0.2471 | 0.2656 |
0.444 | 75.51 | 55500 | 0.0535 | 0.2493 | 0.2675 |
0.4464 | 76.19 | 56000 | 0.0524 | 0.2461 | 0.2635 |
0.4436 | 76.87 | 56500 | 0.0526 | 0.2479 | 0.2641 |
0.4432 | 77.55 | 57000 | 0.0526 | 0.2513 | 0.2641 |
0.4459 | 78.23 | 57500 | 0.0521 | 0.2460 | 0.2625 |
0.4433 | 78.91 | 58000 | 0.0521 | 0.2457 | 0.2622 |
0.4407 | 79.59 | 58500 | 0.0528 | 0.2531 | 0.2659 |
0.4389 | 80.27 | 59000 | 0.0521 | 0.2485 | 0.2631 |
0.4384 | 80.95 | 59500 | 0.0522 | 0.2502 | 0.2653 |
0.4306 | 81.63 | 60000 | 0.0528 | 0.2480 | 0.2665 |
0.4505 | 82.31 | 60500 | 0.0523 | 0.2461 | 0.2637 |
0.4442 | 82.99 | 61000 | 0.0523 | 0.2519 | 0.2641 |
0.4349 | 83.67 | 61500 | 0.0522 | 0.2509 | 0.2625 |
0.4398 | 84.35 | 62000 | 0.0523 | 0.2510 | 0.2659 |
0.4398 | 85.03 | 62500 | 0.0526 | 0.2507 | 0.2648 |
0.4355 | 85.71 | 63000 | 0.0523 | 0.2500 | 0.2653 |
0.4373 | 86.39 | 63500 | 0.0524 | 0.2523 | 0.2650 |
0.4391 | 87.07 | 64000 | 0.0523 | 0.2509 | 0.2635 |
0.4381 | 87.75 | 64500 | 0.0521 | 0.2502 | 0.2635 |
0.4297 | 88.43 | 65000 | 0.0521 | 0.2521 | 0.2632 |
0.44 | 89.12 | 65500 | 0.0520 | 0.2507 | 0.2624 |
0.4313 | 89.8 | 66000 | 0.0519 | 0.2497 | 0.2623 |
0.4402 | 90.48 | 66500 | 0.0517 | 0.2488 | 0.2608 |
0.4324 | 91.16 | 67000 | 0.0512 | 0.2485 | 0.2585 |
0.4317 | 91.84 | 67500 | 0.0513 | 0.2488 | 0.2587 |
0.437 | 92.52 | 68000 | 0.0513 | 0.2473 | 0.2590 |
0.4389 | 93.2 | 68500 | 0.0512 | 0.2472 | 0.2581 |
0.4428 | 93.88 | 69000 | 0.0512 | 0.2475 | 0.2587 |
0.4294 | 94.56 | 69500 | 0.0513 | 0.2489 | 0.2596 |
0.4247 | 95.24 | 70000 | 0.0515 | 0.2499 | 0.2597 |
0.4309 | 95.92 | 70500 | 0.0514 | 0.2493 | 0.2590 |
0.4366 | 96.6 | 71000 | 0.0512 | 0.2492 | 0.2592 |
0.4245 | 97.28 | 71500 | 0.0513 | 0.2493 | 0.2587 |
0.4346 | 97.96 | 72000 | 0.0512 | 0.2478 | 0.2583 |
0.4289 | 98.64 | 72500 | 0.0512 | 0.2489 | 0.2585 |
0.4246 | 99.32 | 73000 | 0.0513 | 0.2487 | 0.2589 |
0.4241 | 100.0 | 73500 | 0.0513 | 0.2491 | 0.2588 |
Framework versions
- Transformers 4.16.2
- Pytorch 1.12.0
- Datasets 2.4.0
- Tokenizers 0.10.3
Credits
Author: Shukrullo Turgunov (aka Vodiylik)