You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

wav2vec2-asr-africa-base-finetuned-naijavoices-hausa-v0.0

This model is a fine-tuned version of asr-africa/wav2vec2-asr-africa-base on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2168
  • Wer: 0.2254
  • Cer: 0.0572

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: 9e-05
  • train_batch_size: 64
  • eval_batch_size: 32
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 128
  • optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.025
  • num_epochs: 100.0
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
2.9988 1.0 4339 0.6527 0.6990 0.1926
1.0919 2.0 8678 0.4615 0.5041 0.1317
0.964 3.0 13017 0.4128 0.4606 0.1195
0.9062 4.0 17356 0.3762 0.4121 0.1078
0.8672 5.0 21695 0.3638 0.3962 0.1048
0.8344 6.0 26034 0.3456 0.3750 0.0990
0.8098 7.0 30373 0.3280 0.3523 0.0923
0.7872 8.0 34712 0.3195 0.3411 0.0892
0.7692 9.0 39051 0.3122 0.3363 0.0880
0.7552 10.0 43390 0.3031 0.3259 0.0855
0.7367 11.0 47729 0.2978 0.3213 0.0837
0.7225 12.0 52068 0.2923 0.3173 0.0831
0.7095 13.0 56407 0.2868 0.3097 0.0806
0.6977 14.0 60746 0.2814 0.3066 0.0796
0.6859 15.0 65085 0.2791 0.2976 0.0779
0.6761 16.0 69424 0.2757 0.2994 0.0778
0.6656 17.0 73763 0.2686 0.2859 0.0742
0.6573 18.0 78102 0.2683 0.2862 0.0748
0.6493 19.0 82441 0.2671 0.2880 0.0752
0.6415 20.0 86780 0.2618 0.2810 0.0727
0.6352 21.0 91119 0.2578 0.2751 0.0715
0.6267 22.0 95458 0.2571 0.2782 0.0720
0.6203 23.0 99797 0.2546 0.2772 0.0716
0.6142 24.0 104136 0.2510 0.2687 0.0696
0.6082 25.0 108475 0.2509 0.2763 0.0708
0.6026 26.0 112814 0.2481 0.2695 0.0695
0.598 27.0 117153 0.2459 0.2621 0.0678
0.5954 28.0 121492 0.2481 0.2632 0.0684
0.5884 29.0 125831 0.2447 0.2652 0.0683
0.5824 30.0 130170 0.2459 0.2583 0.0671
0.5787 31.0 134509 0.2433 0.2631 0.0679
0.5756 32.0 138848 0.2412 0.2589 0.0672
0.5709 33.0 143187 0.2398 0.2608 0.0673
0.5652 34.0 147526 0.2384 0.2573 0.0661
0.5598 35.0 151865 0.2369 0.2568 0.0659
0.5567 36.0 156204 0.2374 0.2555 0.0659
0.5521 37.0 160543 0.2348 0.2537 0.0650
0.5489 38.0 164882 0.2351 0.2526 0.0653
0.545 39.0 169221 0.2331 0.2531 0.0648
0.5409 40.0 173560 0.2327 0.2510 0.0644
0.5384 41.0 177899 0.2314 0.2494 0.0639
0.5355 42.0 182238 0.2306 0.2494 0.0642
0.5323 43.0 186577 0.2295 0.2476 0.0635
0.529 44.0 190916 0.2285 0.2442 0.0628
0.5265 45.0 195255 0.2275 0.2428 0.0622
0.5226 46.0 199594 0.2270 0.2451 0.0630
0.5187 47.0 203933 0.2273 0.2415 0.0621
0.5155 48.0 208272 0.2268 0.2465 0.0627
0.5128 49.0 212611 0.2246 0.2433 0.0621
0.5101 50.0 216950 0.2257 0.2408 0.0617
0.5094 51.0 221289 0.2250 0.2382 0.0611
0.5047 52.0 225628 0.2241 0.2418 0.0620
0.502 53.0 229967 0.2236 0.2391 0.0614
0.4989 54.0 234306 0.2239 0.2394 0.0613
0.4962 55.0 238645 0.2236 0.2406 0.0613
0.494 56.0 242984 0.2224 0.2412 0.0612
0.4915 57.0 247323 0.2222 0.2383 0.0611
0.4896 58.0 251662 0.2224 0.2368 0.0605
0.4871 59.0 256001 0.2215 0.2353 0.0603
0.4844 60.0 260340 0.2208 0.2371 0.0606
0.4811 61.0 264679 0.2196 0.2362 0.0601
0.4797 62.0 269018 0.2221 0.2349 0.0598
0.4776 63.0 273357 0.2194 0.2351 0.0598
0.4746 64.0 277696 0.2186 0.2374 0.0601
0.4745 65.0 282035 0.2185 0.2321 0.0589
0.4739 66.0 286374 0.2200 0.2304 0.0587
0.4705 67.0 290713 0.2189 0.2343 0.0596
0.4679 68.0 295052 0.2183 0.2331 0.0595
0.4653 69.0 299391 0.2196 0.2292 0.0586
0.4642 70.0 303730 0.2166 0.2335 0.0591
0.462 71.0 308069 0.2181 0.2321 0.0590
0.46 72.0 312408 0.2192 0.2296 0.0586
0.4592 73.0 316747 0.2178 0.2315 0.0589
0.4571 74.0 321086 0.2172 0.2335 0.0592
0.4548 75.0 325425 0.2174 0.2294 0.0585
0.4524 76.0 329764 0.2171 0.2316 0.0591
0.4513 77.0 334103 0.2168 0.2295 0.0585
0.4496 78.0 338442 0.2167 0.2277 0.0581
0.4476 79.0 342781 0.2177 0.2307 0.0584
0.4466 80.0 347120 0.2160 0.2290 0.0581
0.4447 81.0 351459 0.2159 0.2289 0.0580
0.4436 82.0 355798 0.2154 0.2276 0.0578
0.4419 83.0 360137 0.2145 0.2269 0.0576
0.4405 84.0 364476 0.2149 0.2261 0.0574
0.4382 85.0 368815 0.2158 0.2262 0.0575
0.4365 86.0 373154 0.2165 0.2271 0.0577
0.4349 87.0 377493 0.2164 0.2257 0.0573
0.4349 88.0 381832 0.2168 0.2254 0.0572

Framework versions

  • Transformers 4.51.3
  • Pytorch 2.6.0+cu124
  • Datasets 3.5.0
  • Tokenizers 0.21.1
Downloads last month
0
Safetensors
Model size
94.4M params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for asr-africa/wav2vec2-asr-africa-base-finetuned-naijavoices-hausa-v0.0

Finetuned
(3)
this model