Modelpublic

jssaluja/fb-mms-1b-punjabi-ccer-train-116393-epochs-15-test-1569

sourceHugging Facecc-by-nc-4.0updated 13d agoView on Hugging Face

0likes237downloads

Model Card

jssaluja/fb-mms-1b-punjabi-ccer-train-116393-epochs-15-test-1569

This model is a fine-tuned version of facebook/mms-1b-all on the jssaluja/rajindersinghcorrected, hustler007/trainingbhaisahibsinghcanadasehajpaath, hustler007/trainingbhaijagtarsinghjachaksehajpaath, hustler007/trainingbhaigurpreetsinghnalvisehajpaath, hustler007/trainingbhaisarwansinghsehajpaath, hustler007/trainingparmindersinghaustraliasehajpaathkirtan, hustler007/trainingbhaiharjindersinghsrinagar, hustler007/trainingbarusahibsehajpaathtantisaazkirtan, hustler007/trainingbhainiranjansinghjawaddikalan, hustler007/traininggurdwarasahibfremontpaath, hustler007/traininggurdwarasahibfremontkirtan, hustler007/trainingakashvanijalandharradiodoctor, hustler007/trainingjailchittian, hustler007/trainingsachisakhisirdarkapursingh, hustler007/trainingranglesajjanakjaudiobook, hustler007/trainingjsktalksdivinetalk, hustler007/trainingbhaiharpalsinghzafarnamakatha, hustler007/trainingbhaisimranjeetsinghtohanaspecialtalk, hustler007/trainingranglesajjangurbaniaudiobook, hustler007/trainingsmtvpodcastsimranjotmakkar, hustler007/trainingbijaysingh, hustler007/trainingkhalaspodcastharsharankaur, hustler007/trainingbhaipinderpalsinghsalokvaaratevadeek, hustler007/trainingsatwantkaur, hustler007/trainingsikhsiyasatdiasporarepresentatives, hustler007/trainingsikhsiyasatgurmeetsingh, hustler007/trainingwikimediapunjabiaudiobooks, hustler007/traininghumhindunahinkahnsinghnabha, hustler007/trainingheerwarisshah, hustler007/trainingbhaipinderpalsinghjaapsahib, hustler007/trainingsarbatmediapodcast, hustler007/trainingsikhsiyasathumanrights, hustler007/trainingsmtvroundtable, hustler007/trainingdiscoverpunjabexclusivestories, hustler007/trainingbhaisewasinghtarmalaakathkatha, hustler007/trainingjhuthepolicemukabledidastan, hustler007/traininggurmatfilmsgurunanaksakhiyan, hustler007/trainingmaskeenjifullkatha, hustler007/trainingpunjabichapterskhalsaraj, hustler007/trainingmehmimoviespunjabishortfilms, hustler007/trainingtuhituhipinderpalsinghkatha, hustler007/traininggurmatelitediscoversikhism, hustler007/traininggyandasagarshorts, hustler007/trainingchaupalchalmeraputt3, hustler007/trainingchaupalpardhan, hustler007/traininglokdhunpunjabilatestmovies, hustler007/trainingrajkaregakhalsashorts, hustler007/trainingjaanbaazrakhaardarshi, hustler007/trainingsafarwithnimratkaur, hustler007/trainingsikhsiyasatdrkanwaljitsingh dataset. It achieves the following results on the evaluation set:

—Loss: 0.2463
—Wer: 0.3195
—Wil: 0.5100
—Mer: 0.3139
—Cer: 0.0798
—Ncer: 0.0317
—Ccer: 0.0862

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: 0.001
—trainbatchsize: 16
—evalbatchsize: 8
—seed: 42
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: linear
—num_epochs: 15
—mixedprecisiontraining: Native AMP

Training results

(table generated directly from `trainer.state.log_history`, keyed by metric name -- avoids a transformers model-card column-ordering bug seen in earlier runs)

Training Loss	Epoch	Step	Ccer	Cer	Validation Loss	Mer	Ncer	Wer	Wil
0.5025	1.0	7274	0.1209	0.1145	0.3569	0.4282	0.0507	0.4436	0.6537
0.2881	2.0	14548	0.1025	0.0958	0.2942	0.3693	0.0419	0.3782	0.5836
0.2618	3.0	21822	0.1121	0.1028	0.3116	0.3894	0.0387	0.4008	0.6110
0.2430	4.0	29096	0.1060	0.0978	0.2971	0.3752	0.0404	0.3820	0.5909
0.2306	5.0	36370	0.0931	0.0886	0.2713	0.3442	0.0370	0.3504	0.5484
0.2194	6.0	43644	0.0911	0.0859	0.2595	0.3418	0.0374	0.3464	0.5449
0.2103	7.0	50918	0.0950	0.0879	0.2727	0.3406	0.0367	0.3469	0.5460
0.2010	8.0	58192	0.0921	0.0856	0.2637	0.3335	0.0351	0.3394	0.5364
0.1942	9.0	65466	0.0892	0.0829	0.2537	0.3281	0.0327	0.3337	0.5289
0.1867	10.0	72740	0.0886	0.0828	0.2590	0.3237	0.0334	0.3295	0.5218
0.1803	11.0	80014	0.0887	0.0828	0.2541	0.3291	0.0334	0.3334	0.5288
0.1751	12.0	87288	0.0867	0.0805	0.2469	0.3167	0.0320	0.3214	0.5135
0.1688	13.0	94562	0.0883	0.0816	0.2474	0.3192	0.0324	0.3248	0.5178
0.1646	14.0	101836	0.0874	0.0810	0.2512	0.3181	0.0324	0.3235	0.5155
0.1611	15.0	109110	0.0862	0.0798	0.2463	0.3139	0.0317	0.3195	0.5100

Framework versions

—Transformers 4.51.3
—Pytorch 2.11.0+cu130
—Datasets 5.0.0
—Tokenizers 0.21.4