muaalem-model-v0

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the None dataset. It achieves the following results on the evaluation set:

• Average Per: 0.0057
• Loss: 0.0092
• Per Ghonna: 0.0012
• Per Hams Or Jahr: 0.0013
• Per Istitala: 0.0008
• Per Itbaq: 0.0009
• Per Phonemes: 0.0027
• Per Qalqla: 0.0036
• Per Safeer: 0.0010
• Per Shidda Or Rakhawa: 0.0457
• Per Tafashie: 0.0008
• Per Tafkheem Or Taqeeq: 0.0014
• Per Tikraar: 0.0035

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: 5e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• lr_scheduler_warmup_ratio: 0.2
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Average Per	Validation Loss	Per Ghonna	Per Hams Or Jahr	Per Istitala	Per Itbaq	Per Phonemes	Per Qalqla	Per Safeer	Per Shidda Or Rakhawa	Per Tafashie	Per Tafkheem Or Taqeeq	Per Tikraar
0.1776	0.2	712	0.0338	0.0562	0.0022	0.0023	0.0015	0.0017	0.0066	0.0084	0.0016	0.3347	0.0013	0.0035	0.0077
0.0287	0.4	1424	0.0080	0.0150	0.0016	0.0016	0.0011	0.0012	0.0048	0.0056	0.0013	0.0601	0.0010	0.0034	0.0064
0.0142	0.6	2136	0.0060	0.0102	0.0014	0.0013	0.0009	0.0010	0.0032	0.0062	0.0011	0.0426	0.0008	0.0021	0.0054
0.0123	0.8	2848	0.0052	0.0105	0.0015	0.0016	0.0010	0.0012	0.0032	0.0037	0.0012	0.0359	0.0010	0.0020	0.0046
0.0112	1.0	3560	0.0057	0.0092	0.0012	0.0013	0.0008	0.0009	0.0027	0.0036	0.0010	0.0457	0.0008	0.0014	0.0035

Framework versions

• Transformers 4.55.0
• Pytorch 2.8.0+cu128
• Datasets 3.3.2
• Tokenizers 0.21.4