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.

mms-khmer-20260110

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

More information needed

More information needed

More information needed

The following hyperparameters were used during training:

learning_rate: 0.0003
train_batch_size: 4
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 2
total_train_batch_size: 16
total_eval_batch_size: 8
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_steps: 500
num_epochs: 20
mixed_precision_training: Native AMP

Training Loss	Epoch	Step	Validation Loss	Wer
1.299	1.0	3335	0.8214	1.0
1.1851	2.0	6670	0.6487	0.9877
0.9869	3.0	10005	0.5669	0.9754
0.9809	4.0	13340	0.4988	0.9446
0.8438	5.0	16675	0.4490	0.9477
0.8106	6.0	20010	0.4240	0.9538
0.7185	7.0	23345	0.4028	0.9292
0.7325	8.0	26680	0.3780	0.9169
0.5794	9.0	30015	0.3685	0.9138
0.6033	10.0	33350	0.3567	0.9108
0.6437	11.0	36685	0.3320	0.8892
0.5419	12.0	40020	0.3406	0.8677
0.4302	13.0	43355	0.3299	0.88
0.394	14.0	46690	0.3414	0.8769
0.3105	15.0	50025	0.3554	0.8892
0.389	16.0	53360	0.3650	0.8831
0.2753	17.0	56695	0.3677	0.8985
0.2678	18.0	60030	0.3707	0.8923
0.2566	19.0	63365	0.3724	0.8677
0.2602	20.0	66700	0.3863	0.9046

Safetensors

Model size

1.0B params

Tensor type

F32