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.

Configuration Parsing Warning:In adapter_config.json: "peft.task_type" must be a string

bm-whisper-large-v4-training-bm-lora-2

This model is a fine-tuned version of djelia/bm-whisper-large-v3-tuned on the djelia/bambara-asr, the djelia/bambara-asr and the djelia/bambara-synthetic-audio datasets.

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0003
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
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: cosine
lr_scheduler_warmup_ratio: 0.05
num_epochs: 2

Framework versions

PEFT 0.17.0
Transformers 4.55.0
Pytorch 2.8.0+cu128
Datasets 3.6.0
Tokenizers 0.21.4

Downloads last month: -

Model tree for djelia/bm-whisper-large-v4-training-bm-lora-2

Base model

djelia/bm-whisper-large-v3-tuned

Adapter

(2)

this model

Datasets used to train djelia/bm-whisper-large-v4-training-bm-lora-2

Evaluation results

wer on sudoping01/bambara-speech-recognition-benchmark/default/eval
self-reported

0.513
cer on sudoping01/bambara-speech-recognition-benchmark/default/eval
self-reported

0.222
wer on djelia/bambara-asr/bm-to-bm-code-switching/test
test set self-reported

0.290
cer on djelia/bambara-asr/bm-to-bm-code-switching/test
test set self-reported

0.108
wer on djelia/bambara-synthetic-audio/semi-clean/train[-1000:]
self-reported

0.173
cer on djelia/bambara-synthetic-audio/semi-clean/train[-1000:]
self-reported

0.105
wer on djelia/bambara-asr/fr-to-fr/test[:1000]
self-reported

0.127
cer on djelia/bambara-asr/fr-to-fr/test[:1000]
self-reported

0.036
wer on djelia/bambara-asr/en-to-en/test[:1000]
self-reported

0.030
cer on djelia/bambara-asr/en-to-en/test[:1000]
self-reported

0.010
wer on djelia/bambara-asr/bm-to-bm/test[:1000]
self-reported

0.203
cer on djelia/bambara-asr/bm-to-bm/test[:1000]
self-reported

0.089
wer on fixie-ai/common_voice_17_0/ar/test[-1000:]
self-reported

1.315
cer on fixie-ai/common_voice_17_0/ar/test[-1000:]
self-reported

1.398