Model Card for Model ID
This model is the fine-tuned version of Whisper-large-v2 model for Speech Recognition task for vulnerable individuals in Tamil.
Preprocessing [optional]
Training Hyperparameters
** training_args = Seq2SeqTrainingArguments(
output_dir="./pretrainedwhisper-medium-native-v2", # change to a repo name of your choice
per_device_train_batch_size=4,
gradient_accumulation_steps=1, # increase by 2x for every 2x decrease in batch size
learning_rate=1e-5,
warmup_steps=200,
max_steps=2000,
gradient_checkpointing=True,
fp16=True,
evaluation_strategy="steps",
per_device_eval_batch_size=8,
predict_with_generate=True,
generation_max_length=225,
save_steps=500,
eval_steps=500,
logging_steps=25,
report_to=["tensorboard"],
load_best_model_at_end=True,
metric_for_best_model="wer",
greater_is_better=False,
push_to_hub=True,
optim="adamw_bnb_8bit"
)
Metrics
WER is the evaluation metrics used here.
