#load pretrained processor and model
processor = Wav2Vec2Processor.from_pretrained("airesearch/wav2vec2-large-xlsr-53-th")
model = Wav2Vec2ForCTC.from_pretrained("airesearch/wav2vec2-large-xlsr-53-th")

#function to resample to 16_000
def speech_file_to_array_fn(batch, 
                            text_col="sentence", 
                            fname_col="path",
                            resampling_to=16000):
    speech_array, sampling_rate = torchaudio.load(batch[fname_col])
    resampler=torchaudio.transforms.Resample(sampling_rate, resampling_to)
    batch["speech"] = resampler(speech_array)[0].numpy()
    batch["sampling_rate"] = resampling_to
    batch["target_text"] = batch[text_col]
    return batch

#get 2 examples as sample input
test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)

#infer
with torch.no_grad():
    logits = model(inputs.input_values,).logits

predicted_ids = torch.argmax(logits, dim=-1)

print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["sentence"][:2])

>> Prediction: ['และ เขา ก็ สัมผัส ดีบุก', 'คุณ สามารถ รับทราบ เมื่อ ข้อความ นี้ ถูก อ่าน แล้ว']
>> Reference: ['และเขาก็สัมผัสดีบุก', 'คุณสามารถรับทราบเมื่อข้อความนี้ถูกอ่านแล้ว']

DatasetDict({
    train: Dataset({
        features: ['path', 'sentence'],
        num_rows: 86586
    })
    test: Dataset({
        features: ['path', 'sentence'],
        num_rows: 2502
    })
    validation: Dataset({
        features: ['path', 'sentence'],
        num_rows: 3027
    })
})

# create model
model = Wav2Vec2ForCTC.from_pretrained(
    "facebook/wav2vec2-large-xlsr-53",
    attention_dropout=0.1,
    hidden_dropout=0.1,
    feat_proj_dropout=0.0,
    mask_time_prob=0.05,
    layerdrop=0.1,
    gradient_checkpointing=True,
    ctc_loss_reduction="mean",
    pad_token_id=processor.tokenizer.pad_token_id,
    vocab_size=len(processor.tokenizer)
)
model.freeze_feature_extractor()
training_args = TrainingArguments(
    output_dir="../data/wav2vec2-large-xlsr-53-thai",
    group_by_length=True,
    per_device_train_batch_size=32,
    gradient_accumulation_steps=1,
    per_device_eval_batch_size=16,
    metric_for_best_model='wer',
    evaluation_strategy="steps",
    eval_steps=1000,
    logging_strategy="steps",
    logging_steps=1000,
    save_strategy="steps",
    save_steps=1000,
    num_train_epochs=100,
    fp16=True,
    learning_rate=1e-4,
    warmup_steps=1000,
    save_total_limit=3,
    report_to="tensorboard"
)