### Run IndoBERTweet Hate Speech Classification Command

Source: https://context7.com/indolem/indobertweet/llms.txt

Example command to run the IndoBERTweet script for hate speech classification (HS1). Configures model, data path, batch size, epochs, learning rate, and patience.

```bash
# python hate_speech_HS1/indobertweet.py \
#   --bert_model indobertweet \
#   --data_path hate_speech_HS1/data/ \
#   --batch_size 30 \
#   --num_train_epochs 20 \
#   --learning_rate 5e-5 \
#   --patience 5
```

--------------------------------

### Load IndoBERTweet Model and Tokenizer

Source: https://github.com/indolem/indobertweet/blob/main/README.md

Load the pre-trained IndoBERTweet model and its corresponding tokenizer using the transformers library. Ensure you have transformers version 3.5.1 or compatible installed.

```python
from transformers import AutoTokenizer, AutoModel
tokenizer = AutoTokenizer.from_pretrained("indolem/indobertweet-base-uncased")
model = AutoModel.from_pretrained("indolem/indobertweet-base-uncased")
```

--------------------------------

### Instantiate and Perform Single Forward Pass with IndoBERTweet

Source: https://context7.com/indolem/indobertweet/llms.txt

Instantiate the IndoBERTweet model for 3-class emotion detection and perform a single forward pass to get logits and predictions. Ensure the model and data are on the correct device.

```python
args = argparse.Namespace(bert_model='indobertweet', vocab_label_size=3)
device = torch.device("cpu")
model = Model(args, device)
model.to(device)

# Single forward pass
src, seg, label, mask_src = Batch(dataset, 0, 2, device).get()
logits = model.forward(src, seg, mask_src)
print(logits.shape)  # torch.Size([2, 3])
preds = model.predict(src, seg, mask_src)
print(preds)         # e.g. [2, 0]
```