digo-prayudha/Llama-3.2-1B-Indonesian-lora

Llama-3.2-1B-Indonesian

This model is a fine-tuned version of meta-llama/Llama-3.2-1B-Instruct that has been optimized for Indonesian language understanding and generation.

The fine-tuning process utilized Low-Rank Adaptation (LoRA) to efficiently adapt the model while minimizing computational and storage overhead. This approach enables effective fine-tuning for specific tasks or domains, particularly in the Indonesian language context.

Training and evaluation data

Ichsan2895/alpaca-gpt4-indonesian

Use WIth Transformers

import torch
from transformers import pipeline

model_id = "digo-prayudha/Llama-3.2-1B-Indonesian"
pipe = pipeline(
    "text-generation",
    model=model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
messages = [
    {"role": "user", "content": "Tentukan subjek dari kalimat berikut: 'Film tersebut dirilis kemarin'."},
]
outputs = pipe(
    messages,
    max_new_tokens=256,
)
print(outputs[0]["generated_text"][-1])

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0002
• train_batch_size: 1
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 6
• total_train_batch_size: 6
• optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

![Train Loss]

Framework versions

• PEFT 0.7.2.dev0
• Transformers 4.46.1
• Pytorch 2.4.0+cu121
• Datasets 2.16.1
• Tokenizers 0.20.1