| language: | |
| - id | |
| - en | |
| tags: | |
| - gemma4 | |
| - aec | |
| - construction | |
| - fine-tuned | |
| base_model: google/gemma-4-E2B-it | |
| library_name: peft | |
| # Gemma 4 AEC - Fine-tuned Model | |
| Model ini di-fine-tune dari `google/gemma-4-E2B-it` menggunakan dataset AEC (Architecture, Engineering, Construction) Indonesia. | |
| ## Dataset | |
| - Sumber: `riosst/gemma4-aec-dataset` | |
| - Jumlah contoh: 10 | |
| - Bahasa: Indonesia | |
| - Domain: Konstruksi, SNI, Peraturan PUPR | |
| ## Cara Menggunakan | |
| ```python | |
| from transformers import AutoModelForCausalLM, AutoTokenizer | |
| model = AutoModelForCausalLM.from_pretrained("riosst/gemma4-aec") | |
| tokenizer = AutoTokenizer.from_pretrained("riosst/gemma4-aec") | |
| # Generate | |
| inputs = tokenizer("Apa itu SNI untuk beton?", return_tensors="pt") | |
| outputs = model.generate(**inputs, max_new_tokens=200) | |
| print(tokenizer.decode(outputs[0], skip_special_tokens=True)) | |
| ``` | |
| ## Fine-tuning Details | |
| - Method: LoRA (QLoRA 4-bit) | |
| - Base Model: google/gemma-4-E2B-it | |
| - Epochs: 3 | |
| - Learning Rate: 2e-4 | |
| - LoRA Rank: 16 | |