Instructions to use rizkyayub/rizbuy-submission-qwen3-indonesian with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use rizkyayub/rizbuy-submission-qwen3-indonesian with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="rizkyayub/rizbuy-submission-qwen3-indonesian")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("rizkyayub/rizbuy-submission-qwen3-indonesian")
model = AutoModelForCausalLM.from_pretrained("rizkyayub/rizbuy-submission-qwen3-indonesian", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use rizkyayub/rizbuy-submission-qwen3-indonesian with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "rizkyayub/rizbuy-submission-qwen3-indonesian"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "rizkyayub/rizbuy-submission-qwen3-indonesian",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/rizkyayub/rizbuy-submission-qwen3-indonesian

SGLang

How to use rizkyayub/rizbuy-submission-qwen3-indonesian with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "rizkyayub/rizbuy-submission-qwen3-indonesian" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "rizkyayub/rizbuy-submission-qwen3-indonesian",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "rizkyayub/rizbuy-submission-qwen3-indonesian" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "rizkyayub/rizbuy-submission-qwen3-indonesian",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Desktop
Docker Model Runner
How to use rizkyayub/rizbuy-submission-qwen3-indonesian with Docker Model Runner:
```
docker model run hf.co/rizkyayub/rizbuy-submission-qwen3-indonesian
```

rizbuy-submission-qwen3-indonesian

Model instruksi bahasa Indonesia hasil fine-tuning dari unsloth/Qwen3-4B menggunakan pustaka Unsloth dan TRL (SFTTrainer). Model dilatih menggunakan metode QLoRA 4-bit dan diekspor menggunakan metode bobot gabungan 16-bit (merged 16-bit) untuk kompatibilitas inferensi langsung tanpa memerlukan dependensi LoRA terpisah.

Spesifikasi Model

Model Dasar: unsloth/Qwen3-4B
Tipe Arsitektur: Causal Language Model (Transformer)
Format Bobot: 16-bit Merged (merged_16bit Safetensors)
Panjang Konteks Maksimal: 1024 token
Bahasa: Indonesia (id)
Lisensi: Apache-2.0

Konfigurasi Pelatihan & Hyperparameter

Model ini dilatih dengan konfigurasi teknis berikut:

1. Dataset

Sumber Data: Ichsan2895/alpaca-gpt4-indonesian
Distribusi Data:
- Data Latih: 40.000 sampel (acak dengan seed 3407)
- Data Evaluasi: 200 sampel (acak dengan seed 3407)
Format Pesan: Chat template qwen2.5 dengan system prompt:

"Kamu adalah asisten AI yang membantu menjawab pertanyaan pengguna berdasarkan data dan fakta yang kamu miliki."

2. Parameter LoRA (PEFT)

Rank ($r$): 8
LoRA Alpha: 16
LoRA Dropout: 0
Target Modules: ["q_proj", "k_proj", "v_proj", "o_proj"]
Bias: none
Gradient Checkpointing: Unsloth native engine

3. Hyperparameter SFTTrainer

Optimizer: paged_adamw_8bit
Learning Rate: $1\times 10^{-5}$ – $2\times 10^{-5}$
LR Scheduler: cosine
Warmup Steps: 50 – 70 langkah
Per-Device Batch Size: 1
Gradient Accumulation Steps: 4 – 8 langkah
Max Steps: 800 langkah
Precision: Mixed Precision (BF16 native jika didukung perangkat keras)

Cara Inferensi

Berikut adalah contoh inferensi menggunakan Hugging Face Transformers sesuai dengan parameter decoding yang digunakan saat validasi:

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer

model_id = "rizkyayub/rizbuy-submission-qwen3-indonesian"

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16 if torch.cuda.is_bf16_supported() else torch.float16,
    device_map="auto"
)

system_prompt = "Kamu adalah asisten AI yang membantu menjawab pertanyaan pengguna berdasarkan data dan fakta yang kamu miliki."

messages = [
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": "Sebutkan 3 bandara terkenal di Indonesia beserta lokasinya!"}
]

inputs = tokenizer.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    return_tensors="pt"
).to(model.device)

attention_mask = torch.ones_like(inputs)
streamer = TextStreamer(tokenizer, skip_prompt=True)

terminators = [tokenizer.eos_token_id] if tokenizer.eos_token_id is not None else [151643]
pad_id = tokenizer.pad_token_id if tokenizer.pad_token_id is not None else terminators[0]

_ = model.generate(
    input_ids=inputs,
    attention_mask=attention_mask,
    streamer=streamer,
    max_new_tokens=256,
    temperature=0.1,
    top_k=20,
    top_p=0.9,
    repetition_penalty=1.05,
    do_sample=True,
    eos_token_id=terminators,
    pad_token_id=pad_id,
)

Downloads last month: 350

Safetensors

Model size

4B params

Tensor type

BF16

Model tree for rizkyayub/rizbuy-submission-qwen3-indonesian

Base model

Qwen/Qwen3-4B-Base

Finetuned

Qwen/Qwen3-4B

Finetuned

unsloth/Qwen3-4B

Adapter

(50)

this model