Instructions to use DimasMP3/Qwen2.5-Exam-GenAI with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use DimasMP3/Qwen2.5-Exam-GenAI with Transformers:

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

pipe = pipeline("text-generation", model="DimasMP3/Qwen2.5-Exam-GenAI")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("DimasMP3/Qwen2.5-Exam-GenAI")
model = AutoModelForCausalLM.from_pretrained("DimasMP3/Qwen2.5-Exam-GenAI", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use DimasMP3/Qwen2.5-Exam-GenAI with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "DimasMP3/Qwen2.5-Exam-GenAI"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "DimasMP3/Qwen2.5-Exam-GenAI",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/DimasMP3/Qwen2.5-Exam-GenAI

SGLang

How to use DimasMP3/Qwen2.5-Exam-GenAI 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 "DimasMP3/Qwen2.5-Exam-GenAI" \
    --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": "DimasMP3/Qwen2.5-Exam-GenAI",
		"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 "DimasMP3/Qwen2.5-Exam-GenAI" \
        --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": "DimasMP3/Qwen2.5-Exam-GenAI",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Unsloth Desktop
Docker Model Runner
How to use DimasMP3/Qwen2.5-Exam-GenAI with Docker Model Runner:
```
docker model run hf.co/DimasMP3/Qwen2.5-Exam-GenAI
```

DimasMP3 commited on Feb 18

Commit

f29a6df

verified ·

1 Parent(s): 81908c8

Update README.md

Browse files

Files changed (1) hide show

README.md +74 -6

README.md CHANGED Viewed

@@ -1,21 +1,89 @@
 ---
 base_model: unsloth/Qwen2.5-7B-Instruct
 tags:
 - text-generation-inference
 - transformers
 - unsloth
 - qwen2
 license: apache-2.0
 language:
 - en
 ---
-# Uploaded finetuned  model
-- **Developed by:** DimasMP3
-- **License:** apache-2.0
-- **Finetuned from model :** unsloth/Qwen2.5-7B-Instruct
-This qwen2 model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.
-[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)

 ---
 base_model: unsloth/Qwen2.5-7B-Instruct
 tags:
+- education
+- indonesia
+- exam-generator
+- kurikulum-merdeka
 - text-generation-inference
 - transformers
 - unsloth
 - qwen2
 license: apache-2.0
 language:
+- id
 - en
 ---
+datasets:
+- DimasMP3/Indo-Elementary-School-Exams (custom)
+metrics:
+- eval_loss: 0.6869
+---
+# 🎓 Qwen2.5-7B-Indo-Exam-Generator-16bit
+[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200" align="right"/>](https://github.com/unslothai/unsloth)
+**Developed by:** [Dimas Maulana Putra (DimasMP3)](https://github.com/DimasMP3)
+**Model Type:** Specialized Fine-Tuned for Indonesian Elementary Education
+**Training Status:** Optimal @ Step 500 (Eval Loss: 0.686)
+---
+## 🌟 Overview
+**Qwen2.5-Indo-Exam-Generator** adalah model bahasa yang telah di-finetuning secara khusus untuk menjawab tantangan dunia pendidikan di Indonesia. Model ini dilatih menggunakan dataset berkualitas tinggi sebanyak **4.500+ soal sekolah dasar** yang disesuaikan dengan **Kurikulum Merdeka**.
+Model ini bukan sekadar AI umum; ia dirancang untuk berperan sebagai **Guru Digital** yang mampu menghasilkan soal ujian (pilihan ganda) lengkap dengan kunci jawaban dan pembahasan yang akurat.
+### ✨ Key Features
+- 🇮🇩 **Native Indonesian Support:** Memahami istilah pendidikan lokal (IPAS, HOTS, Kurikulum Merdeka).
+- 📝 **Structured Output:** Konsisten dalam menghasilkan format Soal, Opsi (A-D), Kunci, dan Pembahasan.
+- 🧠 **RAG Ready:** Dioptimalkan untuk bekerja dengan sistem *Retrieval-Augmented Generation* (pgvector/Drizzle).
+- 🚀 **High Precision:** Fine-tuned dalam format 16-bit untuk akurasi logika yang tajam.
+---
+## 📊 Training Results (WandB Metrics)
+Training dilakukan dengan pengawasan ketat terhadap *Validation Loss* untuk mencegah halusinasi:
+| Metric | Value |
+| :--- | :--- |
+| **Best Step** | 500 |
+| **Validation Loss** | **0.6869** |
+| **Training Loss** | 0.3204 |
+| **Epoch** | 2.43 |
+> **Note:** Berhenti otomatis via *Early Stopping* di Step 650 untuk memastikan bobot terbaik (Step 500) yang tersimpan.
+---
+## 🛠️ Tech Stack
+Model ini lahir dari perpaduan teknologi mutakhir:
+- **Base Model:** `unsloth/Qwen2.5-7B-Instruct`
+- **Fine-tuning Tool:** [Unsloth](https://github.com/unslothai/unsloth) (2x faster training)
+- **Framework:** Huggingface TRL & Transformers
+- **Optimization:** LoRA (Rank 128)
+---
+## 📖 Cara Penggunaan (Inference)
+```python
+from unsloth import FastLanguageModel
+import torch
+model, tokenizer = FastLanguageModel.from_pretrained(
+    model_name = "DimasMP3/Qwen2.5-7B-Indo-Exam-Generator-16bit",
+    max_seq_length = 2048,
+    load_in_4bit = True,
+)
+prompt = """<|im_start|>system
+Anda adalah Guru SD yang ahli. Buatlah soal pilihan ganda berdasarkan konteks materi ini.<|im_end|>
+<|im_start|>user
+Topik: Ekosistem Laut
+Konteks: Terumbu karang adalah tempat tinggal ikan.<|im_end|>
+<|im_start|>assistant
+"""
+# Generate Output...