Instructions to use iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("google/gemma-2-2b")
model = PeftModel.from_pretrained(base_model, "iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP")

Transformers

How to use iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP with Transformers:

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

pipe = pipeline("text-generation", model="iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP

SGLang

How to use iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP 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 "iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP with Docker Model Runner:
```
docker model run hf.co/iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP
```

Model Card for Model ID

Model Details

Model Description

TripleBits-Sinhala-Gemma-2-2b is a Gemma-2-2B-based model that has undergone continual pretraining (CPT) on a diverse Sinhala corpus.

Trained by: Team TripleBits for Shared Task 2025

How to Get Started with the Model

Use the code below to get started with the model.

from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
from huggingface_hub import login

login(token="your_hf_token")

# Load base model
base_model = AutoModelForCausalLM.from_pretrained("google/gemma-2-2b", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("google/gemma-2-2b")

# Load fine-tuned LoRA adapter
model = PeftModel.from_pretrained(base_model, "iCIIT/TripleBits-Sinhala-Gemma-2-2b-CP")

question = "ශ්‍රී ලංකාවේ අගනුවර කුමක්ද?"

instruction = f"""පහත සදහන් ප්‍රශ්නයට නිවැරදි පිළිතුරක් ලබා දෙන්න. පිළිතුරු ලබා දීමේදී ප්‍රශ්නයේ ස්වභාවය අනුව - සරල ප්‍රශ්න සඳහා කෙටි පිළිතුරු ද, සංකීර්ණ ප්‍රශ්න සඳහා විස්තරාත්මක පැහැදිලි කිරීම් ද ලබා දෙන්න.
### ප්‍රශ්නය: {question}
### පිළිතුර:"""

inputs = tokenizer(instruction, return_tensors="pt").to(model.device)
outputs = model.generate(
            **inputs,
            max_new_tokens=100,
            num_beams=10,
            repetition_penalty=1.2,
            no_repeat_ngram_size=3,
            do_sample=False,
            early_stopping=True,
            eos_token_id=tokenizer.eos_token_id, 
            pad_token_id=tokenizer.pad_token_id
            )

generated_answer_tokens = outputs[0][inputs.input_ids.shape[1]:]
generated_answer = tokenizer.decode(generated_answer_tokens, skip_special_tokens=True).strip()
print(generated_answer)