Instructions to use SauravMahalik/kalinga-02-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use SauravMahalik/kalinga-02-GGUF with Transformers:

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

pipe = pipeline("text-generation", model="SauravMahalik/kalinga-02-GGUF")

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("SauravMahalik/kalinga-02-GGUF", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use SauravMahalik/kalinga-02-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf SauravMahalik/kalinga-02-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf SauravMahalik/kalinga-02-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf SauravMahalik/kalinga-02-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf SauravMahalik/kalinga-02-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf SauravMahalik/kalinga-02-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf SauravMahalik/kalinga-02-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf SauravMahalik/kalinga-02-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf SauravMahalik/kalinga-02-GGUF:Q4_K_M

Use Docker

docker model run hf.co/SauravMahalik/kalinga-02-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use SauravMahalik/kalinga-02-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "SauravMahalik/kalinga-02-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "SauravMahalik/kalinga-02-GGUF",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/SauravMahalik/kalinga-02-GGUF:Q4_K_M

SGLang

How to use SauravMahalik/kalinga-02-GGUF 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 "SauravMahalik/kalinga-02-GGUF" \
    --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": "SauravMahalik/kalinga-02-GGUF",
		"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 "SauravMahalik/kalinga-02-GGUF" \
        --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": "SauravMahalik/kalinga-02-GGUF",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Ollama
How to use SauravMahalik/kalinga-02-GGUF with Ollama:
```
ollama run hf.co/SauravMahalik/kalinga-02-GGUF:Q4_K_M
```
Unsloth Desktop
Docker Model Runner
How to use SauravMahalik/kalinga-02-GGUF with Docker Model Runner:
```
docker model run hf.co/SauravMahalik/kalinga-02-GGUF:Q4_K_M
```

Lemonade

How to use SauravMahalik/kalinga-02-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull SauravMahalik/kalinga-02-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.kalinga-02-GGUF-Q4_K_M

List all available models

lemonade list

Atomic Chat

KALINGA - Experiment No. 02 (GGUF)

One language. Done properly. Measured honestly.

Kalinga-02 is a continued-pretrained sarvamai/sarvam-1 (2.5B) with a custom Odia-optimized tokenizer, trained on ~1.13B tokens of Odia. This repo hosts community-friendly GGUF quants.

Files

File	Quant	Size
`kalinga-02-Q4_K_M.gguf`	recommended daily use	~1.5 GB
`kalinga-02-Q5_K_M.gguf`	higher fidelity	~1.7 GB
`kalinga-02-Q8_0.gguf`	near-lossless	~2.7 GB

Converted with llama.cpp convert_hf_to_gguf.py (f16) + llama-quantize.

Run it

ollama create kalinga-02 -f Modelfile   # Modelfile: FROM ./kalinga-02-Q4_K_M.gguf
ollama run kalinga-02

llama-cli -m kalinga-02-Q4_K_M.gguf -p "ଓଡ଼ିଆରେ ଗୋଟିଏ ବାକ୍ୟ ଲେଖ" -n 128

LM Studio / Jan / GPT4All: drop the file in, pick the quant that fits your RAM.

The receipts

Same harness, both models, no cherry-picking.

Held-out perplexity, 2,000 docs, bits-per-char (tokenizer-fair):

Task	sarvam-1 (base)	kalinga-02
overall	1.96	2.91
qa	1.45	2.64
generation	2.48	2.62
translation	1.07	3.70

Multiple-choice accuracy, 64,031 rows, logprob-scored (no generation):

Model	Accuracy
sarvam-1 (base)	40.82%
kalinga-02	38.55%

What the numbers say: this training recipe degrades the base on fluency and task accuracy, and English ability collapses. This quant release exists so the community can reproduce, probe, and build on a documented result - the next experiment starts from evidence.

License / lineage

Base: sarvamai/sarvam-1 (non-commercial terms flow downstream). This is an experiment, tagged accordingly - use with eyes open.

Experiment No. 02 - Kalinga Series.

Downloads last month: 120

GGUF

Model size

2B params

Architecture

llama

Hardware compatibility

4-bit

5-bit

8-bit