Instructions to use dhrubasumatary/sarvam-1-quantized with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dhrubasumatary/sarvam-1-quantized with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="dhrubasumatary/sarvam-1-quantized",
	filename="FP16.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use dhrubasumatary/sarvam-1-quantized with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf dhrubasumatary/sarvam-1-quantized
# Run inference directly in the terminal:
llama-cli -hf dhrubasumatary/sarvam-1-quantized

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf dhrubasumatary/sarvam-1-quantized
# Run inference directly in the terminal:
llama-cli -hf dhrubasumatary/sarvam-1-quantized

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 dhrubasumatary/sarvam-1-quantized
# Run inference directly in the terminal:
./llama-cli -hf dhrubasumatary/sarvam-1-quantized

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 dhrubasumatary/sarvam-1-quantized
# Run inference directly in the terminal:
./build/bin/llama-cli -hf dhrubasumatary/sarvam-1-quantized

Use Docker

docker model run hf.co/dhrubasumatary/sarvam-1-quantized

LM Studio
Jan

vLLM

How to use dhrubasumatary/sarvam-1-quantized with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dhrubasumatary/sarvam-1-quantized"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dhrubasumatary/sarvam-1-quantized",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/dhrubasumatary/sarvam-1-quantized

Ollama
How to use dhrubasumatary/sarvam-1-quantized with Ollama:
```
ollama run hf.co/dhrubasumatary/sarvam-1-quantized
```

Unsloth Studio new

How to use dhrubasumatary/sarvam-1-quantized with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for dhrubasumatary/sarvam-1-quantized to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for dhrubasumatary/sarvam-1-quantized to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for dhrubasumatary/sarvam-1-quantized to start chatting

Docker Model Runner
How to use dhrubasumatary/sarvam-1-quantized with Docker Model Runner:
```
docker model run hf.co/dhrubasumatary/sarvam-1-quantized
```

Lemonade

How to use dhrubasumatary/sarvam-1-quantized with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull dhrubasumatary/sarvam-1-quantized

Run and chat with the model

lemonade run user.sarvam-1-quantized-{{QUANT_TAG}}

List all available models

lemonade list

Sarvam-1 Quantized Model

Model Description

The Sarvam-1 quantized model is a reduced-size version of the original Sarvam-1 model, specifically optimized for efficient inference on local machines using Ollama or similar tools. This quantization preserves the model's capabilities while significantly lowering the computational requirements, making it accessible for wider use.

This model is particularly effective for generating text in 10 Indic languages (bn, gu, hi, kn, ml, mr, or, pa, ta, te) and maintains competitive performance compared to larger models like Llama-3.1-8B.

Key Features

Quantization for Efficiency: This model has been quantized to reduce its memory footprint and enhance inference speed, making it suitable for local deployment.
Support for Multiple Indian Languages: Optimized for generating text in major Indian languages alongside English.
High-Quality Training Data: Trained on a large, curated dataset with a focus on Indic languages, ensuring high-quality outputs.

Model Architecture

Original Model Size: 2 billion parameters
Quantized Model Size: [Specify the size if different]
Key Features: Retains core architecture characteristics of Sarvam-1, including token efficiency and inference capabilities.

Performance

While specific quantitative performance metrics for the quantized model are not provided, it is expected to exhibit similar capabilities to the original Sarvam-1 model, particularly in handling text generation tasks in Indian languages.

Usage

To utilize the quantized model, follow the instructions below:

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer
model_name = "your_username/sarvam-1-quantized"  # Replace with your model's path
model = AutoModelForCausalLM.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)

# Example usage
text = "कर्नाटक की राजधानी है:"
inputs = tokenizer(text, return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=5)
result = tokenizer.decode(outputs[0])
print(result)

Downloads last month: 8

GGUF

Model size

3B params

Architecture

llama

Hardware compatibility

We're not able to determine the quantization variants.

View all variants

Model tree for dhrubasumatary/sarvam-1-quantized

Base model

sarvamai/sarvam-1

Quantized

(17)

this model