Instructions to use Sashvat/HQQ-270M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Sashvat/HQQ-270M with Transformers:

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

pipe = pipeline("text-generation", model="Sashvat/HQQ-270M")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Sashvat/HQQ-270M")
model = AutoModelForCausalLM.from_pretrained("Sashvat/HQQ-270M")
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

vLLM

How to use Sashvat/HQQ-270M with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Sashvat/HQQ-270M"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sashvat/HQQ-270M",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Sashvat/HQQ-270M

SGLang

How to use Sashvat/HQQ-270M 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 "Sashvat/HQQ-270M" \
    --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": "Sashvat/HQQ-270M",
		"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 "Sashvat/HQQ-270M" \
        --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": "Sashvat/HQQ-270M",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Sashvat/HQQ-270M with Docker Model Runner:
```
docker model run hf.co/Sashvat/HQQ-270M
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

🚀 Introducing : HQQ-270M

Overview :-

HQQ-270M model is developed by Sashvat AI by finetuning Gemma-3 It specializes in transforming complex, multi-layered user queries into optimized, high-quality Google search queries .

⚠️ Usage Requirement : All input queries must begin with the prefix HQQ: ( short for High Quality Query ) . This ensures the model knows the input is intended for query optimization .

🔍 What does it do?

Converts Deep research prompts into precise search queries.
Handles Broad or ambiguous questions by breaking them into focused, search-ready chunks.
Enhances information retrieval by optimizing queries for search engines.

This model is ideal for :

Researchers
Students
Analysts
Anyone needing faster + higher-quality search results .

✨ Key Features :

Fine-tuned from Gemma-3 → retains strong language reasoning .
Fast & Efficient → Gemma's architecture is designed to make the model fast & efficient .
Optimized for real-world queries → search queries are short, relevant, and actionable.
Prefix-activated (HQQ:) → ensures model is used for its intended purpose.

📦 How to Use

🔧 Installation

pip install transformers accelerate huggingface_hub

🖥️ Inference

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Sashvat/HQQ-270M")
model = AutoModelForCausalLM.from_pretrained("Sashvat/HQQ-270M")

system_prompt = """
Convert text after "HQQ: " into an optimized Google search query. Extract key terms, remove filler words, focus on searchable keywords.
"""

query = "HQQ: What are the economic, political, and environmental implications of large-scale adoption of nuclear fusion by 2050?"

messages = [
    {"role": "system", "content": system_prompt },
    {"role": "user", "content": query }
]

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]:]))

🚀 Example Output :-

Input:

HQQ: What are the economic, political, and environmental implications of large-scale adoption of nuclear fusion by 2050?

Output :

"economic political environmental implications" "large-scale adoption nuclear fusion" 2050

Training Loss vs Steps :-

Note : 500 Steps / ~6 Epochs

📊 Intended Use :-

This model is intended for :

Query optimization for Google search and other search engines
Information retrieval pipelines .
Assisting deep research tasks .

⚠️ Important : Input must always begin with HQQ: . Without this prefix , results may be unpredictable .

📜 License

This model is released under the Sashvat AI License v1.0.
You may freely use, modify, and distribute this model, including for commercial purposes .
However, any use must clearly state :

"Made by Sashvat AI"

📄 Full license: LICENSE

🏢 About Sashvat AI

We are Sashvat AI, from BHARAT 🕉️ . building next-generation productivity tools, AI agents, and research accelerators .

📌 Citation

If you use this model, please cite :

@misc{2025-SashvatAI-HQQ-270M,
  title   = {HQQ-270M},
  author  = {SashvatAI},
  year    = {2025},
  url     = {https://huggingface.co/Sashvat/HQQ-270M}
}

Downloads last month: 26

Safetensors

Model size

0.3B params

Tensor type

BF16

Model tree for Sashvat/HQQ-270M

Base model

google/gemma-3-270m

Finetuned

google/gemma-3-270m-it

Finetuned

(1097)

this model

Merges

3 models

Quantizations

1 model