Instructions to use parmanu-lcs2/gemma-4-9B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use parmanu-lcs2/gemma-4-9B with Transformers:

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

pipe = pipeline("text-generation", model="parmanu-lcs2/gemma-4-9B", trust_remote_code=True)

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

processor = AutoProcessor.from_pretrained("parmanu-lcs2/gemma-4-9B", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("parmanu-lcs2/gemma-4-9B", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use parmanu-lcs2/gemma-4-9B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "parmanu-lcs2/gemma-4-9B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "parmanu-lcs2/gemma-4-9B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/parmanu-lcs2/gemma-4-9B

SGLang

How to use parmanu-lcs2/gemma-4-9B 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 "parmanu-lcs2/gemma-4-9B" \
    --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": "parmanu-lcs2/gemma-4-9B",
		"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 "parmanu-lcs2/gemma-4-9B" \
        --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": "parmanu-lcs2/gemma-4-9B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use parmanu-lcs2/gemma-4-9B with Docker Model Runner:
```
docker model run hf.co/parmanu-lcs2/gemma-4-9B
```

parmanu-lcs2/gemma-4-9B

This model is a pruned version of gemma-4-9B, compressed using TRIDENT (paper: coming soon).

Pruning details

Setting	Value
Pruning method	TRIDENT (paper: coming soon)
Target compression ratio	25%
Parameters after pruning	8,969,794,864
Training data	SlimOrca (10,000 samples)

Hardware

All pruning was performed on an NVIDIA A100 GPU.

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("parmanu-lcs2/gemma-4-12B", trust_remote_code=True, dtype="auto")
tokenizer = AutoTokenizer.from_pretrained("parmanu-lcs2/gemma-4-12B")

trust_remote_code=True is required because pruning leaves each layer's MLP with a different width. This architecture is defined in the bundled modeling_trident.py.

References

Original model: gemma-4-12B
TRIDENT paper: Coming soon

Downloads last month: -

Safetensors

Model size

9B params

Tensor type

BF16

parmanu-lcs2
/

gemma-4-9B

parmanu-lcs2/gemma-4-9B

Pruning details

Hardware

Usage

References

Dataset used to train parmanu-lcs2/gemma-4-9B