Instructions to use junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8 with Transformers:

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

pipe = pipeline("text-generation", model="junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8")
model = AutoModelForCausalLM.from_pretrained("junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8

SGLang

How to use junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8 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 "junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8" \
    --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": "junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8",
		"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 "junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8" \
        --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": "junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8 with Docker Model Runner:
```
docker model run hf.co/junzzhu/atomllama-33K-5x5-DigitMesh-sparse-q8
```

junzzhu commited on Jan 15

Commit

a4b61a0

verified ·

1 Parent(s): b0c0856

Update README.md

Browse files

Files changed (1) hide show

README.md +4 -4

README.md CHANGED Viewed

@@ -23,17 +23,17 @@ An INT8 quantized version of [atomllama-33K-5x5-DigitMesh-sparse](https://huggin
 ## Model Description
-This is a **50% sparse + INT8 quantized** variant of the AtomLlama-33K-5x5-DigitMesh model, combining structured sparsity with Quantization Aware Training (QAT). This dual compression approach maintains digit recognition accuracy while significantly reducing model size and computational requirements.
 ### Key Features
 - **Base Model**: [junzzhu/atomllama-33K-5x5-DigitMesh-sparse](https://huggingface.co/junzzhu/atomllama-33K-5x5-DigitMesh-sparse)
 - **Sparsity**: ~50% (unstructured)
-- **Quantization**: INT8 with Quantization Aware Training (QAT)
-- **Parameters**: ~33K total, ~16.5K non-zero, 8-bit precision
 - **Architecture**: LlamaForCausalLM
 - **Task**: 5×5 binary digit mesh recognition
-- **Compression**: ~3x smaller than original model (50% sparsity + 4x from INT8)
 ## Usage

 ## Model Description
+This is a **50% sparse + INT8 quantized** variant of the AtomLlama-33K-5x5-DigitMesh model, combining structured [sparsity with Quantization Aware Training (QAT)](https://github.com/junzzhu/axolotl/blob/main/src/axolotl/integrations/sparse_qat/). This dual compression approach maintains digit recognition accuracy while significantly reducing model size and computational requirements.
 ### Key Features
 - **Base Model**: [junzzhu/atomllama-33K-5x5-DigitMesh-sparse](https://huggingface.co/junzzhu/atomllama-33K-5x5-DigitMesh-sparse)
 - **Sparsity**: ~50% (unstructured)
+- **Quantization**: INT8 with [Sparse QAT](https://github.com/junzzhu/axolotl/blob/main/src/axolotl/integrations/sparse_qat/)
+- **Parameters**: ~33K total, ~16.5K non-zero
 - **Architecture**: LlamaForCausalLM
 - **Task**: 5×5 binary digit mesh recognition
+- **Compression**: ~3x smaller than original model (46KB vs. 137KB)
 ## Usage