Instructions to use inference-optimization/TinyQwen3-KV-Cache-Only with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use inference-optimization/TinyQwen3-KV-Cache-Only with Transformers:

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

pipe = pipeline("text-generation", model="inference-optimization/TinyQwen3-KV-Cache-Only")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

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

tokenizer = AutoTokenizer.from_pretrained("inference-optimization/TinyQwen3-KV-Cache-Only")
model = AutoModelForCausalLM.from_pretrained("inference-optimization/TinyQwen3-KV-Cache-Only", device_map="auto")
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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use inference-optimization/TinyQwen3-KV-Cache-Only with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "inference-optimization/TinyQwen3-KV-Cache-Only"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "inference-optimization/TinyQwen3-KV-Cache-Only",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/inference-optimization/TinyQwen3-KV-Cache-Only

SGLang

How to use inference-optimization/TinyQwen3-KV-Cache-Only 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 "inference-optimization/TinyQwen3-KV-Cache-Only" \
    --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": "inference-optimization/TinyQwen3-KV-Cache-Only",
		"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 "inference-optimization/TinyQwen3-KV-Cache-Only" \
        --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": "inference-optimization/TinyQwen3-KV-Cache-Only",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use inference-optimization/TinyQwen3-KV-Cache-Only with Docker Model Runner:
```
docker model run hf.co/inference-optimization/TinyQwen3-KV-Cache-Only
```

TinyQwen3-KV-Cache-Only / README.md

kylesayrs

Add tiny Qwen3 model with only KV cache quantization (8-bit float, tensor, static, symmetric)

4a11240 verified 9 days ago

preview code

Raw History Blame Contribute Delete

1.52 kB

metadata

base_model:
  - nm-testing/tinysmokeqwen3
library_name: transformers
tags:
  - compressed-tensors
  - kv-cache
  - quantization

TinyQwen3 KV-Cache-Only quantized

A tiny Qwen3 model (nm-testing/tinysmokeqwen3) quantized with llm-compressor using only KV cache quantization — the model weights are left unquantized (dense).

Quantization config

kv_cache_scheme: 8-bit float (fp8), tensor strategy, static, symmetric
config_groups: {} (no weight quantization)
quantization_status: compressed

Produced with oneshot and the recipe:

quant_stage:
    quant_modifiers:
        QuantizationModifier:
            kv_cache_scheme:
                num_bits: 8
                type: float
                strategy: tensor
                dynamic: False
                symmetric: True

Calibrated on 16 samples of open_platypus with max_seq_length=512.

Use

Intended for testing compressed-tensors KV-cache-only checkpoints in transformers CI (see tests/quantization/compressed_tensors_integration/test_compressed_tensors.py).

from transformers import AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("inference-optimization/TinyQwen3-KV-Cache-Only")