Instructions to use pytorch/Qwen3-4B-INT8-INT4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use pytorch/Qwen3-4B-INT8-INT4 with Transformers:

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

pipe = pipeline("text-generation", model="pytorch/Qwen3-4B-INT8-INT4")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("pytorch/Qwen3-4B-INT8-INT4")
model = AutoModelForCausalLM.from_pretrained("pytorch/Qwen3-4B-INT8-INT4")
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 pytorch/Qwen3-4B-INT8-INT4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "pytorch/Qwen3-4B-INT8-INT4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "pytorch/Qwen3-4B-INT8-INT4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/pytorch/Qwen3-4B-INT8-INT4

SGLang

How to use pytorch/Qwen3-4B-INT8-INT4 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 "pytorch/Qwen3-4B-INT8-INT4" \
    --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": "pytorch/Qwen3-4B-INT8-INT4",
		"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 "pytorch/Qwen3-4B-INT8-INT4" \
        --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": "pytorch/Qwen3-4B-INT8-INT4",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use pytorch/Qwen3-4B-INT8-INT4 with Docker Model Runner:
```
docker model run hf.co/pytorch/Qwen3-4B-INT8-INT4
```

Qwen3-4B-INT8-INT4 / config.json

metascroy

Upload 3 files

a05b1b3 verified 8 months ago

raw

history blame contribute delete

4.06 kB

	{
	"architectures": [
	"Qwen3ForCausalLM"
	],
	"attention_bias": false,
	"attention_dropout": 0.0,
	"bos_token_id": 151643,
	"dtype": "bfloat16",
	"eos_token_id": 151645,
	"head_dim": 128,
	"hidden_act": "silu",
	"hidden_size": 2560,
	"initializer_range": 0.02,
	"intermediate_size": 9728,
	"layer_types": [
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention",
	"full_attention"
	],
	"max_position_embeddings": 40960,
	"max_window_layers": 36,
	"model_type": "qwen3",
	"num_attention_heads": 32,
	"num_hidden_layers": 36,
	"num_key_value_heads": 8,
	"quantization_config": {
	"include_input_output_embeddings": true,
	"modules_to_not_convert": [],
	"quant_method": "torchao",
	"quant_type": {
	"default": {
	"_data": {
	"module_fqn_to_config": {
	"_default": {
	"_data": {
	"act_mapping_type": {
	"_data": "ASYMMETRIC",
	"_type": "MappingType"
	},
	"intx_packing_format": {
	"_data": "UNPACKED_TO_INT8",
	"_type": "IntxPackingFormat"
	},
	"layout": {
	"_data": {},
	"_type": "QDQLayout",
	"_version": 1
	},
	"weight_dtype": {
	"_data": "int4",
	"_type": "torch.dtype"
	},
	"weight_granularity": {
	"_data": {
	"group_size": 32
	},
	"_type": "PerGroup",
	"_version": 1
	},
	"weight_mapping_type": {
	"_data": "SYMMETRIC",
	"_type": "MappingType"
	},
	"weight_scale_dtype": null
	},
	"_type": "Int8DynamicActivationIntxWeightConfig",
	"_version": 2
	},
	"model.embed_tokens": {
	"_data": {
	"granularity": {
	"_data": {
	"axis": 0
	},
	"_type": "PerAxis",
	"_version": 1
	},
	"intx_packing_format": {
	"_data": "UNPACKED_TO_INT8",
	"_type": "IntxPackingFormat"
	},
	"layout": {
	"_data": {},
	"_type": "QDQLayout",
	"_version": 1
	},
	"mapping_type": {
	"_data": "SYMMETRIC",
	"_type": "MappingType"
	},
	"scale_dtype": null,
	"weight_dtype": {
	"_data": "int8",
	"_type": "torch.dtype"
	}
	},
	"_type": "IntxWeightOnlyConfig",
	"_version": 2
	}
	}
	},
	"_type": "ModuleFqnToConfig",
	"_version": 1
	}
	},
	"quant_type_kwargs": {},
	"untie_embedding_weights": false
	},
	"rms_norm_eps": 1e-06,
	"rope_scaling": null,
	"rope_theta": 1000000,
	"sliding_window": null,
	"tie_word_embeddings": false,
	"transformers_version": "4.57.0.dev0",
	"use_cache": true,
	"use_sliding_window": false,
	"vocab_size": 151936
	}