Instructions to use RefalMachine/DeepSeek-V3-0324-Channel-INT8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use RefalMachine/DeepSeek-V3-0324-Channel-INT8 with Transformers:

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

pipe = pipeline("text-generation", model="RefalMachine/DeepSeek-V3-0324-Channel-INT8", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("RefalMachine/DeepSeek-V3-0324-Channel-INT8", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("RefalMachine/DeepSeek-V3-0324-Channel-INT8", trust_remote_code=True)
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 Settings

vLLM

How to use RefalMachine/DeepSeek-V3-0324-Channel-INT8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "RefalMachine/DeepSeek-V3-0324-Channel-INT8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "RefalMachine/DeepSeek-V3-0324-Channel-INT8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/RefalMachine/DeepSeek-V3-0324-Channel-INT8

SGLang

How to use RefalMachine/DeepSeek-V3-0324-Channel-INT8 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 "RefalMachine/DeepSeek-V3-0324-Channel-INT8" \
    --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": "RefalMachine/DeepSeek-V3-0324-Channel-INT8",
		"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 "RefalMachine/DeepSeek-V3-0324-Channel-INT8" \
        --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": "RefalMachine/DeepSeek-V3-0324-Channel-INT8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use RefalMachine/DeepSeek-V3-0324-Channel-INT8 with Docker Model Runner:
```
docker model run hf.co/RefalMachine/DeepSeek-V3-0324-Channel-INT8
```

RefalMachine commited on Apr 12, 2025

Commit

f32412f

verified ·

1 Parent(s): 658c8fb

Update README.md

Browse files

Files changed (1) hide show

README.md +18 -3

README.md CHANGED Viewed

@@ -10,11 +10,26 @@ base_model:
 - deepseek-ai/DeepSeek-V3-0324
 ---
-This is the BF16 model of DeekSeek V3-0324. Useful for quantization and inference on GPUs that do not support FP8 (Nvidia Ampere)
-BF16 is result of dequantizing the FP8 quantized weights from DeepSeek AI: https://huggingface.co/deepseek-ai/DeepSeek-V3-0324
-[GPTQModel](https://github.com/modelcloud/gptqmodel) is your go-to choice for DeepSeek V3-0324 quantization toolkit for inference on vLLM and SGLang
 # DeepSeek-V3-0324

 - deepseek-ai/DeepSeek-V3-0324
 ---
+# Channel-wise INT8 DeepSeek-V3-0324
+The INT8 quant for SGLang (https://github.com/sgl-project/sglang)
+[PULL REQUEST](https://github.com/sgl-project/sglang/pull/3888)
+## 1. Quantization Process
+We apply INT8 quantization to the BF16 checkpoints.
+The quantization scales are determined by dividing the channnel-wise maximum of element values by the INT8 type maximum.
+To generate this weight, run the provided script in the ``./inference`` directory:
+``
+python3 bf16_cast_channel_int8.py --input-bf16-hf-path /path/to/bf16-weights/ --output-int8-hf-path /path/to/save-int8-weight/
+``
+## 2. Trouble Shooting
+Before inference, you should confirm that there is no attribute "quantization_config" in `config.json`.
+---
 # DeepSeek-V3-0324