Instructions to use nanzhang/QuantLRM-R1-Llama-70B-3-bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nanzhang/QuantLRM-R1-Llama-70B-3-bit with Transformers:

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

pipe = pipeline("text-generation", model="nanzhang/QuantLRM-R1-Llama-70B-3-bit")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("nanzhang/QuantLRM-R1-Llama-70B-3-bit")
model = AutoModelForCausalLM.from_pretrained("nanzhang/QuantLRM-R1-Llama-70B-3-bit", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use nanzhang/QuantLRM-R1-Llama-70B-3-bit with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nanzhang/QuantLRM-R1-Llama-70B-3-bit"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nanzhang/QuantLRM-R1-Llama-70B-3-bit",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/nanzhang/QuantLRM-R1-Llama-70B-3-bit

SGLang

How to use nanzhang/QuantLRM-R1-Llama-70B-3-bit 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 "nanzhang/QuantLRM-R1-Llama-70B-3-bit" \
    --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": "nanzhang/QuantLRM-R1-Llama-70B-3-bit",
		"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 "nanzhang/QuantLRM-R1-Llama-70B-3-bit" \
        --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": "nanzhang/QuantLRM-R1-Llama-70B-3-bit",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use nanzhang/QuantLRM-R1-Llama-70B-3-bit with Docker Model Runner:
```
docker model run hf.co/nanzhang/QuantLRM-R1-Llama-70B-3-bit
```

Add pipeline tag, library name, and base_model metadata

by nielsr HF Staff - opened Feb 10

base: refs/heads/main

←

from: refs/pr/1

Discussion Files changed

-5

Files changed (1) hide show

README.md +4 -5

README.md CHANGED Viewed

@@ -4,7 +4,11 @@ tags:
 - 3-bit
 - Quantization
 - Pseudo-Quantization
 ---
 # QuantLRM-R1-Llama-70B-3-bit
 3-bit quantized `DeepSeek-R1-Distill-Llama-70B` based on [QuantLRM](https://www.arxiv.org/abs/2602.02581), a state-of-the-art quantization method of large reasoning models via fine-tuning signals
@@ -15,15 +19,12 @@ This is the pseudo-quantized model (weights are dequantized back to full-precisi
 ### Model Description
-<!-- Provide a longer summary of what this model is. -->
 - **Developed by:** Nan Zhang (njz5124@psu.edu)
 - **Model type:** 3-bit pseudo-quantized version of `DeepSeek-R1-Distill-Llama-70B`
 ### Model Sources
-<!-- Provide the basic links for the model. -->
 - **Repository:** https://github.com/psunlpgroup/QuantLRM
 - **Paper:** https://www.arxiv.org/abs/2602.02581
@@ -31,7 +32,6 @@ This is the pseudo-quantized model (weights are dequantized back to full-precisi
 ## Uses
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
 This model is designed to be used with `vLLM` due to its inference optimization. Please use the tokenizer of `deepseek-ai/DeepSeek-R1-Distill-Llama-70B`.
@@ -49,7 +49,6 @@ This model achieves 2.12% improvement (based on average scores of various reason
 ## Citation
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
 **BibTeX:**

 - 3-bit
 - Quantization
 - Pseudo-Quantization
+pipeline_tag: text-generation
+library_name: transformers
+base_model: deepseek-ai/DeepSeek-R1-Distill-Llama-70B
 ---
 # QuantLRM-R1-Llama-70B-3-bit
 3-bit quantized `DeepSeek-R1-Distill-Llama-70B` based on [QuantLRM](https://www.arxiv.org/abs/2602.02581), a state-of-the-art quantization method of large reasoning models via fine-tuning signals
 ### Model Description
 - **Developed by:** Nan Zhang (njz5124@psu.edu)
 - **Model type:** 3-bit pseudo-quantized version of `DeepSeek-R1-Distill-Llama-70B`
 ### Model Sources
 - **Repository:** https://github.com/psunlpgroup/QuantLRM
 - **Paper:** https://www.arxiv.org/abs/2602.02581
 ## Uses
 This model is designed to be used with `vLLM` due to its inference optimization. Please use the tokenizer of `deepseek-ai/DeepSeek-R1-Distill-Llama-70B`.
 ## Citation
 **BibTeX:**