CoolFace
Modelpublic

edge-inference/DSR1-14B-llmc-awq-w4

sourceHugging Facebsd-3-clauseupdated 1y agoView on Hugging Face
0likes13downloads
Model Card

DSR1-14B-llmc-awq-w4

AWQ quantized version of DeepSeek-R1-Distill-Qwen-14B using llm-compressor.

Model Details

  • —Base Model: DeepSeek-R1-Distill-Qwen-14B
  • —Quantization: AWQ W4A16 (4-bit weights, 16-bit activations)
  • —Group Size: 128
  • —Framework: llm-compressor
  • —Memory: 9.3GB (vs 28GB original)

Usage

vLLM

python
from vllm import LLM

model = LLM("edge-inference/DSR1-14B-llmc-awq-w4")
output = model.generate("Hello, how are you?")
print(output[0].outputs[0].text)

Transformers

python
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("edge-inference/DSR1-14B-llmc-awq-w4")
tokenizer = AutoTokenizer.from_pretrained("edge-inference/DSR1-14B-llmc-awq-w4")

Performance

  • —Memory: 67% reduction (9.3GB vs 28GB)
  • —Speed: Faster inference due to reduced memory bandwidth
  • —Quality: Minimal degradation with AWQ quantization

License

Same as base model (DeepSeek License)