CoolFace
Modelpublic

ModelCloud/QwQ-32B-Preview-gptqmodel-4bit-vortex-v1

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face
51likes64downloads
Model Card

<center> ๐Ÿ› Update: We have discovered a regression in the QwQ Vortext v1 model that may output extra strings such as "Edited Text" that was result of invalid calibration data auto-injected by our vortex pipeline. <br> Please use v2 of the model here: <a href="https://huggingface.co/ModelCloud/QwQ-32B-Preview-gptqmodel-4bit-vortex-v2">https://huggingface.co/ModelCloud/QwQ-32B-Preview-gptqmodel-4bit-vortex-v2</a> </center>

image/png

This model has been quantized using GPTQModel.

  • โ€”bits: 4
  • โ€”dynamic: null
  • โ€”group_size: 32
  • โ€”desc_act: true
  • โ€”static_groups: false
  • โ€”sym: true
  • โ€”lm_head: false
  • โ€”true_sequential: true
  • โ€”quant_method: "gptq"
  • โ€”checkpoint_format: "gptq"
  • โ€”meta๏ผš
  • โ€”quantizer: gptqmodel:1.2.2
  • โ€”uri: https://github.com/modelcloud/gptqmodel
  • โ€”damp_percent: 0.1
  • โ€”damp_auto_increment: 0.0015

Example:

python
from transformers import AutoTokenizer
from gptqmodel import GPTQModel

tokenizer = AutoTokenizer.from_pretrained("ModelCloud/QwQ-32B-Preview-gptqmodel-4bit-vortex-v1")
model = GPTQModel.load("ModelCloud/QwQ-32B-Preview-gptqmodel-4bit-vortex-v1")

messages = [
    {"role": "system", "content": "You are Qwen, created by Alibaba Cloud. You are a helpful assistant."},
    {"role": "user", "content": "How can I design a data structure in C++ to store the top 5 largest integer numbers?"},
]
input_tensor = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")

outputs = model.generate(input_ids=input_tensor.to(model.device), max_new_tokens=512)
result = tokenizer.decode(outputs[0][input_tensor.shape[1]:], skip_special_tokens=True)

print(result)