ibrahimkettaneh/QwQ-32B-Preview-abliterated-4.5bpw-h8-exl2

I kindly ask that you follow me for more quantizations, please, because it lets me know you are interested in more work like this 😁

Credits goes to all those who have contributed to the community, in this case, specifically: huihui-ai/QwQ-32B-Preview-abliterated

Benchmark

Source: 🐺🐦‍⬛ LLM Comparison/Test: 25 SOTA LLMs (including QwQ) through 59 MMLU-Pro CS benchmark runs

Credits go to for their helpful and informative benchmark: Wolfram Ravenwolf

Recommendation for Best Performance

To increase performance, increase the max new output when running inference from the default to 16384 tokens.

Detailed Table

Duration	Total	%	TIGER-Lab	Correct Random Guesses	Prompt tokens	tk/s	Completion tokens	tk/s
QwQ-32B-Preview (8.0bpw EXL2, max_tokens=16384)	bartowski/QwQ-32B-Preview-exl2_8_0	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	38436MiB	1/2
QwQ-32B-Preview (8.0bpw EXL2, max_tokens=16384)	bartowski/QwQ-32B-Preview-exl2_8_0	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	38436MiB	2/2
QwQ-32B-Preview (4.25bpw EXL2, max_tokens=16384)	bartowski/QwQ-32B-Preview-exl2_4_25	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	27636MiB	1/2
QwQ-32B-Preview (4.25bpw EXL2, max_tokens=16384)	bartowski/QwQ-32B-Preview-exl2_4_25	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	27636MiB	2/2
QwQ-32B-Preview (8.0bpw EXL2)	bartowski/QwQ-32B-Preview-exl2_8_0	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	38528MiB	1/4
QwQ-32B-Preview (8.0bpw EXL2)	bartowski/QwQ-32B-Preview-exl2_8_0	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	38528MiB	2/4
QwQ-32B-Preview (8.0bpw EXL2)	bartowski/QwQ-32B-Preview-exl2_8_0	-	32B	EXL2	TabbyAPI	RTX 6000	37000MiB	3/4
QwQ-32B-Preview (8.0bpw EXL2)	bartowski/QwQ-32B-Preview-exl2_8_0	-	32B	EXL2	TabbyAPI	RTX 6000	37000MiB	4/4
QwQ-32B-Preview-abliterated (4.5bpw EXL2, max_tokens=16384)	ibrahimkettaneh_QwQ-32B-Preview-abliterated-4.5bpw-h8-exl2	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	28556MiB	1/2
QwQ-32B-Preview-abliterated (4.5bpw EXL2, max_tokens=16384)	ibrahimkettaneh_QwQ-32B-Preview-abliterated-4.5bpw-h8-exl2	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	28556MiB	2/2
QwQ-32B-Preview (4.25bpw EXL2)	bartowski/QwQ-32B-Preview-exl2_4_25	-	32B	EXL2	TabbyAPI	RTX 6000	26198MiB	1/4
QwQ-32B-Preview (4.25bpw EXL2)	bartowski/QwQ-32B-Preview-exl2_4_25	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	27750MiB	2/4
QwQ-32B-Preview (4.25bpw EXL2)	bartowski/QwQ-32B-Preview-exl2_4_25	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	27750MiB	3/4
QwQ-32B-Preview (4.25bpw EXL2)	bartowski/QwQ-32B-Preview-exl2_4_25	-	32B	EXL2	TabbyAPI	RTX 6000	26198MiB	4/4
QwQ-32B-Preview (3.0bpw EXL2, max_tokens=8192)	bartowski/QwQ-32B-Preview-exl2_3_0	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	22990MiB	1/2
QwQ-32B-Preview (3.0bpw EXL2, max_tokens=8192)	bartowski/QwQ-32B-Preview-exl2_3_0	Qwen/Qwen2.5-Coder-0.5B-Instruct	32B	EXL2	TabbyAPI	RTX 6000	22990MiB	2/2
QwQ-32B-Preview (3.0bpw EXL2)	bartowski/QwQ-32B-Preview-exl2_3_0	-	32B	EXL2	TabbyAPI	RTX 6000	21574MiB	1/2
QwQ-32B-Preview (3.0bpw EXL2)	bartowski/QwQ-32B-Preview-exl2_3_0	-	32B	EXL2	TabbyAPI	RTX 6000	21574MiB	2/2

For more context, details, and comparisons, you can refer to the original article by Ravenwolf.

Context

This is an uncensored version of Qwen/QwQ-32B-Preview created with abliteration (see remove-refusals-with-transformers to know more about it).

This is a crude, proof-of-concept implementation to remove refusals from an LLM model without using TransformerLens.

QwQ-32B-Preview

Introduction

QwQ-32B-Preview is an experimental research model developed by the Qwen Team, focused on advancing AI reasoning capabilities. As a preview release, it demonstrates promising analytical abilities while having several important limitations:

1. Language Mixing and Code-Switching: The model may mix languages or switch between them unexpectedly, affecting response clarity.
2. Recursive Reasoning Loops: The model may enter circular reasoning patterns, leading to lengthy responses without a conclusive answer.
3. Safety and Ethical Considerations: The model requires enhanced safety measures to ensure reliable and secure performance, and users should exercise caution when deploying it.
4. Performance and Benchmark Limitations: The model excels in math and coding but has room for improvement in other areas, such as common sense reasoning and nuanced language understanding.

Specification:

• Type: Causal Language Models
• Training Stage: Pretraining & Post-training
• Architecture: transformers with RoPE, SwiGLU, RMSNorm, and Attention QKV bias
• Number of Parameters: 32.5B
• Number of Paramaters (Non-Embedding): 31.0B
• Number of Layers: 64
• Number of Attention Heads (GQA): 40 for Q and 8 for KV
• Context Length: Full 32,768 tokens

For more details, please refer to our blog. You can also check Qwen2.5 GitHub, and Documentation.

Requirements

The code of Qwen2.5 has been in the latest Hugging face transformers and we advise you to use the latest version of transformers.

With transformers<4.37.0, you will encounter the following error:

KeyError: 'qwen2'

Quickstart

Here provides a code snippet with apply_chat_template to show you how to load the tokenizer and model and how to generate contents.

from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "Qwen/QwQ-32B-Preview"
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)
prompt = "How many r in strawberry."
messages = [
    {"role": "system", "content": "You are a helpful and harmless assistant. You are Qwen developed by Alibaba. You should think step-by-step."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]

Citation

If you find our work helpful, feel free to give us a cite.

@misc{qwq-32b-preview,
    title = {QwQ: Reflect Deeply on the Boundaries of the Unknown},
    url = {https://qwenlm.github.io/blog/qwq-32b-preview/},
    author = {Qwen Team},
    month = {November},
    year = {2024}
}
@article{qwen2,
      title={Qwen2 Technical Report}, 
      author={An Yang and Baosong Yang and Binyuan Hui and Bo Zheng and Bowen Yu and Chang Zhou and Chengpeng Li and Chengyuan Li and Dayiheng Liu and Fei Huang and Guanting Dong and Haoran Wei and Huan Lin and Jialong Tang and Jialin Wang and Jian Yang and Jianhong Tu and Jianwei Zhang and Jianxin Ma and Jin Xu and Jingren Zhou and Jinze Bai and Jinzheng He and Junyang Lin and Kai Dang and Keming Lu and Keqin Chen and Kexin Yang and Mei Li and Mingfeng Xue and Na Ni and Pei Zhang and Peng Wang and Ru Peng and Rui Men and Ruize Gao and Runji Lin and Shijie Wang and Shuai Bai and Sinan Tan and Tianhang Zhu and Tianhao Li and Tianyu Liu and Wenbin Ge and Xiaodong Deng and Xiaohuan Zhou and Xingzhang Ren and Xinyu Zhang and Xipin Wei and Xuancheng Ren and Yang Fan and Yang Yao and Yichang Zhang and Yu Wan and Yunfei Chu and Yuqiong Liu and Zeyu Cui and Zhenru Zhang and Zhihao Fan},
      journal={arXiv preprint arXiv:2407.10671},
      year={2024}
}