Llamacpp imatrix Quantizations of LongWriter-Zero-32B by THU-KEG

Using llama.cpp release b5753 for quantization.

Original model: https://huggingface.co/THU-KEG/LongWriter-Zero-32B

All quants made using imatrix option with dataset from here

Run them in LM Studio

Run them directly with llama.cpp, or any other llama.cpp based project

Prompt format

[gMASK]<sop>A conversation between the user and the assistant. The user provides a writing/general task, and the assistant completes it. The assistant first deeply thinks through the writing/answering process in their mind before providing the final written work to the user. The assistant should engage in comprehensive and in-depth planning to ensure that every aspect of the writing/general task is detailed and well-structured. If there is any uncertainty or ambiguity in the writing request, the assistant should reflect, ask themselves clarifying questions, and explore multiple writing approaches to ensure the final output meets the highest quality standards. Since writing is both a creative and structured task, the assistant should analyze it from multiple perspectives, considering coherence, clarity, style, tone, audience, purpose, etc.. Additionally, the assistant should review and refine the work to enhance its expressiveness. The writing thought process and the final written work should be enclosed within <think> </think> and <answer> </answer> tags, respectively, as shown below: <think>A comprehensive strategy for writing that encompasses detailed planning and structural design—including brainstorming, outlining, style selection, audience adaptation, self-reflection, quality assurance, etc..</think> <answer>The final written work after thorough optimization and refinement.</answer> <|user|>: {system_prompt} <|assistant|>:

Download a file (not the whole branch) from below:

Embed/output weights

Some of these quants (Q3_K_XL, Q4_K_L etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to.

Downloading using huggingface-cli

ARM/AVX information

Previously, you would download Q4_0_4_4/4_8/8_8, and these would have their weights interleaved in memory in order to improve performance on ARM and AVX machines by loading up more data in one pass.

Now, however, there is something called "online repacking" for weights. details in this PR. If you use Q4_0 and your hardware would benefit from repacking weights, it will do it automatically on the fly.

As of llama.cpp build b4282 you will not be able to run the Q4_0_X_X files and will instead need to use Q4_0.

Additionally, if you want to get slightly better quality for , you can use IQ4_NL thanks to this PR which will also repack the weights for ARM, though only the 4_4 for now. The loading time may be slower but it will result in an overall speed incrase.

Which file should I choose?

Credits

Thank you kalomaze and Dampf for assistance in creating the imatrix calibration dataset.

Thank you ZeroWw for the inspiration to experiment with embed/output.

Thank you to LM Studio for sponsoring my work.

Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski

Base model

Qwen/Qwen2.5-32B

THU-KEG/LongWriter-Zero-32B

(9)