peterbuitho/VietPoet-Qwen3.5-4B-GGUF

VietPoet Qwen3.5-4B (GGUF)

A Qwen3.5-4B fine-tuned (QLoRA, 8,000 poems, 2 epochs) to write Vietnamese lục bát poems, exported for LM Studio / llama.cpp. It is meant to be used with the line-by-line sampler and rule checker from github.com/peterbuitho/ThoLucBat, whose local installation for Windows (VietPoet-win.zip) sets it up for you. Used on its own it writes the right shape but breaks the tone rules more often.

File	Size	Notes
VietPoet-Qwen3.5-4B-Q8_0.gguf	4.6 GB	near-lossless; use with 10 GB+ of graphics memory
VietPoet-Qwen3.5-4B-Q4_K_M.gguf	2.8 GB	for smaller graphics cards or CPU

What it does and does not do

With the sampler (16 samples per line, only lines that satisfy the 6/8 length, tone and rhyme rules are kept), on 40 held-out "8 câu" prompts against LM Studio: rule score 0.991 (Q8_0), 95% of poems fully valid; Q4_K_M with 4 samples per line: 0.982, 87.5% valid. Without the sampler the raw model scores about 0.81 and breaks the 6th/8th-syllable tone rule in 38% of bát lines.

These numbers measure form, not poetry. Poems are correct lục bát but the meaning is often loose or off-topic (training prompts only had the poem title as topic). Judge the poetry yourself.

Prompt format

Qwen chat format with thinking off, and the poem written by appending lines to the assistant turn:

<|im_start|>system
Bạn là nhà thơ Việt Nam chuyên sáng tác thơ lục bát.<|im_end|>
<|im_start|>user
Viết một bài thơ lục bát 8 câu về mùa thu quê em.<|im_end|>
<|im_start|>assistant
<think>

</think>

The request wording varies (several templates); they are in app/prompts.py in the GitHub repo.

Credits

• Training data: phamson02/vietnamese-poetry-corpus (CC BY 4.0), filtered to poems that pass a lục bát rule checker.
• Base model: Qwen/Qwen3.5-4B (Apache-2.0).
• Scoring idea: Vietnamese Poem Generation & the Prospect of Cross-Language Poem-to-Poem Translation (arXiv:2401.01078).