GoKawiilDeveloper fine-tunes Qwen3-1.7B to generate text without the letter 'e'
A developer fine-tuned the open-source Qwen3-1.7B language model on a single 16GB consumer GPU so it could respond to everyday prompts while avoiding the letter 'e' entirely, a style known as a lipogram. The project, documented on GitHub, shows before-and-after examples including a poem about trees that successfully omits 'e' after training. The creator noted the task was difficult because language models process text in tokens rather than letters, making it hard for them to track which characters appear in a word.
GoKawiil's interpretation of the reporting above, not reported fact.
The project illustrates a known limitation of large language models: their tokenized view of text makes character-level constraints, like counting letters or avoiding one, surprisingly hard to satisfy even for small, fine-tuned models. It also shows that hobbyist-scale fine-tuning on consumer hardware can still coax specialized, constrained behaviors out of an open-weight model without large-scale infrastructure. Such experiments could inform broader interest in how fine-tuning might be used to enforce stylistic or structural rules in generated text.
- A developer fine-tuned Qwen3-1.7B on a 16GB GPU to avoid using the letter 'e' in its responses.
- The task highlights that LLMs process text as tokens, not letters, making character-level constraints difficult.
- The project, with code shared on GitHub, demonstrates consumer-hardware fine-tuning for niche stylistic goals.
NVIDIA GeForce RTX 4060 Ti 16GB — If you're fine-tuning LLMs like Qwen3 locally, having a 16GB GPU is key to running these experiments without hitting memory walls. The RTX 4060 Ti 16GB offers that VRAM headroom for local fine-tuning projects at home, letting hobbyists replicate work like this lipogram experiment on their own hardware.
See NVIDIA GeForce RTX 4060 Ti 16GB on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: kevin-wu.ca, 2026-10-25
Published there as: “Fine-tuned Qwen on my 16gb GPU to respond without the letter e”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.