Show HN: I finetuned 1.5B Qwen to near GPT-4o level bash generation perf

A developer released ec-1.5b-gguf and ec-0.6b-gguf, two finetuned Qwen models, along with the full synthetic training dataset at huggingface.co/datasets/dirac-run/ec-training-data and a CLI at github.com/dirac-run/ec, claiming the 1.5B model reaches near GPT-4o-level bash generation performance. The project was built as a hobby side project using mostly automated training pipelines, and the developer says the data may be freely used for training.

Purely a hobby side project to see how far I can push a really small model, using mostly automated training pipelines Original mention: https://news.ycombinator.com/item?id=49869735 https://news.ycombinator.com/item?id=49869735 There were a bunch of requests to release it. Full synthetic data: https://huggingface.co/datasets/dirac-run/ec-training-data https://huggingface.co/datasets/dirac-run/ec-training-data Models: https://huggingface.co/dirac-run/ec-1.5b-gguf https://huggingface.co/dirac-run/ec-1.5b-gguf and https://huggingface.co/dirac-run/ec-0.6b-gguf https://huggingface.co/dirac-run/ec-0.6b-gguf Cli https://github.com/dirac-run/ec https://github.com/dirac-run/ec feel free to train/use the data as you wish. Comments URL: https://news.ycombinator.com/item?id=49966238 https://news.ycombinator.com/item?id=49966238 Points: 1 Comments: 0