Show HN: I finetuned 1.5B Qwen to near GPT-4o level bash generation perf
Purely a hobby side project to see how far I can push a really small model, using (mostly) automated training pipelines<p>Original mention: <a href="https://news.ycombinator.com/item?id=49869735">https://news.ycombinator.com/item?id=49869735</a><p>There were a bunch of requests to release it. Full synthetic data: <a href="https://huggingface.co/datasets/dirac-run/ec-training-data" rel="nofollow">https://huggingface.co/datasets/dirac-run/ec-training-data</a><p>Models: <a href="https://huggingface.co/dirac-run/ec-1.5b-gguf" rel="nofollow">https://huggingface.co/dirac-run/ec-1.5b-gguf</a> and <a href="https://huggingface.co/dirac-run/ec-0.6b-gguf" rel="nofollow">https://huggingface.co/dirac-run/ec-0.6b-gguf</a><p>Cli <a href="https://github.com/dirac-run/ec" rel="nofollow">https://github.com/dirac-run/ec</a><p>feel free to train/use the data as you wish.
競合・代替製品
GPT4o.so
免费使用 GPT-4o 模型，不用注册和登录。（[@sunnyjingliu](https://github.com/ruanyf/weekly/issues/4617) 投稿）
Show HN: A decent text-to-image model trained for ~$20K (Apache 2.0 Weights)
My brother and I run a 2-person lab that trains our own generative video models from scratch. The ultimate goal is a se…
Show HN: Interactive Diesel Price Forecaster
Interactive model of diesel price I've been working on with my (sightspinner) agent. Takes into account the main (known…
Show HN: HackerRepos – HackerTube for repos mentioned on HN
Karpathy posted hackertube.org 5 days ago, and mentioned that it took him 2 hours to build. A day after another fellow…
Show HN: Pick your dream LLM H/W setup and simulate its throughput
Also, you can share your simulation with query parameters, ex) <a href="https://cro.sh/specmaxx/?m=deepseek-ai/DeepSeek…