@sudoingX I love what you're doing, man. It takes so much time and effort to test each and every model, and most of the times it's heartbreaking or disappointing.
@sudoingX What hyper-parameters are you using? I have a 4060 Ti with 16GB VRAM. I have tried running a 9B bf16 model using Ollama. I don't understand the hype around this model, though. In my tests, it performed very poorly. Am I missing something?
@HackerTwins@ZenMagnets@TheAhmadOsman Is the model actually that good? I tried using it with ollama and didn't like it. Maybe I need to change some parameters and use llama.cpp. But I don't think its anywhere near sonnet 4
Turns out with claude code, my decades long strategy of NOT deeply learning:
- regexs
- sql
- nginx confs
- elaborate shell commands
- advanced shell scripting
- any javascript framework
- perf optimization
- webpack, cdns, bundlers
- 1000 other things
...was entirely correct.