I'm trying to train my own llm for specifically and only coding with complex java algorithms. I have already tried qwen2.5 7b, but that was obviously too much. Are there any good model recommendations for my case? The dataset is 7500 rows and I'm training using unsloth.
I'd also be using low context length (1024-2048)