Closed reneleonhardt closed 4 months ago
FIM seems to be possible with this model. However, you need to specify the correct FIM template in order for this to work.
It supports code completion, FIM tasks, and understands nuanced programming queries, enabling developers to work more efficiently and effectively with less effort.
https://stability.ai/news/introducing-stable-code-instruct-3b
I reused the CodeQwen infill template (like the prompt template), it seems to work.
In addition I sorted the LLaMA models in settings because it seems fairer than appending all new models to the end where no one can see them... it would be cool to persist the plugin version between updates/reinstall, then we could mark all new ones (according to each user personally). Similar to the changelog, now IntelliJ showed only What's New 2.7.1, but I was coming from 2.6.3, so I can't see what changed in 2.7.0... strange "changelog" only showing one change/release 🤷
it would be cool to persist the plugin version between updates/reinstall, then we could mark all new ones (according to each user personally).
Not sure I follow. What do you mean by persisting the plugin version between updates?
Similar to the changelog, now IntelliJ showed only What's New 2.7.1, but I was coming from 2.6.3, so I can't see what changed in 2.7.0... strange "changelog" only showing one change/release 🤷
Yeah, it probably just shows the latest/recent changes. I had to patch a fix really quickly due to some backward compatibility issues.
https://stability.ai/news/introducing-stable-code-instruct-3b https://huggingface.co/stabilityai/stable-code-instruct-3b https://huggingface.co/bartowski/stable-code-instruct-3b-GGUF https://ollama.com/library/stable-code
Prompt seems identical to Code Qwen, but I had to add
<|im_end|>
as stop token. Can you try it out, FIM isn't possible with the instruct model, right?Too bad we can't run 2 models after each other like I proposed...
For example https://huggingface.co/google/codegemma-7b https://huggingface.co/google/codegemma-7b-it https://huggingface.co/meta-llama/Meta-Llama-3-8B https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct