|
--- |
|
license: other |
|
license_name: deepseek-license |
|
license_link: LICENSE |
|
--- |
|
DeepSeek-Coder-V2-Lite-Base finetuned for 0.25 epochs on adamo1139/ise-uiuc_Magicoder-Evol-Instruct-110K-ShareGPT via llama-factory at 3000ctx with qlora, rank 32 and alpha 32. |
|
|
|
Prompt format is ChatML but ChatML-specific tokens are not in the tokenizer, so it's sometimes spilling random tokens. Definitely something to fix in the next version. |
|
|
|
It's an early WIP, unless you are dying to try DeepSeek-Coder-V2-Lite finetunes I suggest you don't use it :) |
