A collection of Italian, Spanish, Russian and Chinese SFT samples (in LLAMA chat format) in code generation settings. These are sourced from:
frankminors123/codellama-2-20k-zh (ZH)
jean1/45k_python_code_chinese_instruction (ZH)
Den4ikAI/russian_code_qa (RU)
mii-llm/code-ita-dpo-small (IT)
Rodr16020/code_instructions_7_5k_alpaca_spanish (ES)
These are then filtered using guesslang to retain samples in the languages we seek to continue pretraining on (refer to MaLA-LM/stack-final).
