You can train using these datasets with LLaMA-Factory if you add this to your data/datasets.json files.
"example_dataset": {
"hf_hub_url": "SkillFactory/SFT_DATA-cd3args-ablation-Qwen2.5-1.5B-Instruct-no_reflections",
"formatting": "sharegpt",
"columns": {
"messages": "conversations"},
"tags": {
"user_tag": "user",
"assistant_tag": "assistant",
"role_tag": "role",
"content_tag": "content"
},
"subset": "sft_train"
}
