training_metadata.json

{
  "base_model": "unsloth/DeepSeek-R1-Distill-Llama-8B",
  "training_dataset": "nvidia/OpenCodeReasoning",
  "training_method": "LoRA",
  "lora_rank": 16,
  "lora_alpha": 16,
  "difficulty_filter": [
    "VERY_HARD",
    10,
    11
  ],
  "max_seq_length": 2048,
  "training_date": "2025-07-12",
  "framework": "unsloth",
  "compatible_for_further_training": true
}