{
  "format": "custom_pretraining_gpt_safetensors",
  "created_at_utc": "2026-10-05T00:31:06.308156+00:00",
  "source_checkpoint": "/workspace/pretraining/outputs/checkpoints/sft-5b-v3/latest.pt",
  "source_config": "/workspace/pretraining/configs/decoder_sft_5b.json",
  "repo_id": "ChapAF/intent-0.1-0.5b",
  "tokenizer_name": "meta-llama/Llama-2-7b-chat-hf",
  "tokenizer_class": "LlamaTokenizer",
  "tokenizer_is_fast": true,
  "parameter_count": 455933952,
  "tensor_count": 98,
  "thinking_default": false,
  "thinking_opt_in": "prepend <think> and request reasoning explicitly",
  "checkpoint_scalars": {
    "optimizer_sharded": true,
    "step": 15852,
    "seen_tokens": 6363521024,
    "tokens_per_step": 401408
  },
  "gguf": {
    "file": "model-f16.gguf",
    "format": "GGUF",
    "architecture": "arcee",
    "output_type": "f16",
    "llama_cpp_compatible": true,
    "exact_architecture": false,
    "approximation": "Q/K RMSNorm is omitted because the upstream Arcee graph does not support it"
  }
}