{
  "mic": "0.3",
  "name": "MiniCPM3 causal LM",
  "usage": "AutoTokenizer + AutoModelForCausalLM, batched right padding, partial training on first-layer MLP gate and final norm",
  "architecture": "MiniCPM3ForCausalLM",
  "checkpoints": {
    "openbmb/MiniCPM3-4B": {
      "fixture": "huggingface-mic/MiniCPM3-4B-mic"
    }
  },
  "fixtures": {
    "huggingface-mic/MiniCPM3-4B-mic": {
      "built_from": "openbmb/MiniCPM3-4B",
      "covers": [
        "openbmb/MiniCPM3-4B"
      ],
      "build": {
        "seed": 17,
        "model_class": "MiniCPM3ForCausalLM"
      }
    }
  },
  "integrations": {
    "transformers": {
      "entry_point": "integrations/transformers/usage.md",
      "format": "runnable-markdown",
      "requires": [
        "transformers>=5",
        "torch"
      ],
      "targets": [
        "fixtures",
        "checkpoints"
      ],
      "variants": {
        "default": {},
        "eager": {
          "attn_implementation": "eager"
        },
        "bfloat16": {
          "dtype": "bfloat16"
        }
      },
      "training": true,
      "resources": {
        "fixture": {
          "memory_gb": 1,
          "download_mb": 10
        },
        "openbmb/MiniCPM3-4B": {
          "devices": [
            "cuda"
          ],
          "gpu_memory_gb": 12,
          "memory_gb": 16,
          "download_mb": 8500
        }
      }
    }
  },
  "outputs": {
    "runtime": {
      "compare": "exact"
    },
    "logits": {
      "compare": "allclose",
      "atol": 0.0001,
      "rtol": 0.0001
    },
    "training.loss": {
      "compare": "allclose",
      "atol": 0.0001,
      "rtol": 0.0001
    },
    "training.gradients": {
      "compare": "all_true"
    },
    "training.updated": {
      "compare": "true"
    }
  }
}