candle-mi 0.1.13

Mechanistic interpretability for language models in Rust, built on candle
Documentation
{
  "model_repo": "casperhansen/llama-3.2-1b-instruct-awq",
  "methodology": "AWQ forward-pass oracle (transformers + AutoAWQ, fp16 compute on CUDA); validates candle-mi's anamnesis AWQ dequant load path",
  "torch_version": "2.10.0+cu130",
  "transformers_version": "5.1.0",
  "platform": "Windows-11-10.0.26200-SP0",
  "model_type": "llama",
  "quant_method": "QuantizationMethod.AWQ",
  "hidden_size": 2048,
  "num_layers": 16,
  "vocab_size": 128256,
  "head_dim": 64,
  "test_cases": [
    {
      "prompt": "The capital of France is",
      "tokens": [
        128000,
        791,
        6864,
        315,
        9822,
        374
      ],
      "top_10": [
        {
          "index": 12366,
          "logit": 18.71875
        },
        {
          "index": 539,
          "logit": 15.5625
        },
        {
          "index": 264,
          "logit": 15.3515625
        },
        {
          "index": 3604,
          "logit": 15.1015625
        },
        {
          "index": 1101,
          "logit": 14.8203125
        },
        {
          "index": 1131,
          "logit": 14.515625
        },
        {
          "index": 279,
          "logit": 14.5
        },
        {
          "index": 20437,
          "logit": 14.3203125
        },
        {
          "index": 2663,
          "logit": 14.2109375
        },
        {
          "index": 3967,
          "logit": 13.9296875
        }
      ]
    },
    {
      "prompt": "Two plus two equals",
      "tokens": [
        128000,
        11874,
        5636,
        1403,
        17239
      ],
      "top_10": [
        {
          "index": 3116,
          "logit": 20.375
        },
        {
          "index": 4330,
          "logit": 18.59375
        },
        {
          "index": 4400,
          "logit": 16.75
        },
        {
          "index": 1131,
          "logit": 16.6875
        },
        {
          "index": 198,
          "logit": 16.671875
        },
        {
          "index": 8223,
          "logit": 16.609375
        },
        {
          "index": 220,
          "logit": 16.59375
        },
        {
          "index": 11,
          "logit": 16.328125
        },
        {
          "index": 4848,
          "logit": 16.25
        },
        {
          "index": 8254,
          "logit": 16.09375
        }
      ]
    },
    {
      "prompt": "Once upon a time, there was a",
      "tokens": [
        128000,
        12805,
        5304,
        264,
        892,
        11,
        1070,
        574,
        264
      ],
      "top_10": [
        {
          "index": 3995,
          "logit": 19.390625
        },
        {
          "index": 2678,
          "logit": 18.390625
        },
        {
          "index": 6366,
          "logit": 18.046875
        },
        {
          "index": 2697,
          "logit": 17.890625
        },
        {
          "index": 24632,
          "logit": 16.984375
        },
        {
          "index": 13987,
          "logit": 16.71875
        },
        {
          "index": 1912,
          "logit": 16.5625
        },
        {
          "index": 893,
          "logit": 16.546875
        },
        {
          "index": 3828,
          "logit": 16.421875
        },
        {
          "index": 41792,
          "logit": 16.375
        }
      ]
    }
  ]
}