candle-mi 0.1.13

Mechanistic interpretability for language models in Rust, built on candle
Documentation
{
  "model_repo": "shuyuej/Llama-3.2-1B-Instruct-GPTQ",
  "methodology": "GPTQ forward-pass oracle (transformers + gptqmodel, fp16 compute on CUDA); validates candle-mi's anamnesis GPTQ 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.GPTQ",
  "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": 264,
          "logit": 15.4375
        },
        {
          "index": 38269,
          "logit": 15.1015625
        },
        {
          "index": 1131,
          "logit": 14.9453125
        },
        {
          "index": 279,
          "logit": 14.8125
        },
        {
          "index": 311,
          "logit": 14.3515625
        },
        {
          "index": 20437,
          "logit": 14.3125
        },
        {
          "index": 539,
          "logit": 14.2890625
        },
        {
          "index": 11,
          "logit": 14.0234375
        },
        {
          "index": 25,
          "logit": 13.8515625
        }
      ]
    },
    {
      "prompt": "Two plus two equals",
      "tokens": [
        128000,
        11874,
        5636,
        1403,
        17239
      ],
      "top_10": [
        {
          "index": 3116,
          "logit": 20.265625
        },
        {
          "index": 4330,
          "logit": 19.578125
        },
        {
          "index": 8254,
          "logit": 18.171875
        },
        {
          "index": 1131,
          "logit": 17.46875
        },
        {
          "index": 2380,
          "logit": 17.328125
        },
        {
          "index": 5899,
          "logit": 17.203125
        },
        {
          "index": 8223,
          "logit": 17.15625
        },
        {
          "index": 4848,
          "logit": 17.078125
        },
        {
          "index": 832,
          "logit": 16.703125
        },
        {
          "index": 1403,
          "logit": 16.5
        }
      ]
    },
    {
      "prompt": "Once upon a time, there was a",
      "tokens": [
        128000,
        12805,
        5304,
        264,
        892,
        11,
        1070,
        574,
        264
      ],
      "top_10": [
        {
          "index": 3995,
          "logit": 18.390625
        },
        {
          "index": 2678,
          "logit": 17.109375
        },
        {
          "index": 6366,
          "logit": 16.8125
        },
        {
          "index": 893,
          "logit": 16.5
        },
        {
          "index": 5333,
          "logit": 16.109375
        },
        {
          "index": 2697,
          "logit": 15.9609375
        },
        {
          "index": 3828,
          "logit": 15.8359375
        },
        {
          "index": 1912,
          "logit": 15.2890625
        },
        {
          "index": 24632,
          "logit": 15.1953125
        },
        {
          "index": 8334,
          "logit": 15.109375
        }
      ]
    }
  ]
}