libmir-cuda 0.3.0

CUDA inference backend for libmir
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130

use models::weights::{
    ExpertProjectionLayout, LayerTensorRole, LogicalTensorRole, TensorStorage, WeightBindingPlan,
};

use crate::{Error, Result};

#[derive(Clone, Copy, Debug, Eq, PartialEq)]
pub(super) enum ClampedRoutedLayout {
    Native,
    Mlx,
    Dense,
}

impl ClampedRoutedLayout {
    pub(super) fn discover(bindings: &WeightBindingPlan) -> Result<Self> {
        if bindings.tensors.iter().any(|binding| {
            matches!(
                binding.role,
                LogicalTensorRole::Layer {
                    tensor: LayerTensorRole::ExpertProjection { .. },
                    ..
                }
            ) && matches!(binding.storage, TensorStorage::Dense { .. })
        }) {
            return Ok(Self::Dense);
        }
        match bindings.expert_projection_layout() {
            Some(ExpertProjectionLayout::InterleavedGateUp) => Ok(Self::Native),
            Some(ExpertProjectionLayout::SeparateGateUp) => Ok(Self::Mlx),
            None => Err(Error::UnsupportedDecoderLayer(
                "binding plan has no complete clamped-routed expert projection layout".into(),
            )),
        }
    }

    pub(super) const fn storage(self) -> &'static str {
        match self {
            Self::Native => "interleaved MXFP4 gate/up blocks",
            Self::Mlx => "separate MLX affine MXFP4 gate/up blocks",
            Self::Dense => "dense BF16 selected-expert matrices",
        }
    }
}

#[cfg(test)]
mod tests {
    use models::weights::{
        AffineBits, AffineGroupAxis, AffinePacking, AffineParameterDType, AffineSignedness,
        AffineStorageDType, AffineZeroPointMode, BindingTransform, BlockQuantization,
        ExpertProjectionRole, GroupedAffineQuantization, LayerTensorRole, LogicalTensorRole,
        TensorBinding, TensorPacking, TensorStorage, WeightBindingPlan,
    };

    use super::*;

    #[test]
    fn distinguishes_native_and_mlx_checkpoints() -> Result<()> {
        let native = plan(vec![binding(
            ExpertProjectionRole::GateUp,
            TensorStorage::BlockQuantized {
                format: BlockQuantization::MXFP4,
                scales: "scales".into(),
                global_scale: None,
                input_scale: None,
                bias: None,
                packing: TensorPacking::InterleavedGateUp,
            },
        )]);
        assert_eq!(ClampedRoutedLayout::discover(&native)?, ClampedRoutedLayout::Native);

        let affine = || TensorStorage::AffineQuantized {
            format: GroupedAffineQuantization {
                bits: AffineBits::Four,
                group_size: 32,
                group_axis: AffineGroupAxis::Input,
                signedness: AffineSignedness::Unsigned,
                zero_point: AffineZeroPointMode::None,
                packing: AffinePacking::Mlx,
                storage_dtype: AffineStorageDType::U32,
                scale_dtype: AffineParameterDType::F16,
                bias_dtype: None,
            },
            scales: "scales".into(),
            biases: None,
            output_bias: None,
        };
        let mlx = plan(vec![
            binding(ExpertProjectionRole::Gate, affine()),
            binding(ExpertProjectionRole::Up, affine()),
        ]);
        assert_eq!(ClampedRoutedLayout::discover(&mlx)?, ClampedRoutedLayout::Mlx);

        let dense = plan(vec![binding(
            ExpertProjectionRole::GateUp,
            TensorStorage::Dense { dtype: "BF16".into(), bias: None },
        )]);
        assert_eq!(ClampedRoutedLayout::discover(&dense)?, ClampedRoutedLayout::Dense);
        Ok(())
    }

    #[test]
    fn rejects_a_plan_without_an_expert_transform() {
        assert!(ClampedRoutedLayout::discover(&plan(Vec::new())).is_err());
    }

    fn plan(tensors: Vec<TensorBinding>) -> WeightBindingPlan {
        WeightBindingPlan { tensors }
    }

    fn binding(projection: ExpertProjectionRole, storage: TensorStorage) -> TensorBinding {
        TensorBinding {
            role: LogicalTensorRole::Layer {
                index: 0,
                tensor: LayerTensorRole::ExpertProjection { expert: None, projection },
            },
            source: String::new(),
            shape: Vec::new(),
            logical_shape: None,
            transforms: match projection {
                ExpertProjectionRole::GateUp => {
                    vec![BindingTransform::FusedGateUp { interleaved: true }]
                },
                ExpertProjectionRole::Gate
                | ExpertProjectionRole::Up
                | ExpertProjectionRole::Down => Vec::new(),
            },
            storage,
        }
    }
}