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

mod execution;
mod scratch;

use mircuda::Stream;

use self::scratch::PrefillBlockScratch;
use super::{
    DecodeMoeBlockConfig, NvFp4ExpertBank,
    experts::{ExpertWeights, Experts},
    validate,
};
use crate::{
    Bf16Linear, Bf16LinearPair, CudaBackend, Error, ExecutionPhase, PrefillAttentionBf16, Result,
    RmsNormBf16, RouterBf16,
    kernels::{ElementwiseBf16, PackedGatedBf16, RouterSpec},
};

/// Fixed-chunk routed-MoE prefill sharing K/V state with decode.
#[derive(Debug)]
pub struct PrefillMoeBlockBf16 {
    attention: PrefillAttentionBf16,
    post_attention_norm: RmsNormBf16,
    pre_dense_norm: RmsNormBf16,
    dense_gate_up: Bf16LinearPair,
    dense_down: Bf16Linear,
    post_dense_norm: RmsNormBf16,
    router: RouterBf16,
    pre_expert_norm: RmsNormBf16,
    experts: Experts,
    post_expert_norm: RmsNormBf16,
    post_feed_forward_norm: RmsNormBf16,
    hidden_ops: ElementwiseBf16,
    dense_activation: PackedGatedBf16,
    scratch: PrefillBlockScratch,
    stream: Stream,
    config: DecodeMoeBlockConfig,
    tokens: usize,
}

impl CudaBackend {
    #[allow(clippy::too_many_arguments)]
    pub fn prepare_prefill_moe_block_bf16(
        &self,
        config: DecodeMoeBlockConfig,
        gate: NvFp4ExpertBank,
        up: NvFp4ExpertBank,
        down: NvFp4ExpertBank,
        tokens: usize,
    ) -> Result<PrefillMoeBlockBf16> {
        PrefillMoeBlockBf16::new(
            self,
            config,
            &ExpertWeights::NvFp4 {
                gate,
                up,
                down,
                activation_mode: models::weights::BlockActivationMode::WeightAndActivation,
            },
            tokens,
        )
    }
}

impl PrefillMoeBlockBf16 {
    pub(super) fn new(
        backend: &CudaBackend,
        config: DecodeMoeBlockConfig,
        expert_weights: &ExpertWeights,
        tokens: usize,
    ) -> Result<Self> {
        validate(config)?;
        if tokens == 0 {
            return Err(Error::InvalidDecoderKernel("prefill MoE block batch is empty"));
        }
        let hidden = config.attention.hidden_size;
        let dense = config.dense_intermediate;
        let epsilon = config.attention.rms_norm_epsilon;
        let hidden_elements = elements(tokens, hidden)?;
        let norm = || RmsNormBf16::new(backend, tokens, hidden, epsilon);
        Ok(Self {
            attention: backend.prepare_prefill_attention_bf16(config.attention, tokens)?,
            post_attention_norm: norm()?,
            pre_dense_norm: norm()?,
            dense_gate_up: Bf16LinearPair::new(
                backend,
                ExecutionPhase::Prefill,
                tokens,
                hidden,
                dense,
            )?,
            dense_down: Bf16Linear::new(backend, tokens, dense, hidden)?,
            post_dense_norm: norm()?,
            router: backend.prepare_router_batch_bf16(
                RouterSpec {
                    hidden,
                    experts: config.experts,
                    top_k: config.top_k,
                    epsilon,
                    norm_multiplier: config.router_norm_multiplier,
                },
                tokens,
            )?,
            pre_expert_norm: norm()?,
            experts: Experts::new(
                backend,
                ExecutionPhase::Prefill,
                tokens,
                config.top_k,
                config.activation,
                expert_weights,
            )?,
            post_expert_norm: norm()?,
            post_feed_forward_norm: norm()?,
            hidden_ops: ElementwiseBf16::compile(&backend.inner.compiler, hidden_elements)?,
            dense_activation: PackedGatedBf16::compile(&backend.inner.compiler, tokens, dense)?,
            scratch: PrefillBlockScratch::new(backend, tokens, hidden, dense)?,
            stream: backend.inner.stream.clone(),
            config,
            tokens,
        })
    }
}

fn elements(tokens: usize, width: usize) -> Result<usize> {
    tokens
        .checked_mul(width)
        .ok_or(Error::InvalidDecoderKernel("prefill MoE block size overflow"))
}