aprender-serve 0.51.0

Pure Rust ML inference engine built from scratch - model serving for GGUF and safetensors
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19

//! Single-token forward pass with KV cache
//!
//! Contains forward_single_with_cache and forward_single_with_cache_adaptive.
//! These are the decode-phase entry points for autoregressive generation.

use crate::brick::BrickProfiler;
use crate::error::Result;
use crate::gguf::ops;
#[cfg(feature = "gpu")]
use crate::gguf::DispatchMetrics;
use crate::gguf::{
    InferenceScratchBuffer, OwnedQuantizedKVCache, OwnedQuantizedLayer, OwnedQuantizedModel,
    GGUF_TYPE_Q4_K, GGUF_TYPE_Q5_K, GGUF_TYPE_Q6_K,
};

include!("results.rs");
include!("forward_single_profiled.rs");
// PMAT-809: Gemma-v1 arch dispatch helpers (rms_norm_arch, gemma_gate_activation).
include!("gemma_dispatch.rs");