Files
zeekayandHanzo Dev 6a363f7ee3 merge: upstream main ddf1b879 and the aarch64 repack branch 35d7ae7c
Three-way per file through the rename: base 39355c6c91, theirs the upstream
merge of main with 35d7ae7c (via 7ef6a694d), ours origin/main.

Lands the barrier thread pool, runtime CPU feature dispatch, aarch64 and x86
repacked quantized kernels (indexed_gemv, gemv_fused_shared_lhs), quantized
embedding on CPU/Metal/CUDA, paged flash-attn kernels, cutile fused MoE, the
in-place op aliasing API, parking_lot storage locks, Metal buffer builders and
debug labels, safetensors 0.8, hf-hub 1.0, tokenizers 0.23.

Ours keeps ROCm, Vulkan, wgpu, the IQ/MXFP4/NVFP4/TQ dtypes, and the i-quant
weight/activation block split in matmul. Our backends ride InplaceOpN and the
quantized embedding; Metal get_rows gains the IQ kernels. ug leaves, as upstream.

Co-authored-by: Hanzo Dev <[email protected]>
2026-09-15 15:52:28 -07:00

99 lines
2.9 KiB
Rust

//! EfficientNet implementation.
//!
//! https://arxiv.org/abs/1905.11946
#[cfg(feature = "mkl")]
extern crate intel_mkl_src;
#[cfg(feature = "accelerate")]
extern crate accelerate_src;
use clap::{Parser, ValueEnum};
use hanzo_ml::{DType, IndexOp, D};
use hanzo_nn::{Module, VarBuilder};
use hanzo_transformers::models::efficientnet::{EfficientNet, MBConvConfig};
#[derive(Clone, Copy, Debug, ValueEnum)]
enum Which {
B0,
B1,
B2,
B3,
B4,
B5,
B6,
B7,
}
#[derive(Parser)]
struct Args {
#[arg(long)]
model: Option<String>,
#[arg(long)]
image: String,
/// Run on CPU rather than on GPU.
#[arg(long)]
cpu: bool,
/// Variant of the model to use.
#[arg(value_enum, long, default_value_t = Which::B2)]
which: Which,
}
pub fn main() -> anyhow::Result<()> {
let args = Args::parse();
let device = hanzo_ml_examples::device(args.cpu)?;
let image = hanzo_ml_examples::imagenet::load_image224(args.image)?.to_device(&device)?;
println!("loaded image {image:?}");
let model_file = match args.model {
None => {
let api = hanzo_ml_examples::hub::Api::new()?;
let api = api.model("lmz/hanzo-ml-efficientnet");
let filename = match args.which {
Which::B0 => "efficientnet-b0.safetensors",
Which::B1 => "efficientnet-b1.safetensors",
Which::B2 => "efficientnet-b2.safetensors",
Which::B3 => "efficientnet-b3.safetensors",
Which::B4 => "efficientnet-b4.safetensors",
Which::B5 => "efficientnet-b5.safetensors",
Which::B6 => "efficientnet-b6.safetensors",
Which::B7 => "efficientnet-b7.safetensors",
};
api.get(filename)?
}
Some(model) => model.into(),
};
let vb = unsafe { VarBuilder::from_mmaped_safetensors(&[model_file], DType::F32, &device)? };
let cfg = match args.which {
Which::B0 => MBConvConfig::b0(),
Which::B1 => MBConvConfig::b1(),
Which::B2 => MBConvConfig::b2(),
Which::B3 => MBConvConfig::b3(),
Which::B4 => MBConvConfig::b4(),
Which::B5 => MBConvConfig::b5(),
Which::B6 => MBConvConfig::b6(),
Which::B7 => MBConvConfig::b7(),
};
let model = EfficientNet::new(vb, cfg, hanzo_ml_examples::imagenet::CLASS_COUNT as usize)?;
println!("model built");
let logits = model.forward(&image.unsqueeze(0)?)?;
let prs = hanzo_nn::ops::softmax(&logits, D::Minus1)?
.i(0)?
.to_vec1::<f32>()?;
let mut prs = prs.iter().enumerate().collect::<Vec<_>>();
prs.sort_by(|(_, p1), (_, p2)| p2.total_cmp(p1));
for &(category_idx, pr) in prs.iter().take(5) {
println!(
"{:24}: {:.2}%",
hanzo_ml_examples::imagenet::CLASSES[category_idx],
100. * pr
);
}
Ok(())
}