huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-19 03:54:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'main' into ivarflakstad/metal-prng

Browse Source
This commit is contained in:
Ivar Flakstad
2024-01-12 07:19:58 +01:00
parent 87efb5d8eb 41915184bb
commit e63bb8661b
53 changed files with 1035 additions and 1051 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
candle-core/benches/benchmarks/random.rs
View File

@ -1,4 +1,4 @@
use crate::benchmarks::{bench_name, device, BenchDevice};
use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
use candle_core::{DType, Device, Tensor};
use criterion::{black_box, criterion_group, Criterion, Throughput};
use std::time::Instant;
@ -11,19 +11,18 @@ fn rand_normal(a: &Tensor) {
a.randn_like(100.0, 15.0).unwrap();
}
fn criterion_benchmark(c: &mut Criterion) {
fn run_random_bench(c: &mut Criterion, device: &Device) {
let b = 1;
let rows = 2048;
let cols = 2048;
let d = device().unwrap();
let dtype = DType::F32;
let tensor = Tensor::zeros((b, rows, cols), dtype, &d).unwrap();
let tensor = Tensor::zeros((b, rows, cols), dtype, device).unwrap();
let flops = b * rows * cols * dtype.size_in_bytes();
let mut group = c.benchmark_group(bench_name("random_uniform"));
let mut group = c.benchmark_group(device.bench_name("random_uniform"));
group.throughput(Throughput::Bytes(flops as u64));
group.bench_function("iter", move |benches| {
benches.iter_custom(|iters| {
@ -31,16 +30,15 @@ fn criterion_benchmark(c: &mut Criterion) {
for _i in 0..iters {
rand_uniform(black_box(&tensor));
}
d.sync().unwrap();
device.sync().unwrap();
start.elapsed()
})
});
group.finish();
let d = device().unwrap();
let tensor = Tensor::zeros((b, rows, cols), dtype, &d).unwrap();
let tensor = Tensor::zeros((b, rows, cols), dtype, device).unwrap();
let mut group = c.benchmark_group(bench_name("random_normal"));
let mut group = c.benchmark_group(device.bench_name("random_normal"));
group.throughput(Throughput::Bytes(flops as u64));
group.bench_function("iter", move |benches| {
benches.iter_custom(|iters| {
@ -48,11 +46,18 @@ fn criterion_benchmark(c: &mut Criterion) {
for _i in 0..iters {
rand_normal(black_box(&tensor));
}
d.sync().unwrap();
device.sync().unwrap();
start.elapsed()
})
});
group.finish();
}
fn criterion_benchmark(c: &mut Criterion) {
let handler = BenchDeviceHandler::new().unwrap();
for device in handler.devices {
run_random_bench(c, &device);
}
}
criterion_group!(benches, criterion_benchmark);