Update dependencies (#3135)

ivarflakstad · web-flow · commit 701205a345e6 · 2025-10-17T00:25:53.000+02:00
* Update candle metal dependencies

* Update benches black_box usage

* Update candle onnx dependencies

* Update tokio in examples/book
diff --git a/Cargo.toml b/Cargo.toml
@@ -42,7 +42,7 @@ candle-nn = { path = "./candle-nn", version = "0.9.1" }
 candle-onnx = { path = "./candle-onnx", version = "0.9.1" }
 candle-transformers = { path = "./candle-transformers", version = "0.9.1" }
 clap = { version = "4.2.4", features = ["derive"] }
-criterion = { version = "0.5.1", default-features = false }
+criterion = { version = "0.7.0", default-features = false }
 cudarc = { version = "0.17.3", features = [
     "std",
     "cublas",
diff --git a/candle-book/Cargo.toml b/candle-book/Cargo.toml
@@ -25,7 +25,7 @@ cudarc = { workspace = true, optional = true }
 half = { workspace = true, optional = true }
 image = { workspace = true, optional = true }
 anyhow = { workspace = true }
-tokio = "1.43.0"
+tokio = "1.48.0"
 
 [dev-dependencies]
 byteorder = { workspace = true }
diff --git a/candle-core/benches/benchmarks/affine.rs b/candle-core/benches/benchmarks/affine.rs
@@ -1,6 +1,7 @@
 use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle_core::{DType, Device, Tensor};
-use criterion::{black_box, criterion_group, Criterion, Throughput};
+use criterion::{criterion_group, Criterion, Throughput};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn run(a: &Tensor) {
diff --git a/candle-core/benches/benchmarks/conv_transpose2d.rs b/candle-core/benches/benchmarks/conv_transpose2d.rs
@@ -1,6 +1,7 @@
 use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle_core::{DType, Device, Tensor};
-use criterion::{black_box, criterion_group, Criterion, Throughput};
+use criterion::{criterion_group, Criterion, Throughput};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn run(
diff --git a/candle-core/benches/benchmarks/copy.rs b/candle-core/benches/benchmarks/copy.rs
@@ -1,6 +1,7 @@
 use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle_core::{Device, Tensor, WithDType};
-use criterion::{black_box, criterion_group, Criterion, Throughput};
+use criterion::{criterion_group, Criterion, Throughput};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn run_copy_mask_benchmark<D: WithDType>(c: &mut Criterion, device: &Device, name: &str) {
diff --git a/candle-core/benches/benchmarks/matmul.rs b/candle-core/benches/benchmarks/matmul.rs
@@ -1,6 +1,7 @@
 use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle_core::{DType, Device, Tensor};
-use criterion::{black_box, criterion_group, Criterion, Throughput};
+use criterion::{criterion_group, Criterion, Throughput};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn run(a: &Tensor, b: &Tensor) {
diff --git a/candle-core/benches/benchmarks/qmatmul.rs b/candle-core/benches/benchmarks/qmatmul.rs
@@ -3,7 +3,8 @@ use candle_core::{
     quantized::{self, GgmlDType, QMatMul},
     Device, Module, Tensor,
 };
-use criterion::{black_box, criterion_group, Criterion, Throughput};
+use criterion::{criterion_group, Criterion, Throughput};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn run(matmul: &QMatMul, x: &Tensor) {
diff --git a/candle-core/benches/benchmarks/random.rs b/candle-core/benches/benchmarks/random.rs
@@ -1,6 +1,7 @@
 use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle_core::{DType, Device, Tensor};
-use criterion::{black_box, criterion_group, Criterion, Throughput};
+use criterion::{criterion_group, Criterion, Throughput};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn rand_uniform(a: &Tensor) {
diff --git a/candle-core/benches/benchmarks/reduce.rs b/candle-core/benches/benchmarks/reduce.rs
@@ -1,7 +1,8 @@
 use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle_core::{DType, Device, Tensor};
-use criterion::{black_box, criterion_group, Criterion, Throughput};
+use criterion::{criterion_group, Criterion, Throughput};
 use half::{bf16, f16};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn run_sum(a: &Tensor) {
diff --git a/candle-core/benches/benchmarks/unary.rs b/candle-core/benches/benchmarks/unary.rs
@@ -1,6 +1,7 @@
 use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle_core::{DType, Device, Tensor};
-use criterion::{black_box, criterion_group, Criterion, Throughput};
+use criterion::{criterion_group, Criterion, Throughput};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn run(a: &Tensor) {
diff --git a/candle-core/benches/benchmarks/where_cond.rs b/candle-core/benches/benchmarks/where_cond.rs
@@ -1,6 +1,7 @@
 use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle_core::{DType, Device, Tensor};
-use criterion::{black_box, criterion_group, Criterion, Throughput};
+use criterion::{criterion_group, Criterion, Throughput};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn run(a: &Tensor, b: &Tensor, c: &Tensor) {
diff --git a/candle-examples/Cargo.toml b/candle-examples/Cargo.toml
@@ -55,7 +55,7 @@ tracing = { workspace = true }
 tracing-chrome = { workspace = true }
 tracing-subscriber = { workspace = true }
 # Necessary to disambiguate with tokio in wasm examples which are 1.28.1
-tokio = "1.43.0"
+tokio = "1.48.0"
 
 [build-dependencies]
 anyhow = { workspace = true }
diff --git a/candle-metal-kernels/Cargo.toml b/candle-metal-kernels/Cargo.toml
@@ -16,23 +16,23 @@ half = { version = "2.5.0", features = [
     "use-intrinsics",
     "rand_distr",
 ] }
-once_cell = "1.18.0"
-thiserror = "1"
-tracing = "0.1.37"
-objc2-metal = "0.3.1"
-objc2 = "0.6.1"
-objc2-foundation = "0.3.1"
+once_cell = "1.21"
+thiserror = "2"
+tracing = "0.1.41"
+objc2-metal = "0.3.2"
+objc2 = "0.6.3"
+objc2-foundation = "0.3.2"
 
 [dev-dependencies]
-clap = { version = "4.2.4", features = ["derive"] }
-half = { version = "2.3.1", features = [
+clap = { version = "4.5.49", features = ["derive"] }
+half = { version = "2.7.1", features = [
     "num-traits",
     "use-intrinsics",
     "rand_distr",
 ] }
 anyhow = "1"
-rand = "0.8.5"
-rand_distr = "0.4.3"
+rand = "0.9.2"
+rand_distr = "0.5.1"
 
 [profile.profiling]
 inherits = "release"
diff --git a/candle-metal-kernels/src/kernel.rs b/candle-metal-kernels/src/kernel.rs
@@ -115,7 +115,7 @@ impl Kernels {
                 let source_content = self.get_library_source(source);
                 let compile_options = MTLCompileOptions::new();
                 //unsafe { compile_options.setEnableLogging(true) };
-                unsafe { compile_options.setMathMode(MTLMathMode::Fast) };
+                compile_options.setMathMode(MTLMathMode::Fast);
                 device
                     .new_library_with_source(source_content, Some(&compile_options))
                     .map_err(|e| MetalKernelError::LoadLibraryError(e.to_string()))?
diff --git a/candle-metal-kernels/src/metal/command_buffer.rs b/candle-metal-kernels/src/metal/command_buffer.rs
@@ -58,7 +58,7 @@ impl CommandBuffer {
     }
 
     pub fn wait_until_completed(&self) {
-        unsafe { self.raw.waitUntilCompleted() }
+        self.raw.waitUntilCompleted()
     }
 }
 
diff --git a/candle-metal-kernels/src/tests.rs b/candle-metal-kernels/src/tests.rs
@@ -3,8 +3,7 @@ use crate::metal::create_command_buffer;
 use core::ffi::c_void;
 use half::{bf16, f16};
 use rand::prelude::SliceRandom;
-use rand::thread_rng;
-use rand::Rng;
+use rand::{rng, Rng};
 
 fn read_to_vec<T: Clone>(buffer: &Buffer, n: usize) -> Vec<T> {
     let ptr = buffer.contents() as *const T;
@@ -999,7 +998,7 @@ fn reduce_sum_case<const N: usize, const D: usize>() {
     let mut v = create_array::<N>();
     if D == 1 {
         // Hardens 1-dimensional test cases
-        v.shuffle(&mut thread_rng());
+        v.shuffle(&mut rng());
     }
     let results = run_reduce(&v, N, D, "fast_sum_f32");
     assert_eq!(approx(results, 4), correct_sum::<N, D>());
@@ -1009,7 +1008,7 @@ fn reduce_max_case<const N: usize, const D: usize>() {
     let mut v = create_array::<N>();
     if D == 1 {
         // Hardens 1-dimensional test cases
-        v.shuffle(&mut thread_rng());
+        v.shuffle(&mut rng());
     }
     let results = run_reduce(&v, N, D, "fast_max_f32");
     assert_eq!(approx(results, 4), correct_max::<N, D>());
@@ -1019,7 +1018,7 @@ fn reduce_argmax_case<const N: usize, const D: usize>() {
     let mut v = create_array::<N>();
     if D == 1 {
         // Hardens 1-dimensional test cases
-        v.shuffle(&mut thread_rng());
+        v.shuffle(&mut rng());
     }
     let results: Vec<u32> = run_reduce(&v, N, D, "fast_argmax_f32");
     assert_eq!(results, correct_argmax::<N, D>(v));
@@ -1233,7 +1232,7 @@ fn run_where_cond<I: Clone, T: Clone>(
         buffer: &right,
         offset_in_bytes: cond_offset,
     };
-    call_where_cond_strided(
+    call_where_cond(
         &device,
         &command_buffer,
         &kernels,
@@ -2388,8 +2387,8 @@ fn const_fill() {
         name: &'static str,
         f: F,
     ) {
-        let len = rand::thread_rng().gen_range(2..16) * rand::thread_rng().gen_range(4..16);
-        let value = rand::thread_rng().gen_range(1. ..19.);
+        let len = rand::rng().random_range(2..16) * rand::rng().random_range(4..16);
+        let value = rand::rng().random_range(1. ..19.);
         let value = f(value);
         let v = constant_fill::<T>(name, len, value);
         assert_eq!(v, vec![value; len])
diff --git a/candle-nn/benches/benchmarks/conv.rs b/candle-nn/benches/benchmarks/conv.rs
@@ -1,7 +1,8 @@
 use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle::{DType, Device, Module, Tensor};
 use candle_nn::{Conv2d, Conv2dConfig};
-use criterion::{black_box, criterion_group, Criterion};
+use criterion::{criterion_group, Criterion};
+use std::hint::black_box;
 use std::time::Instant;
 
 const B: usize = 1;
diff --git a/candle-nn/benches/benchmarks/layer_norm.rs b/candle-nn/benches/benchmarks/layer_norm.rs
@@ -1,7 +1,8 @@
 use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle::{DType, Device, Module, Tensor};
 use candle_nn::LayerNorm;
-use criterion::{black_box, criterion_group, Criterion};
+use criterion::{criterion_group, Criterion};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn run(input: &Tensor, weight: &Tensor, bias: &Tensor) {
diff --git a/candle-nn/benches/benchmarks/softmax.rs b/candle-nn/benches/benchmarks/softmax.rs
@@ -2,7 +2,8 @@ use crate::benchmarks::{BenchDevice, BenchDeviceHandler};
 use candle::{DType, Device, Tensor};
 use candle_nn::ops::softmax_last_dim;
 use criterion::Throughput;
-use criterion::{black_box, criterion_group, Criterion};
+use criterion::{criterion_group, Criterion};
+use std::hint::black_box;
 use std::time::Instant;
 
 fn run(input: &Tensor) {
diff --git a/candle-onnx/Cargo.toml b/candle-onnx/Cargo.toml
@@ -12,11 +12,11 @@ license = "MIT OR Apache-2.0"
 [dependencies]
 candle = { path = "../candle-core", package = "candle-core", version = "0.9.1" }
 candle-nn = { path = "../candle-nn", version = "0.9.1" }
-prost = "0.12.1"
+prost = "0.14.1"
 
 [build-dependencies]
-prost-build = "0.12.1"
+prost-build = "0.14.1"
 
 [dev-dependencies]
 anyhow = { version = "1", features = ["backtrace"] }
-clap = { version = "4.2.4", features = ["derive"] }
+clap = { version = "4.5.49", features = ["derive"] }
diff --git a/candle-onnx/src/eval.rs b/candle-onnx/src/eval.rs
@@ -2320,7 +2320,7 @@ fn simple_eval_(
 
                 let indices_shape = indices.dims();
                 let data_shape = data.dims();
-                let updates_shape = updates.dims();
+                let _updates_shape = updates.dims();
 
                 // Last dimension of indices represents the depth of indexing
                 let k = indices_shape.last().unwrap().clone();
diff --git a/candle-onnx/tests/ops.rs b/candle-onnx/tests/ops.rs
@@ -6442,6 +6442,7 @@ fn test_selu_operator() -> Result<()> {
     Ok(())
 }
 
+#[test]
 fn test_hard_swish() -> candle::Result<()> {
     {
         let manual_graph = create_model_proto_with_graph(Some(GraphProto {
