Merge remote-tracking branch 'origin/main' into lukim/upgrade-candle

Author: lukekim

Cargo.toml

@@ -77,6 +77,7 @@ image = { version = "0.25.2", default-features = false, features = [
 imageproc = { version = "0.24.0", default-features = false }
 intel-mkl-src = { version = "0.8.1", features = ["mkl-static-lp64-iomp"] }
 libc = { version = "0.2.147" }
+libm = { version = "0.2.15" }
 log = "0.4"
 memmap2 = { version = "0.9.3", features = ["stable_deref_trait"] }
 num_cpus = "1.15.0"

candle-core/Cargo.toml

@@ -22,6 +22,7 @@ half = { workspace = true }
 float8 = { workspace = true }
 intel-mkl-src = { workspace = true, optional = true }
 libc = { workspace = true, optional = true }
+libm = { workspace = true }
 memmap2 = { workspace = true }
 num-traits = { workspace = true }
 num_cpus = { workspace = true }

candle-core/benches/benchmarks/affine.rs

@@ -38,8 +38,7 @@ fn criterion_benchmark(c: &mut Criterion) {
         run_affine_benchmark(c, &device, DType::F32, "affine_f32");
         run_affine_benchmark(c, &device, DType::F16, "affine_f16");
         run_affine_benchmark(c, &device, DType::BF16, "affine_bf16");
-        #[cfg(feature = "metal")]
-        continue;
+        #[cfg(not(feature = "metal"))]
         run_affine_benchmark(c, &device, DType::F8E4M3, "affine_fp8");
     }
 }

candle-core/benches/benchmarks/mod.rs

@@ -32,7 +32,7 @@ impl BenchDevice for Device {
             }
             Device::Metal(device) => {
                 #[cfg(feature = "metal")]
-                return Ok(device.wait_until_completed()?);
+                return device.wait_until_completed();
                 #[cfg(not(feature = "metal"))]
                 panic!("Metal device without metal feature enabled: {:?}", device)
             }