gitadmin/foxhunt
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix: cudarc 0.19 API fixes in tests — CudaDevice→CudaContext, no Arc double-wrap

Browse Source 
Test code used CudaDevice (cudarc 0.17) instead of CudaContext (0.19),
and wrapped Arc::new(CudaContext::new().new_stream()) which double-wraps
since new_stream() already returns Arc<CudaStream>.

740 tests pass across ml-core, ml-ppo, ml-supervised, ml-ensemble.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
jgrusewski
2026-03-18 08:05:32 +01:00
parent f607335820
commit d6fa11ec6c
9 changed files with 34 additions and 52 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
crates/ml-core/src/cuda_autograd/activations.rs
View File

@@ -595,12 +595,10 @@ mod tests {
use super::*;
fn make_stream() -> Arc<CudaStream> {
Arc::new(
cudarc::driver::CudaDevice::new(0)
cudarc::driver::CudaContext::new(0)
.expect("CUDA required")
.fork_default_stream()
.expect("fork stream"),
)
.new_stream()
.expect("fork stream")
}
#[test]

32
crates/ml-core/src/cuda_autograd/gpu_tensor.rs
View File

@@ -892,12 +892,10 @@ mod tests {
#[test]
fn test_zeros_shape() {
let stream = Arc::new(
cudarc::driver::CudaDevice::new(0)
let stream = cudarc::driver::CudaContext::new(0)
.expect("CUDA required")
.fork_default_stream()
.expect("fork stream"),
);
.new_stream()
.expect("fork stream");
let t = GpuTensor::zeros(&[4, 8], &stream).unwrap();
assert_eq!(t.shape(), &[4, 8]);
assert_eq!(t.numel(), 32);
@@ -905,12 +903,10 @@ mod tests {
#[test]
fn test_roundtrip() {
let stream = Arc::new(
cudarc::driver::CudaDevice::new(0)
let stream = cudarc::driver::CudaContext::new(0)
.expect("CUDA required")
.fork_default_stream()
.expect("fork stream"),
);
.new_stream()
.expect("fork stream");
let host_data: Vec<f32> = (0..12).map(|i| i as f32).collect();
let t = GpuTensor::from_host(&host_data, vec![3, 4], &stream).unwrap();
let back = t.to_host(&stream).unwrap();
@@ -919,12 +915,10 @@ mod tests {
#[test]
fn test_shape_mismatch() {
let stream = Arc::new(
cudarc::driver::CudaDevice::new(0)
let stream = cudarc::driver::CudaContext::new(0)
.expect("CUDA required")
.fork_default_stream()
.expect("fork stream"),
);
.new_stream()
.expect("fork stream");
let host_data = vec![1.0_f32; 10];
let result = GpuTensor::from_host(&host_data, vec![3, 4], &stream);
assert!(result.is_err());
@@ -932,12 +926,10 @@ mod tests {
#[test]
fn test_reshape() {
let stream = Arc::new(
cudarc::driver::CudaDevice::new(0)
let stream = cudarc::driver::CudaContext::new(0)
.expect("CUDA required")
.fork_default_stream()
.expect("fork stream"),
);
.new_stream()
.expect("fork stream");
let t = GpuTensor::zeros(&[6, 4], &stream).unwrap();
let t2 = t.reshape(vec![2, 12]).unwrap();
assert_eq!(t2.shape(), &[2, 12]);

8
crates/ml-core/src/cuda_autograd/init.rs
View File

@@ -141,12 +141,10 @@ mod tests {
use super::*;
fn make_stream() -> Arc<CudaStream> {
Arc::new(
cudarc::driver::CudaDevice::new(0)
cudarc::driver::CudaContext::new(0)
.expect("CUDA required")
.fork_default_stream()
.expect("fork stream"),
)
.new_stream()
.expect("fork stream")
}
#[test]

8
crates/ml-core/src/cuda_autograd/linear.rs
View File

@@ -362,12 +362,10 @@ mod tests {
use super::super::var_store::GpuVarStore;
fn make_stream() -> Arc<CudaStream> {
Arc::new(
cudarc::driver::CudaDevice::new(0)
cudarc::driver::CudaContext::new(0)
.expect("CUDA required")
.fork_default_stream()
.expect("fork stream"),
)
.new_stream()
.expect("fork stream")
}
fn make_cublas(stream: &Arc<CudaStream>) -> CudaBlas {

8
crates/ml-core/src/cuda_autograd/loss.rs
View File

@@ -218,12 +218,10 @@ mod tests {
use super::*;
fn make_stream() -> Arc<CudaStream> {
Arc::new(
cudarc::driver::CudaDevice::new(0)
cudarc::driver::CudaContext::new(0)
.expect("CUDA required")
.fork_default_stream()
.expect("fork stream"),
)
.new_stream()
.expect("fork stream")
}
#[test]

8
crates/ml-core/src/cuda_autograd/optimizer.rs
View File

@@ -368,12 +368,10 @@ mod tests {
use super::super::init;
fn make_stream() -> Arc<CudaStream> {
Arc::new(
cudarc::driver::CudaDevice::new(0)
cudarc::driver::CudaContext::new(0)
.expect("CUDA required")
.fork_default_stream()
.expect("fork stream"),
)
.new_stream()
.expect("fork stream")
}
#[test]

10
crates/ml-core/src/cuda_autograd/var_store.rs
View File

@@ -353,12 +353,10 @@ mod tests {
use super::*;
fn make_stream() -> Arc<CudaStream> {
Arc::new(
cudarc::driver::CudaDevice::new(0)
.expect("CUDA required")
.fork_default_stream()
.expect("fork stream"),
)
cudarc::driver::CudaContext::new(0)
.expect("CUDA required")
.new_stream()
.expect("fork stream")
}
#[test]

3
crates/ml-supervised/src/mamba/loss.rs
View File

@@ -88,7 +88,8 @@ pub fn directional_mse_loss(
#[cfg(test)]
mod tests {
use super::*;
use cudarc::driver::CudaContext;
use std::sync::Arc;
use cudarc::driver::{CudaContext, CudaStream};
fn test_stream() -> Arc<CudaStream> {
let ctx = CudaContext::new(0).expect("CUDA context required");

1
crates/ml-supervised/src/mamba/mod.rs
View File

@@ -3019,6 +3019,7 @@ impl Clone for Mamba2SSM {
mod tests {
use super::*;
use anyhow::Result;
use cudarc::driver::CudaContext;
#[tokio::test]
async fn test_mamba_creation() -> Result<()> {