PortPilot
Benchmark · pre-generated, always available← back to leaderboard
Migration Readiness Report

NVIDIA-AI-IOT/torch2trt

https://github.com/NVIDIA-AI-IOT/torch2trt ↗
Ready to run on ROCm

NVIDIA-AI-IOT/torch2trt scored 93/100 for ROCm readiness (Ready to run on ROCm). Of 295 CUDA-related findings, 274 work as-is on ROCm PyTorch, 2 need a mechanical HIPify-style change, and 19 are manual blockers. Focus first on the manual blockers — they gate the migration.

274Works as-is
2Mechanical change
19Manual blocker
Findings
295
Python files
70
Files scanned
140
Custom CUDA kernels
detected

Detected but out of scope (not analyzed): C++, C/C++ header, JavaScript

Findings by file

295 findings · 38 files
examples/contrib/quantization_aware_training/infer.py· 5
C
TensorRT dependency
import tensorrt as trt
:13

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# examples/contrib/quantization_aware_training/infer.py:13
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
A
torch.cuda API usage
args.cuda = not args.no_cuda and torch.cuda.is_available()
:19

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
torch.cuda API usage
torch.cuda.manual_seed(args.seed)
:24

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
.cuda() tensor/module move
model = model.cuda().eval()
:44

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
rand_in = torch.randn([128,3,32,32],dtype=torch.float32).cuda()
:55

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

examples/contrib/quantization_aware_training/train.py· 8
C
TensorRT dependency
import tensorrt as trt
:15

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# examples/contrib/quantization_aware_training/train.py:15
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
A
torch.cuda API usage
args.cuda = not args.no_cuda and torch.cuda.is_available()
:37

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
torch.cuda API usage
torch.cuda.manual_seed(args.seed)
:42

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
.cuda() tensor/module move
model = model.cuda()
:66

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = inputs.cuda()
:107

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
labels = labels.cuda()
:108

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
model=model.cuda().eval()
:151

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
rand_in = torch.randn([128,3,32,32],dtype=torch.float32).cuda()
:156

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

examples/easyocr/optimize_detector.py· 3
C
TensorRT dependency
import tensorrt as trt
:5

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# examples/easyocr/optimize_detector.py:5
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
A
torch.cuda API usage
torch.cuda.current_stream().synchronize()
:56

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
torch.cuda API usage
torch.cuda.current_stream().synchronize()
:60

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

examples/easyocr/optimize_recognizer.py· 3
C
TensorRT dependency
import tensorrt as trt
:5

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# examples/easyocr/optimize_recognizer.py:5
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
A
torch.cuda API usage
torch.cuda.current_stream().synchronize()
:82

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
torch.cuda API usage
torch.cuda.current_stream().synchronize()
:86

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

requirements/requirements_10.txt· 1
C
TensorRT dependency
tensorrt==10.0.1
:1

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# requirements/requirements_10.txt:1
- tensorrt==10.0.1
+ tensorrt==10.0.1  # advisory: see ROCm note above
requirements/requirements_8.txt· 1
C
TensorRT dependency
tensorrt==8.6.1
:1

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# requirements/requirements_8.txt:1
- tensorrt==8.6.1
+ tensorrt==8.6.1  # advisory: see ROCm note above
setup.py· 1
C
TensorRT dependency
import tensorrt
:2

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# setup.py:2
- import tensorrt
+ import tensorrt  # advisory: see ROCm note above
tests/feature_tests/test_dataset_calibrator.py· 13
C
TensorRT dependency
import tensorrt as trt
:2

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# tests/feature_tests/test_dataset_calibrator.py:2
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
A
.cuda() tensor/module move
self.conv = nn.Conv2d(3, 6, kernel_size=3, stride=1, padding=1).cuda().eval()
:20

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda(),
:28

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda()
:29

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:32

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda(),
:49

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda()
:50

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
self.conv = nn.Conv2d(3, 6, kernel_size=3, stride=1, padding=1).cuda().eval()
:66

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda(),
:74

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda()
:75

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:78

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda(),
:95

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda()
:96

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_dynamic_shape.py· 11
C
TensorRT dependency
import tensorrt as trt
:4

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# tests/feature_tests/test_dynamic_shape.py:4
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
A
.cuda() tensor/module move
module = nn.Conv2d(3, 6, kernel_size=3, stride=1, padding=1).cuda().eval()
:13

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
dataset.insert((torch.randn(1, 3, 224, 224).cuda(),))
:16

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
dataset.insert((torch.randn(1, 3, 64, 64).cuda(),))
:17

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
dataset.insert((torch.randn(1, 3, 128, 128).cuda(),))
:18

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
dataset.insert((torch.randn(4, 3, 32, 32).cuda(),))
:19

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = nn.Conv2d(3, 6, kernel_size=3, stride=1, padding=1).cuda().eval()
:41

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
dataset.insert((torch.randn(1, 3, 224, 224).cuda(),))
:44

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
dataset.insert((torch.randn(1, 3, 64, 64).cuda(),))
:45

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
dataset.insert((torch.randn(1, 3, 128, 128).cuda(),))
:46

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
dataset.insert((torch.randn(4, 3, 32, 32).cuda(),))
:47

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

torch2trt/__init__.py· 1
C
TensorRT dependency
import tensorrt as trt
:3

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# torch2trt/__init__.py:3
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
torch2trt/contrib/qat/converters/QuantConv.py· 2
C
TensorRT dependency
import tensorrt as trt
:3

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# torch2trt/contrib/qat/converters/QuantConv.py:3
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
A
Device string "cuda"
@add_module_test(torch.float32, torch.device('cuda'), [(1, 10, 224, 224)], enabled=trt_version() >= '7.0')
:62

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

torch2trt/contrib/qat/converters/QuantConvBN.py· 2
C
TensorRT dependency
import tensorrt as trt
:3

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# torch2trt/contrib/qat/converters/QuantConvBN.py:3
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
A
Device string "cuda"
@add_module_test(torch.float32, torch.device('cuda'), [(1, 10, 224, 224)], enabled=trt_version() >= '7.0')
:62

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

torch2trt/contrib/qat/converters/QuantRelu.py· 1
C
TensorRT dependency
import tensorrt as trt
:2

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# torch2trt/contrib/qat/converters/QuantRelu.py:2
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
torch2trt/dataset_calibrator.py· 1
C
TensorRT dependency
import tensorrt as trt
:2

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# torch2trt/dataset_calibrator.py:2
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
torch2trt/misc_utils.py· 3
C
TensorRT dependency
import tensorrt as trt
:3

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# torch2trt/misc_utils.py:3
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
A
Device string "cuda"
if device.type == torch.device("cuda").type:
:41

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
return torch.device("cuda")
:51

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

torch2trt/torch2trt.py· 1
C
TensorRT dependency
import tensorrt as trt
:2

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# torch2trt/torch2trt.py:2
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
torch2trt/trt_module.py· 3
C
TensorRT dependency
import tensorrt as trt
:2

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# torch2trt/trt_module.py:2
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
A
torch.cuda API usage
bindings, torch.cuda.current_stream().cuda_stream
:121

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
torch.cuda API usage
self.context.execute_async_v3(torch.cuda.current_stream().cuda_stream)
:154

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

torch2trt/utils.py· 1
C
TensorRT dependency
import tensorrt as trt
:2

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# torch2trt/utils.py:2
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
torch2trt/version_utils.py· 1
C
TensorRT dependency
import tensorrt as trt
:2

TensorRT is an NVIDIA-only inference compiler with no ROCm equivalent. Migration means re-targeting to MIGraphX / ONNX Runtime-ROCm and revalidating accuracy.

RecommendRe-implement the inference path on AMD MIGraphX or ORT-ROCm.

Suggested change · advisory
# torch2trt/version_utils.py:2
- import tensorrt as trt
+ import tensorrt as trt  # advisory: see ROCm note above
plugins/src/example_plugin.cu· 1
B
Custom CUDA kernels (.cu/.cuh)
CUDA kernel source
file

Hand-written .cu/.cuh kernels compile only with nvcc. AMD's HIPify tools translate the vast majority of CUDA C++ to HIP automatically.

RecommendRun hipify-perl over the .cu/.cuh sources and build with hipcc.

Suggested change · advisory
# plugins/src/example_plugin.cu:?
- CUDA kernel source
+ CUDA kernel source  # advisory: see ROCm note above
plugins/src/reflection_pad_2d_plugin.cu· 1
B
Custom CUDA kernels (.cu/.cuh)
CUDA kernel source
file

Hand-written .cu/.cuh kernels compile only with nvcc. AMD's HIPify tools translate the vast majority of CUDA C++ to HIP automatically.

RecommendRun hipify-perl over the .cu/.cuh sources and build with hipcc.

examples/contrib/quantization_aware_training/utils/utilities.py· 4
A
.cuda() tensor/module move
images = images.cuda()
:217

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
labels = labels.cuda()
:218

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
torch.cuda API usage
torch.cuda.synchronize()
:233

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
torch.cuda API usage
torch.cuda.synchronize()
:241

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

scripts/profile_timm.py· 8
A
torch.cuda API usage
torch.cuda.current_stream().synchronize()
:24

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
torch.cuda API usage
torch.cuda.current_stream().synchronize()
:28

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
torch.cuda API usage
torch.cuda.current_stream().synchronize()
:38

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
torch.cuda API usage
torch.cuda.current_stream().synchronize()
:42

Calls under torch.cuda.* are aliased by ROCm builds of PyTorch. The same code runs on AMD Instinct GPUs unchanged — torch.cuda.is_available(), streams, events and AMP all map onto HIP.

RecommendNo change required. Install the ROCm build of PyTorch (pip install torch --index-url https://download.pytorch.org/whl/rocm6.1).

A
.cuda() tensor/module move
model = model.cuda().eval()
:121

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(self.batch_size, 3, self.size, self.size).cuda()
:122

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(self.batch_size, 3, self.size, self.size).cuda()
:150

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(self.batch_size, 3, self.size, self.size).cuda()
:162

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/converter_tests/test_converters.py· 115
A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.leaky_relu(x)).cuda().eval()
:64

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4).cuda()]
:65

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.elu(x)).cuda().eval()
:71

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4).cuda()]
:72

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.selu(x)).cuda().eval()
:78

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4).cuda()]
:79

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.selu(x)).cuda().eval()
:85

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4).cuda()]
:86

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.softplus(x)).cuda().eval()
:92

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4).cuda()]
:93

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.adaptive_avg_pool2d(x, output_size)).cuda().eval()
:103

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4, 4).cuda()]
:104

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.adaptive_avg_pool3d(x, output_size)).cuda().eval()
:114

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4, 4, 4).cuda()]
:115

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.adaptive_max_pool2d(x, output_size)).cuda().eval()
:125

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4, 4).cuda()]
:126

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.adaptive_max_pool3d(x, output_size)).cuda().eval()
:136

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4, 4, 4).cuda()]
:137

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(lambda a, b: a + b).cuda().eval()
:142

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4).cuda(), torch.randn(1, 3, 4).cuda()]
:143

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(lambda a, b: torch.add(a, b)).cuda().eval()
:148

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4).cuda(), torch.randn(1, 3, 4).cuda()]
:149

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = IAdd().cuda().eval()
:162

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.ones(1, 3, 4).cuda(), torch.ones(1, 3, 4).cuda()]
:163

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: 1 + x).cuda().eval()
:168

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4).cuda()]
:169

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: 1.0 + x).cuda().eval()
:174

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4).cuda()]
:175

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = nn.Sequential(*modules).cuda().eval()
:199

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(*input_size).cuda()]
:203

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.cat([x, x], dim=dim)).cuda().eval()
:209

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 3).cuda()]
:210

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.chunk(x, chunks, dim)).cuda().eval()
:219

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 3).cuda()]
:220

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.split(x, split_sections_or_size, dim)).cuda().eval()
:230

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 3).cuda()]
:231

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.clamp(x, min, max)).cuda().eval()
:241

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 8, 8).cuda()]
:242

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: x.clone()).cuda().eval()
:247

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 8, 8).cuda()]
:248

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(lambda x, y: x > y).cuda().eval()
:253

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 4, 4).cuda(), torch.randn(1, 4, 4).cuda()]
:254

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: x > 0).cuda().eval()
:260

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 4, 4).cuda()]
:261

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(lambda x, y: x < y).cuda().eval()
:266

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 4, 4).cuda(), torch.randn(1, 4, 4).cuda()]
:267

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: x < 0).cuda().eval()
:273

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 4, 4).cuda()]
:274

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(lambda x, y: x == y).cuda().eval()
:278

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.zeros(1, 4, 4).cuda(), torch.zeros(1, 4, 4).cuda()]
:279

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: x == 0).cuda().eval()
:287

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.zeros(1, 4, 4).cuda()]
:288

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = cls(3, 3,
:322

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(*shape).cuda()]
:331

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = cls(3, 3,
:364

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(*shape).cuda()]
:373

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(lambda x, y: x / y).cuda().eval()
:378

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 4, 4).cuda(), torch.ones(1, 4, 4).cuda()*2]
:379

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: x / val).cuda().eval()
:386

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 4, 4).cuda()]
:387

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(fn).cuda().eval()
:398

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.ones(1, 4, 4).cuda()]
:399

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(lambda x, y: torch.einsum(einsum_expr, x, y)).cuda().eval()
:409

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 4).cuda(), torch.randn(1, 4, 5).cuda()]
:410

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: x.expand(*sizes)).cuda().eval()
:422

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(3, 1).cuda()]
:423

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.flatten(x, start_dim, end_dim)).cuda().eval()
:436

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 2, 3, 4, 5).cuda()]
:437

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(lambda x, y: x // y).cuda().eval()
:443

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.ones(1, 2, 3, 4, 5).cuda()]
:444

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: x // scalar).cuda().eval()
:455

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 2, 3).cuda()]
:456

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = nn.GELU().cuda().eval()
:461

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 2, 3).cuda()]
:462

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: expr(x)).cuda().eval()
:474

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 2, 3, 4).cuda()]
:475

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = nn.GroupNorm(num_groups, num_channels, affine=affine).cuda().eval()
:486

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, num_channels, 3, 4).cuda()]
:487

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(*shape).cuda()]
:497

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.interpolate(
:517

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(*input_size).cuda()]
:524

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = nn.LayerNorm(8).cuda().eval()
:530

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 4, 8).cuda()]
:532

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = nn.Linear(4, 8).cuda().eval()
:539

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(*input_shapes).cuda()]
:540

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = nn.LogSoftmax(dim).cuda().eval()
:546

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 2).cuda()]
:547

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(lambda x, y: torch.matmul(x, y)).cuda().eval()
:558

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(*shape_a).cuda(), torch.randn(*shape_b).cuda()]
:560

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = cls(kernel_size,stride,padding,dilation,ceil_mode=ceil_mode).cuda().eval()
:578

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(*input_size).cuda()]
:580

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = cls(kernel_size,stride,padding,ceil_mode=ceil_mode, count_include_pad=count_include_pad).cuda().eval()
:597

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(*input_size).cuda()]
:599

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(fn).cuda().eval()
:613

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 3).cuda(), torch.randn(1, 3, 3).cuda()]
:614

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(unary_1d_randn_ops[op]).cuda().eval()
:674

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 3).cuda()]
:675

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(unary_1d_positive_ops[op]).cuda().eval()
:696

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [0.5 * torch.ones(1, 3, 3).cuda()]
:697

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(op_map[op]).cuda().eval()
:711

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 3, 3).cuda()]
:713

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.narrow(x, dim, start, length)).cuda().eval()
:723

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(3, 3).cuda()]
:725

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = BinaryModule(lambda x, y: x != y).cuda().eval()
:730

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.zeros(1, 3, 3).cuda()]
:731

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(lambda x: torch.nn.functional.normalize(x, p, dim)).cuda().eval()
:741

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.zeros(1, 3, 3).cuda()]
:742

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(
:752

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(3, 3, 4, 2).cuda()]
:757

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(
:767

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(*sizes).cuda()]
:772

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = UnaryModule(fn).cuda().eval()
:796

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 2).cuda()]
:797

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = nn.PReLU(4).cuda().eval()
:802

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
inputs = [torch.randn(1, 4, 3, 3).cuda()]
:803

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/converter_tests/test_getitem.py· 16
A
.cuda() tensor/module move
module = YOLOXFocusTestModule().cuda().eval()
:48

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, 112, 112).cuda()
:50

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, 112, 112).cuda()
:54

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(4, 3, 112, 112).cuda()
:57

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:82

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, 112, 112).cuda()
:84

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, 112, 112).cuda()
:88

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(4, 3, 112, 112).cuda()
:91

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:116

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, 112, 112).cuda()
:118

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, 112, 112).cuda()
:122

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(4, 3, 112, 112).cuda()
:125

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:150

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, 112, 112).cuda()
:152

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, 112, 112).cuda()
:156

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(4, 3, 112, 112).cuda()
:159

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_contiguous.py· 2
A
.cuda() tensor/module move
net.eval().cuda()
:10

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
test_tensor = torch.randn((1, 25, 25, 3)).cuda().permute((0, 3, 1, 2))
:12

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_dataset.py· 10
A
.cuda() tensor/module move
self.conv = nn.Conv2d(3, 6, kernel_size=3, stride=1, padding=1).cuda().eval()
:54

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda(),
:62

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda()
:63

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:66

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
self.conv = nn.Conv2d(3, 6, kernel_size=3, stride=1, padding=1).cuda().eval()
:85

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda(),
:93

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda()
:94

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:97

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
self.conv = nn.Conv2d(3, 6, kernel_size=3, stride=1, padding=1).cuda().eval()
:116

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
Device string "cuda"
device = torch.device('cuda:0')
:123

The literal device string "cuda" resolves to the active AMD GPU on ROCm PyTorch. .to("cuda") / device="cuda" need no edits.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_flatten_dynamic.py· 12
A
.cuda() tensor/module move
module = FlattenModule(start_dim=0, end_dim=-1).cuda().eval()
:19

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 4, 5).cuda()
:21

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 4, 5).cuda()
:25

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 4, 5).cuda()
:28

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = FlattenModule(start_dim=1, end_dim=-1).cuda().eval()
:34

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 4, 5).cuda()
:36

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 4, 5).cuda()
:40

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 4, 5).cuda()
:43

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = FlattenModule(start_dim=0, end_dim=1).cuda().eval()
:49

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 4, 5).cuda()
:51

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 4, 5).cuda()
:55

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 4, 5).cuda()
:58

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_flatten_module.py· 4
A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:13

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda(),
:16

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda(),
:18

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
torch.randn(1, 3, 32, 32).cuda()
:19

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_interpolate_dynamic.py· 8
A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:17

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:19

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:25

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 3, 64, 64).cuda()
:28

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:39

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:41

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:47

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 3, 64, 64).cuda()
:50

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_legacy_max_batch_size.py· 8
A
.cuda() tensor/module move
model = nn.Conv2d(3, 6, kernel_size=1).cuda().eval()
:8

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, 32, 32).cuda()
:10

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, 32, 32).cuda()
:15

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(4, 3, 32, 32).cuda()
:22

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
model = nn.Conv1d(10, 20, kernel_size=1).cuda().eval()
:30

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 10, 32).cuda()
:32

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 10, 32).cuda()
:37

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(4, 10, 32).cuda()
:44

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_save_load.py· 2
A
.cuda() tensor/module move
model = torch.nn.Conv2d(3, 3, 1).cuda().eval().half()
:8

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn((1, 3, 224, 224)).cuda().half()
:9

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_tensor_ne.py· 3
A
.cuda() tensor/module move
module = NotEqual().cuda().eval()
:14

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 40, 20).cuda()
:16

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
y = torch.randn(1, 3, 1, 20).cuda()
:17

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_tensor_shape_div_batch.py· 2
A
.cuda() tensor/module move
module = DivConstantBatch().cuda().eval()
:15

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 10, 10).cuda()
:17

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/feature_tests/test_tensor_shape.py· 27
A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:19

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:21

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:27

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 3, 32, 32).cuda()
:30

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:41

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:43

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:47

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 3, 32, 32).cuda()
:50

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:61

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:63

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:67

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 3, 32, 32).cuda()
:70

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:81

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:83

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:87

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 3, 32, 32).cuda()
:90

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:101

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:103

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:107

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 3, 32, 32).cuda()
:110

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:121

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:123

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:127

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(4, 3, 32, 32).cuda()
:130

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
module = TestModule().cuda().eval()
:141

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:143

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
x = torch.randn(1, 3, 32, 32).cuda()
:147

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/model_tests/timm/test_maxvit.py· 4
A
.cuda() tensor/module move
model = model.cuda()
:12

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, *shape).cuda()
:13

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
_cross_validate_module(maxvit_tiny_rw_224().cuda().eval(), (224, 224))
:21

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
_cross_validate_module(maxvit_rmlp_small_rw_224().cuda().eval(), (224, 224))
:25

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/model_tests/torchvision/test_classification_models.py· 3
A
.cuda() tensor/module move
model = model.cuda().eval()
:7

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, *shape).cuda()
:8

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, *shape).cuda()
:10

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

tests/model_tests/torchvision/test_segmentation_models.py· 3
A
.cuda() tensor/module move
model = model.cuda().eval()
:15

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, *shape).cuda()
:16

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.

A
.cuda() tensor/module move
data = torch.randn(1, 3, *shape).cuda()
:18

.cuda() on tensors and modules is honoured by ROCm PyTorch and moves data to the AMD GPU. No change needed.

RecommendNo change required on ROCm PyTorch.